-
Notifications
You must be signed in to change notification settings - Fork 38
/
cuComplex.h
338 lines (300 loc) · 11.6 KB
/
cuComplex.h
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
/*
* Copyright 1993-2012 NVIDIA Corporation. All rights reserved.
*
* NOTICE TO LICENSEE:
*
* This source code and/or documentation ("Licensed Deliverables") are
* subject to NVIDIA intellectual property rights under U.S. and
* international Copyright laws.
*
* These Licensed Deliverables contained herein is PROPRIETARY and
* CONFIDENTIAL to NVIDIA and is being provided under the terms and
* conditions of a form of NVIDIA software license agreement by and
* between NVIDIA and Licensee ("License Agreement") or electronically
* accepted by Licensee. Notwithstanding any terms or conditions to
* the contrary in the License Agreement, reproduction or disclosure
* of the Licensed Deliverables to any third party without the express
* written consent of NVIDIA is prohibited.
*
* NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
* LICENSE AGREEMENT, NVIDIA MAKES NO REPRESENTATION ABOUT THE
* SUITABILITY OF THESE LICENSED DELIVERABLES FOR ANY PURPOSE. IT IS
* PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND.
* NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THESE LICENSED
* DELIVERABLES, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY,
* NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE.
* NOTWITHSTANDING ANY TERMS OR CONDITIONS TO THE CONTRARY IN THE
* LICENSE AGREEMENT, IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY
* SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, OR ANY
* DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
* WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
* ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE
* OF THESE LICENSED DELIVERABLES.
*
* U.S. Government End Users. These Licensed Deliverables are a
* "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT
* 1995), consisting of "commercial computer software" and "commercial
* computer software documentation" as such terms are used in 48
* C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government
* only as a commercial end item. Consistent with 48 C.F.R.12.212 and
* 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), all
* U.S. Government End Users acquire the Licensed Deliverables with
* only those rights set forth herein.
*
* Any use of the Licensed Deliverables in individual and commercial
* software must include, in the user documentation and internal
* comments to the code, the above Disclaimer and U.S. Government End
* Users Notice.
*/
#if !defined(CU_COMPLEX_H_)
#define CU_COMPLEX_H_
/* When trying to include C header file in C++ Code extern "C" is required
* But the Standard QNX headers already have ifdef extern in them when compiling C++ Code
* extern "C" cannot be nested
* Hence keep the header out of extern "C" block
*/
#include <math.h> /* import fabsf, sqrt */
#if defined(__cplusplus)
extern "C" {
#endif /* __cplusplus */
#include "vector_types.h"
typedef float2 cuFloatComplex;
__host__ __device__ static __inline__ float cuCrealf (cuFloatComplex x)
{
return x.x;
}
__host__ __device__ static __inline__ float cuCimagf (cuFloatComplex x)
{
return x.y;
}
__host__ __device__ static __inline__ cuFloatComplex make_cuFloatComplex
(float r, float i)
{
cuFloatComplex res;
res.x = r;
res.y = i;
return res;
}
__host__ __device__ static __inline__ cuFloatComplex cuConjf (cuFloatComplex x)
{
return make_cuFloatComplex (cuCrealf(x), -cuCimagf(x));
}
__host__ __device__ static __inline__ cuFloatComplex cuCaddf (cuFloatComplex x,
cuFloatComplex y)
{
return make_cuFloatComplex (cuCrealf(x) + cuCrealf(y),
cuCimagf(x) + cuCimagf(y));
}
__host__ __device__ static __inline__ cuFloatComplex cuCsubf (cuFloatComplex x,
cuFloatComplex y)
{
return make_cuFloatComplex (cuCrealf(x) - cuCrealf(y),
cuCimagf(x) - cuCimagf(y));
}
/* This implementation could suffer from intermediate overflow even though
* the final result would be in range. However, various implementations do
* not guard against this (presumably to avoid losing performance), so we
* don't do it either to stay competitive.
