Eigen problem #116

pmollo · 2019-11-12T08:08:13Z

Hi,
I have to solve eigen value problems, so I first try to solve a validation example... But it does not work.

** Expected behavior **
I take a diagonal matrix (4x4) with 4 distinct integers on diagonal, then I use the "EigenValue" function following the help page (https://doc.freefem.org/models/eigen-value-problems.html).
So I expect to find 4 eigen values/vectors, but the function only returns the 3 highest ones.
I try to grow in dimension, but the function always returns the (n-1) highest eigen values.

The code for the example :
`matrix A =
[ [4, 0, 0, 0],
[0, 3 ,0 ,0],
[0, 0, 2, 0],
[0, 0, 0, 5] ];
set(A, solver=UMFPACK);

cout << A << endl;
matrix B =
[ [1, 0, 0, 0],
[0, 1 ,0 ,0],
[0, 0, 1, 0],
[0, 0, 0, 1] ];
set(B, solver=UMFPACK);

// Solve
int Nev = 4 ;
real[int] ev(Nev); //to store eigen values
real[int,int] Vec(Nev,Nev); //to store eigen vectors

int k = EigenValue(A, B, sym=true, value=ev, rawvector=Vec, nev=Nev, maxit=0, ncv=0);
cout << k ;
cout << " vp : " << ev << endl;
cout << " Vp : " << Vec << endl;
`

I try multiple parameters, but I still don't know why it is not working.

** Desktop : **

OS: Ubuntu 18.04
Freefem version : v4.4.2

Thank you for your help,
Pierre.

prj- · 2019-11-12T08:24:23Z

You should not use the EigenValue routine, it's based on an old interface to ARPACK.
Here are two different options:

using SLEPc (useful for sparse problems)
using LAPACK (useful for dense linear systems)

load "SLEPc"
matrix A =
[ [4, 0, 0, 0],
[0, 3 ,0 ,0],
[0, 0, 2, 0],
[0, 0, 0, 5] ];
matrix B =
[ [1, 0, 0, 0],
[0, 1 ,0 ,0],
[0, 0, 1, 0],
[0, 0, 0, 1] ];
Mat dA(A);
Mat dB(B);
real[int] nev(4);
real[int, int] neV(4, 4);
EPSSolve(dA, dB, values = nev, array = neV);
cout << nev << endl;

load "lapack"
real[int, int] fullA = [ [4, 0, 0, 0],
[0, 3 ,0 ,0],
[0, 0, 2, 0],
[0, 0, 0, 5] ];
real[int, int] fullB = [ [1, 0, 0, 0],
[0, 1 ,0 ,0],
[0, 0, 1, 0],
[0, 0, 0, 1] ];
complex[int] cnev(4);
complex[int, int] cneV(4, 4);
dggev(fullA, fullB, cnev, nev, cneV);
cout << cnev << endl;

You can run this program using, for example, mpirun -np 1 FreeFem++-mpi ev.edp -eps_view_values ::ascii_matlab -v 0. Expected output:

Lambda_EPS_0x84000000_0 = [
5.0000000000000027e+00
4.0000000000000044e+00
3.0000000000000027e+00
2.0000000000000000e+00
];
4
	  5	  4	  3	  2
4
	(4,0)	(3,0)	(2,0)	(5,0)

pmollo · 2019-11-12T13:24:20Z

Hi,
Thank you for your help !
It will be useful to update the help page about eigen values :)

prj- · 2019-11-12T13:26:48Z

There are some available informations at this URL. All SLEPc examples are available there. If you don't have any further issue, could you please close the issue when you get the chance to? Thanks.

frederichecht · 2019-11-12T14:24:23Z

Hi, the problem is an arpack, you can to compute all eigen value I think you can only n-2 eigen value no more.

