forked from NVIDIA/CUDALibrarySamples
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbicgstab_example.c
665 lines (642 loc) · 32.7 KB
/
bicgstab_example.c
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
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
/*
* Copyright 1993-2022 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.
*/
#include <cublas_v2.h>
#include <cuda_runtime.h>
#include <cusparse.h>
#include <stdio.h> // fopen
#include <stdlib.h> // EXIT_FAILURE
#include <string.h> // strtok
#define CHECK_CUDA(func) \
{ \
cudaError_t status = (func); \
if (status != cudaSuccess) { \
printf("CUDA API failed at line %d with error: %s (%d)\n", \
__LINE__, cudaGetErrorString(status), status); \
return EXIT_FAILURE; \
} \
}
#define CHECK_CUSPARSE(func) \
{ \
cusparseStatus_t status = (func); \
if (status != CUSPARSE_STATUS_SUCCESS) { \
printf("cuSPARSE API failed at line %d with error: %s (%d)\n", \
__LINE__, cusparseGetErrorString(status), status); \
return EXIT_FAILURE; \
} \
}
#define CHECK_CUBLAS(func) \
{ \
cublasStatus_t status = (func); \
if (status != CUBLAS_STATUS_SUCCESS) { \
printf("CUBLAS API failed at line %d with error: %d\n", \
__LINE__, status); \
return EXIT_FAILURE; \
} \
}
#if defined(NDEBUG)
# define PRINT_INFO(var)
#else
# define PRINT_INFO(var) printf(" " #var ": %f\n", var);
#endif
typedef struct VecStruct {
cusparseDnVecDescr_t vec;
double* ptr;
} Vec;
//==============================================================================
void mtx_header(const char* file_path,
int* num_lines,
int* num_rows,
int* num_cols,
int* nnz,
int* is_symmetric) {
char buffer[256];
FILE* file = fopen(file_path, "r");
if (file == NULL) {
printf("Error: unable to open the file %s\n", file_path);
exit(EXIT_FAILURE);
}
fgets(buffer, 256, file); // skip comments
char *token = strtok(buffer, " ");
if (strcmp(token, "%%MatrixMarket") != 0) {
printf("Unsupported file format. Only MTX format is supported");
exit(EXIT_FAILURE);
}
strtok(NULL, " "); // skip word
strtok(NULL, " "); // skip word
token = strtok(NULL, " "); // check data type
if (strcmp(token, "real") != 0) {
printf("Only real (double) matrices are supported");
exit(EXIT_FAILURE);
}
token = strtok(NULL, " \n"); // symmetric, unsymmetric
*is_symmetric = (strcmp(token, "symmetric") == 0);
while (fgetc(file) == '%')
fgets(buffer, 256, file); // skip % comments
fseek(file, -1, SEEK_CUR);
fscanf(file, "%d %d %d", num_rows, num_cols, num_lines);
*nnz = (*is_symmetric) ? *num_lines * 2 : *num_lines;
fclose(file);
}
void mtx_parsing(const char* file_path,
int num_lines,
int num_rows,
int nnz,
int* rows_offsets,
int* columns,
double* values,
int base) {
typedef struct IdxType {
int row, col;
double val;
} Idx;
int sort_by_row(const void *a, const void *b) {
return ((Idx*) a)->row - ((Idx*) b)->row;
}
char buffer[256];
FILE* file = fopen(file_path, "r");
while (fgetc(file) == '%')
fgets(buffer, 256, file); // skip comments
fgets(buffer, 256, file); // skip num row, cols, nnz
Idx* idx_tmp = (Idx*) malloc(nnz * sizeof(Idx));
for (int i = 0; i < num_lines; i++) {
int row, column;
double value;
fscanf(file, "%d %d %lf ", &row, &column, &value);
row -= (1 - base);
column -= (1 - base);
idx_tmp[i].row = row;
idx_tmp[i].col = column;
idx_tmp[i].val = value;
if (nnz != num_lines) { // is stored symmetric
idx_tmp[i + num_lines].row = column;
idx_tmp[i + num_lines].col = row;
idx_tmp[i + num_lines].val = value;
}
}