*/
__host__ __device__ static __inline__ cuFloatComplex cuCmulf (cuFloatComplex x,
cuFloatComplex y)
{
cuFloatComplex prod;
prod = make_cuFloatComplex ((cuCrealf(x) * cuCrealf(y)) -
(cuCimagf(x) * cuCimagf(y)),
(cuCrealf(x) * cuCimagf(y)) +
(cuCimagf(x) * cuCrealf(y)));
return prod;
}
/* This implementation guards against intermediate underflow and overflow
* by scaling. Such guarded implementations are usually the default for
* complex library implementations, with some also offering an unguarded,
* faster version.
*/
__host__ __device__ static __inline__ cuFloatComplex cuCdivf (cuFloatComplex x,
cuFloatComplex y)
{
cuFloatComplex quot;
float s = fabsf(cuCrealf(y)) + fabsf(cuCimagf(y));
float oos = 1.0f / s;
float ars = cuCrealf(x) * oos;
float ais = cuCimagf(x) * oos;
float brs = cuCrealf(y) * oos;
float bis = cuCimagf(y) * oos;
s = (brs * brs) + (bis * bis);
oos = 1.0f / s;
quot = make_cuFloatComplex (((ars * brs) + (ais * bis)) * oos,
((ais * brs) - (ars * bis)) * oos);
return quot;
}
/*
* We would like to call hypotf(), but it's not available on all platforms.
* This discrete implementation guards against intermediate underflow and
* overflow by scaling. Otherwise we would lose half the exponent range.
* There are various ways of doing guarded computation. For now chose the
* simplest and fastest solution, however this may suffer from inaccuracies
* if sqrt and division are not IEEE compliant.
*/
__host__ __device__ static __inline__ float cuCabsf (cuFloatComplex x)
{
float a = cuCrealf(x);
float b = cuCimagf(x);
float v, w, t;
a = fabsf(a);
b = fabsf(b);
if (a > b) {
v = a;
w = b;
} else {
v = b;
w = a;
}
t = w / v;
t = 1.0f + t * t;
t = v * sqrtf(t);
if ((v == 0.0f) || (v > 3.402823466e38f) || (w > 3.402823466e38f)) {
t = v + w;
}
return t;
}
/* Double precision */
typedef double2 cuDoubleComplex;
__host__ __device__ static __inline__ double cuCreal (cuDoubleComplex x)
{
return x.x;
}
__host__ __device__ static __inline__ double cuCimag (cuDoubleComplex x)
{
return x.y;
}
__host__ __device__ static __inline__ cuDoubleComplex make_cuDoubleComplex
(double r, double i)
{
cuDoubleComplex res;
res.x = r;
res.y = i;
return res;
}
__host__ __device__ static __inline__ cuDoubleComplex cuConj(cuDoubleComplex x)
{
return make_cuDoubleComplex (cuCreal(x), -cuCimag(x));
}
__host__ __device__ static __inline__ cuDoubleComplex cuCadd(cuDoubleComplex x,
cuDoubleComplex y)
{
return make_cuDoubleComplex (cuCreal(x) + cuCreal(y),
cuCimag(x) + cuCimag(y));
}
__host__ __device__ static __inline__ cuDoubleComplex cuCsub(cuDoubleComplex x,
cuDoubleComplex y)
{
return make_cuDoubleComplex (cuCreal(x) - cuCreal(y),
cuCimag(x) - cuCimag(y));
}
/* This implementation could suffer from intermediate overflow even though
* the final result would be in range. However, various implementations do
* not guard against this (presumably to avoid losing performance), so we
* don't do it either to stay competitive.