…

Le 12 nov. 2019 à 09:08, Pmollo ***@***.***> a écrit : Hi, I have to solve eigen value problems, so I first try to solve a validation example... But it does not work. ** Expected behavior ** I take a diagonal matrix (4x4) with 4 distinct integers on diagonal, then I use the "EigenValue" function following the help page (https://doc.freefem.org/models/eigen-value-problems.html <https://doc.freefem.org/models/eigen-value-problems.html>). So I expect to find 4 eigen values/vectors, but the function only returns the 3 highest ones. I try to grow in dimension, but the function always returns the (n-1) highest eigen values. The code for the example : `matrix A = [ [4, 0, 0, 0], [0, 3 ,0 ,0], [0, 0, 2, 0], [0, 0, 0, 5] ]; set(A, solver=UMFPACK); cout << A << endl; matrix B = [ [1, 0, 0, 0], [0, 1 ,0 ,0], [0, 0, 1, 0], [0, 0, 0, 1] ]; set(B, solver=UMFPACK); // Solve int Nev = 4 ; real[int] ev(Nev); //to store eigen values real[int,int] Vec(Nev,Nev); //to store eigen vectors int k = EigenValue(A, B, sym=true, value=ev, rawvector=Vec, nev=Nev, maxit=0, ncv=0); cout << k ; cout << " vp : " << ev << endl; cout << " Vp : " << Vec << endl; ` I try multiple parameters, but I still don't know why it is not working. ** Desktop : ** OS: Ubuntu 18.04 Freefem version : v4.4.2 Thank you for your help, Pierre. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#116?email_source=notifications&email_token=ABLPNDB3H6CH4NKRXZLQZYDQTJP65A5CNFSM4JL7XS3KYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HYTZRVQ>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABLPNDEXF5HQWEEOSU4WJK3QTJP65ANCNFSM4JL7XS3A>.

pmollo · 2019-11-12T19:04:53Z

Hi again,
I still can't solve my eigen problems :

I'm unable to install/load SLEPc module. I try to reinstall FF and compile src codes, but it does not work.
LAPACK works & the function dggev finds e.v. but the eigen vectors are wrong (full of zeros or 1e-310). Beside that, the help page about LAPACK is empty (https://doc.freefem.org/references/external-libraries.html#dggev).

Thank you for your help
(Sorry for the re-opened/closed before)

prj- · 2019-11-12T19:06:12Z

What is the problem with SLEPc install? You can also call LAPACK directly from SLEPc. Here is what I get with -eps_view_vectors ::ascii_matlab -eps_type lapack.

%Vec Object: X0_EPS_0x84000000_0 1 MPI processes
%  type: seq
X0_EPS_0x84000000_0 = [
0.0000000000000000e+00
0.0000000000000000e+00
0.0000000000000000e+00
-1.0000000000000000e+00
];
%Vec Object: X1_EPS_0x84000000_0 1 MPI processes
%  type: seq
X1_EPS_0x84000000_0 = [
1.0000000000000000e+00
0.0000000000000000e+00
0.0000000000000000e+00
0.0000000000000000e+00
];
%Vec Object: X2_EPS_0x84000000_0 1 MPI processes
%  type: seq
X2_EPS_0x84000000_0 = [
0.0000000000000000e+00
1.0000000000000000e+00
0.0000000000000000e+00
0.0000000000000000e+00
];
%Vec Object: X3_EPS_0x84000000_0 1 MPI processes
%  type: seq
X3_EPS_0x84000000_0 = [
0.0000000000000000e+00
0.0000000000000000e+00
1.0000000000000000e+00
0.0000000000000000e+00
];

pmollo · 2019-11-12T19:38:46Z

This is what I've got when I try to compile PETSc :
"
===============================================================================
Configuring PETSc to compile on your system
===============================================================================
TESTING: checkCCompiler from config.setCompilers(config/BuildSystem/config/setCompilers.py:620) *******************************************************************************
UNABLE to CONFIGURE with GIVEN OPTIONS (see configure.log for details):

C compiler you provided with -with-cc= does not work.

Makefile:245: recipe for target 'petsc-3.12.0/tag-conf-real' failed
make: *** [petsc-3.12.0/tag-conf-real] Error 1
"

And this is what I've got when I use "mpirun -np 1 FreeFem++-mpi test.edp -eps_view_values ::ascii_matlab -eps_type lapack -ne" :

"
~/Documents$ mpirun -np 1 FreeFem++-mpi test.edp -eps_view_values ::ascii_matlab -eps_type lapack -ne
-- FreeFem++ v4.400002 (Fri Oct 25 11:37:05 CEST 2019 - git v4.4.2)
Load: lg_fem lg_mesh lg_mesh3 eigenvalue parallelempi
Add lapack interface ... sizestack + 1024 =1456 ( 432 )

4 4
(6.902534736e-310,0) (0,0) (0,0) (0,0)
(6.902534736e-310,0) (0,0) (0,0) (0,0)
(0,0) (0,0) (0,0) (0,0)
(0,0) (0,0) (0,0) (0,0)

times: compile 0.002496s, execution 0.000117s, mpirank:0
CodeAlloc : nb ptr 3460, size :462648 mpirank: 0
Ok: Normal End
"