qsort(idx_tmp, nnz, sizeof(Idx), sort_by_row); // sort by row
memset(rows_offsets, 0x0, (num_rows + 1) * sizeof(int));
for (int i = 0; i < nnz; i++)
rows_offsets[idx_tmp[i].row + 1]++;
// prefix-scan
for (int i = 1; i <= num_rows; i++)
rows_offsets[i] = rows_offsets[i] + rows_offsets[i - 1];
for (int i = 0; i < nnz; i++) {
columns[i] = idx_tmp[i].col;
values[i] = idx_tmp[i].val;
}
fclose(file);
free(idx_tmp);
}
//==============================================================================
int gpu_BiCGStab(cublasHandle_t cublasHandle,
cusparseHandle_t cusparseHandle,
int m,
cusparseSpMatDescr_t matA,
cusparseSpMatDescr_t matM_lower,
cusparseSpMatDescr_t matM_upper,
Vec d_B,
Vec d_X,
Vec d_R0,
Vec d_R,
Vec d_P,
Vec d_P_aux,
Vec d_S,
Vec d_S_aux,
Vec d_V,
Vec d_T,
Vec d_tmp,
void* d_bufferMV,
int maxIterations,
double tolerance) {
const double zero = 0.0;
const double one = 1.0;
const double minus_one = -1.0;
//--------------------------------------------------------------------------
// Create opaque data structures that holds analysis data between calls
double coeff_tmp;
size_t bufferSizeL, bufferSizeU;
void* d_bufferL, *d_bufferU;
cusparseSpSVDescr_t spsvDescrL, spsvDescrU;
CHECK_CUSPARSE( cusparseSpSV_createDescr(&spsvDescrL) )
CHECK_CUSPARSE( cusparseSpSV_bufferSize(
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
&coeff_tmp, matM_lower, d_P.vec, d_tmp.vec, CUDA_R_64F,
CUSPARSE_SPSV_ALG_DEFAULT, spsvDescrL, &bufferSizeL) )
CHECK_CUDA( cudaMalloc(&d_bufferL, bufferSizeL) )
CHECK_CUSPARSE( cusparseSpSV_analysis(
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
&coeff_tmp, matM_lower, d_P.vec, d_tmp.vec, CUDA_R_64F,
CUSPARSE_SPSV_ALG_DEFAULT, spsvDescrL, d_bufferL) )
// Calculate UPPER buffersize
CHECK_CUSPARSE( cusparseSpSV_createDescr(&spsvDescrU) )
CHECK_CUSPARSE( cusparseSpSV_bufferSize(
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
&coeff_tmp, matM_upper, d_tmp.vec, d_P_aux.vec,
CUDA_R_64F, CUSPARSE_SPSV_ALG_DEFAULT, spsvDescrU,
&bufferSizeU) )
CHECK_CUDA( cudaMalloc(&d_bufferU, bufferSizeU) )
CHECK_CUSPARSE( cusparseSpSV_analysis(
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
&coeff_tmp, matM_upper, d_tmp.vec, d_P_aux.vec,
CUDA_R_64F, CUSPARSE_SPSV_ALG_DEFAULT, spsvDescrU,
d_bufferU) )
//--------------------------------------------------------------------------
// ### 1 ### R0 = b - A * X0 (using initial guess in X)
// (a) copy b in R0
CHECK_CUDA( cudaMemcpy(d_R0.ptr, d_B.ptr, m * sizeof(double),
cudaMemcpyDeviceToDevice) )
// (b) compute R = -A * X0 + R
CHECK_CUSPARSE( cusparseSpMV(cusparseHandle,
CUSPARSE_OPERATION_NON_TRANSPOSE,
&minus_one, matA, d_X.vec, &one, d_R0.vec,
CUDA_R_64F, CUSPARSE_SPMV_ALG_DEFAULT,
d_bufferMV) )
//--------------------------------------------------------------------------
double alpha, delta, delta_prev, omega;
CHECK_CUBLAS( cublasDdot(cublasHandle, m, d_R0.ptr, 1, d_R0.ptr, 1,
&delta) )
delta_prev = delta;
// R = R0
CHECK_CUDA( cudaMemcpy(d_R.ptr, d_R0.ptr, m * sizeof(double),
cudaMemcpyDeviceToDevice) )
//--------------------------------------------------------------------------
// nrm_R0 = ||R||
double nrm_R;
CHECK_CUBLAS( cublasDnrm2(cublasHandle, m, d_R0.ptr, 1, &nrm_R) )
double threshold = tolerance * nrm_R;
printf(" Initial Residual: Norm %e' threshold %e\n", nrm_R, threshold);
//--------------------------------------------------------------------------
// ### 2 ### repeat until convergence based on max iterations and
// and relative residual