*/
__host__ __device__ static __inline__ cuDoubleComplex cuCmul(cuDoubleComplex x,
cuDoubleComplex y)
{
cuDoubleComplex prod;
prod = make_cuDoubleComplex ((cuCreal(x) * cuCreal(y)) -
(cuCimag(x) * cuCimag(y)),
(cuCreal(x) * cuCimag(y)) +
(cuCimag(x) * cuCreal(y)));
return prod;
}
/* This implementation guards against intermediate underflow and overflow
* by scaling. Such guarded implementations are usually the default for
* complex library implementations, with some also offering an unguarded,
* faster version.
*/
__host__ __device__ static __inline__ cuDoubleComplex cuCdiv(cuDoubleComplex x,
cuDoubleComplex y)
{
cuDoubleComplex quot;
double s = (fabs(cuCreal(y))) + (fabs(cuCimag(y)));
double oos = 1.0 / s;
double ars = cuCreal(x) * oos;
double ais = cuCimag(x) * oos;
double brs = cuCreal(y) * oos;
double bis = cuCimag(y) * oos;
s = (brs * brs) + (bis * bis);
oos = 1.0 / s;
quot = make_cuDoubleComplex (((ars * brs) + (ais * bis)) * oos,
((ais * brs) - (ars * bis)) * oos);
return quot;
}
/* This implementation guards against intermediate underflow and overflow
* by scaling. Otherwise we would lose half the exponent range. There are
* various ways of doing guarded computation. For now chose the simplest
* and fastest solution, however this may suffer from inaccuracies if sqrt
* and division are not IEEE compliant.
*/
__host__ __device__ static __inline__ double cuCabs (cuDoubleComplex x)
{
double a = cuCreal(x);
double b = cuCimag(x);
double v, w, t;
a = fabs(a);
b = fabs(b);
if (a > b) {
v = a;
w = b;
} else {
v = b;
w = a;
}
t = w / v;
t = 1.0 + t * t;
t = v * sqrt(t);
if ((v == 0.0) ||
(v > 1.79769313486231570e+308) || (w > 1.79769313486231570e+308)) {
t = v + w;
}
return t;
}
#if defined(__cplusplus)
}
#endif /* __cplusplus */
/* aliases */
typedef cuFloatComplex cuComplex;
__host__ __device__ static __inline__ cuComplex make_cuComplex (float x,
float y)
{
return make_cuFloatComplex (x, y);
}
/* float-to-double promotion */
__host__ __device__ static __inline__ cuDoubleComplex cuComplexFloatToDouble
(cuFloatComplex c)
{
return make_cuDoubleComplex ((double)cuCrealf(c), (double)cuCimagf(c));
}
__host__ __device__ static __inline__ cuFloatComplex cuComplexDoubleToFloat
(cuDoubleComplex c)
{
return make_cuFloatComplex ((float)cuCreal(c), (float)cuCimag(c));
}
__host__ __device__ static __inline__ cuComplex cuCfmaf( cuComplex x, cuComplex y, cuComplex d)
{
float real_res;
float imag_res;
real_res = (cuCrealf(x) * cuCrealf(y)) + cuCrealf(d);
imag_res = (cuCrealf(x) * cuCimagf(y)) + cuCimagf(d);
real_res = -(cuCimagf(x) * cuCimagf(y)) + real_res;
imag_res = (cuCimagf(x) * cuCrealf(y)) + imag_res;
return make_cuComplex(real_res, imag_res);
}
__host__ __device__ static __inline__ cuDoubleComplex cuCfma( cuDoubleComplex x, cuDoubleComplex y, cuDoubleComplex d)
{
double real_res;
double imag_res;
real_res = (cuCreal(x) * cuCreal(y)) + cuCreal(d);
imag_res = (cuCreal(x) * cuCimag(y)) + cuCimag(d);
real_res = -(cuCimag(x) * cuCimag(y)) + real_res;
imag_res = (cuCimag(x) * cuCreal(y)) + imag_res;
return make_cuDoubleComplex(real_res, imag_res);
}
#endif /* !defined(CU_COMPLEX_H_) */