And with "mpirun -np 1 FreeFem++-mpi test.edp -eps_view_vectors ::ascii_matlab -eps_type lapack -ne":
"
~/Documents$ mpirun -np 1 FreeFem++-mpi test.edp -eps_view_vectors ::ascii_matlab -eps_type lapack -ne
-- FreeFem++ v4.400002 (Fri Oct 25 11:37:05 CEST 2019 - git v4.4.2)
Load: lg_fem lg_mesh lg_mesh3 eigenvalue parallelempi
Add lapack interface ... sizestack + 1024 =1456 ( 432 )

4 4
(6.913082725e-310,0) (0,0) (0,0) (0,0)
(6.913082725e-310,0) (0,0) (0,0) (0,0)
(0,0) (0,0) (0,0) (0,0)
(0,0) (0,0) (0,0) (0,0)

times: compile 0.00245s, execution 0.00011s, mpirank:0
CodeAlloc : nb ptr 3460, size :462648 mpirank: 0
Ok: Normal End
"

The code is :
`
real[int] nev(4);
real[int, int] neV(4, 4);

load "lapack"
real[int, int] fullA = [ [4, 0, 0, 0],
[0, pi ,0 ,0],
[0, 0, 2, 0],
[0, 0, 0, 5] ];
real[int, int] fullB = [ [1, 0, 0, 0],
[0, 1 ,0 ,0],
[0, 0, 1, 0],
[0, 0, 0, 1] ];
complex[int] cnev(4);
complex[int, int] cneV(4, 4);
cneV=0.;
dggev(fullA, fullB, cnev, nev, cneV);
cout << cneV << endl;
`

prj- · 2019-11-12T19:42:14Z

FreeFEM configure does not find any MPI wrapper. Do you have MPI installed on your machine? Could you share your configure.log, please?

pmollo · 2019-11-12T20:03:02Z

Here my config.log file :
config.txt

prj- · 2019-11-12T20:23:20Z

Could you please follow this procedure and let me know if things still don't work out?

$ ./configure --enable-download --enable-optim --prefix=${PWD}
$ ./3rdparty/getall -a
$ cd 3rdparty/ff-petsc
$ make petsc-slepc
$ cd -
$ ./reconfigure
$ make

You are missing a couple of flags in your initial ./configure, so this may explain this issue. All this is detailed here, I've added the --prefix option because it is a bad idea to have to use sudo for installing FreeFEM.

pmollo · 2019-11-12T22:56:08Z

Ok, I think that something went wrong during my first attempt, but I tried in a VM and it's working.

Thank you again for your time & help

pmollo closed this as completed Nov 12, 2019

pmollo reopened this Nov 12, 2019

pmollo closed this as completed Nov 12, 2019

pmollo reopened this Nov 12, 2019

pmollo closed this as completed Nov 12, 2019

cirosantilli mentioned this issue Jun 9, 2020

v4.6 build "make petsc-slepc" fails with "C compiler you provided with -with-cc= cannot be found or does not work." on Ubuntu 20.04 #141

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Eigen problem #116

Eigen problem #116

pmollo commented Nov 12, 2019

prj- commented Nov 12, 2019 •

edited

Loading

pmollo commented Nov 12, 2019

prj- commented Nov 12, 2019 •

edited

Loading

frederichecht commented Nov 12, 2019 via email

pmollo commented Nov 12, 2019

prj- commented Nov 12, 2019 •

edited

Loading

pmollo commented Nov 12, 2019 •

edited

Loading

prj- commented Nov 12, 2019

pmollo commented Nov 12, 2019

prj- commented Nov 12, 2019 •

edited

Loading

pmollo commented Nov 12, 2019

Eigen problem #116

Eigen problem #116

Comments

pmollo commented Nov 12, 2019

prj- commented Nov 12, 2019 • edited Loading

pmollo commented Nov 12, 2019

prj- commented Nov 12, 2019 • edited Loading

frederichecht commented Nov 12, 2019 via email

pmollo commented Nov 12, 2019

prj- commented Nov 12, 2019 • edited Loading

pmollo commented Nov 12, 2019 • edited Loading

prj- commented Nov 12, 2019

pmollo commented Nov 12, 2019

prj- commented Nov 12, 2019 • edited Loading

pmollo commented Nov 12, 2019

prj- commented Nov 12, 2019 •

edited

Loading

prj- commented Nov 12, 2019 •

edited

Loading

prj- commented Nov 12, 2019 •

edited

Loading

pmollo commented Nov 12, 2019 •

edited

Loading

prj- commented Nov 12, 2019 •

edited

Loading