for (int i = 1; i <= maxIterations; i++) {
printf(" Iteration = %d; Error Norm = %e\n", i, nrm_R);
//----------------------------------------------------------------------
// ### 4, 7 ### P_i = R_i
CHECK_CUDA( cudaMemcpy(d_P.ptr, d_R.ptr, m * sizeof(double),
cudaMemcpyDeviceToDevice) )
if (i > 1) {
//------------------------------------------------------------------
// ### 6 ### beta = (delta_i / delta_i-1) * (alpha / omega_i-1)
// (a) delta_i = (R'_0, R_i-1)
CHECK_CUBLAS( cublasDdot(cublasHandle, m, d_R0.ptr, 1, d_R.ptr, 1,
&delta) )
// (b) beta = (delta_i / delta_i-1) * (alpha / omega_i-1);
double beta = (delta / delta_prev) * (alpha / omega);
delta_prev = delta;
//------------------------------------------------------------------
// ### 7 ### P = R + beta * (P - omega * V)
// (a) P = - omega * V + P
double minus_omega = -omega;
CHECK_CUBLAS( cublasDaxpy(cublasHandle, m, &minus_omega, d_V.ptr, 1,
d_P.ptr, 1) )
// (b) P = beta * P
CHECK_CUBLAS( cublasDscal(cublasHandle, m, &beta, d_P.ptr, 1) )
// (c) P = R + P
CHECK_CUBLAS( cublasDaxpy(cublasHandle, m, &one, d_R.ptr, 1,
d_P.ptr, 1) )
}
//----------------------------------------------------------------------
// ### 9 ### P_aux = M_U^-1 M_L^-1 P_i
// (a) M_L^-1 P_i => tmp (triangular solver)
CHECK_CUDA( cudaMemset(d_tmp.ptr, 0x0, m * sizeof(double)) )
CHECK_CUDA( cudaMemset(d_P_aux.ptr, 0x0, m * sizeof(double)) )
CHECK_CUSPARSE( cusparseSpSV_solve(cusparseHandle,
CUSPARSE_OPERATION_NON_TRANSPOSE,
&one, matM_lower, d_P.vec, d_tmp.vec,
CUDA_R_64F,
CUSPARSE_SPSV_ALG_DEFAULT,
spsvDescrL) )
// (b) M_U^-1 tmp => P_aux (triangular solver)
CHECK_CUSPARSE( cusparseSpSV_solve(cusparseHandle,
CUSPARSE_OPERATION_NON_TRANSPOSE,
&one, matM_upper, d_tmp.vec,
d_P_aux.vec, CUDA_R_64F,
CUSPARSE_SPSV_ALG_DEFAULT,
spsvDescrU) )
//----------------------------------------------------------------------
// ### 10 ### alpha = (R'0, R_i-1) / (R'0, A * P_aux)
// (a) V = A * P_aux
CHECK_CUSPARSE( cusparseSpMV(cusparseHandle,
CUSPARSE_OPERATION_NON_TRANSPOSE, &one,
matA, d_P_aux.vec, &zero, d_V.vec,
CUDA_R_64F, CUSPARSE_SPMV_ALG_DEFAULT,
d_bufferMV) )
// (b) denominator = R'0 * V
double denominator;
CHECK_CUBLAS( cublasDdot(cublasHandle, m, d_R0.ptr, 1, d_V.ptr, 1,
&denominator) )
alpha = delta / denominator;
PRINT_INFO(delta)
PRINT_INFO(alpha)
//----------------------------------------------------------------------
// ### 11 ### X_i = X_i-1 + alpha * P_aux
CHECK_CUBLAS( cublasDaxpy(cublasHandle, m, &alpha, d_P_aux.ptr, 1,
d_X.ptr, 1) )
//----------------------------------------------------------------------
// ### 12 ### S = R_i-1 - alpha * (A * P_aux)
// (a) S = R_i-1
CHECK_CUDA( cudaMemcpy(d_S.ptr, d_R.ptr, m * sizeof(double),
cudaMemcpyDeviceToDevice) )
// (b) S = -alpha * V + R_i-1
double minus_alpha = -alpha;
CHECK_CUBLAS( cublasDaxpy(cublasHandle, m, &minus_alpha, d_V.ptr, 1,
d_S.ptr, 1) )
//----------------------------------------------------------------------
// ### 13 ### check ||S|| < threshold
double nrm_S;
CHECK_CUBLAS( cublasDnrm2(cublasHandle, m, d_S.ptr, 1, &nrm_S) )
PRINT_INFO(nrm_S)
if (nrm_S < threshold)
break;
//----------------------------------------------------------------------
// ### 14 ### S_aux = M_U^-1 M_L^-1 S
// (a) M_L^-1 S => tmp (triangular solver)
cudaMemset(d_tmp.ptr, 0x0, m * sizeof(double));
cudaMemset(d_S_aux.ptr, 0x0, m * sizeof(double));
CHECK_CUSPARSE( cusparseSpSV_solve(cusparseHandle,
CUSPARSE_OPERATION_NON_TRANSPOSE,
&one, matM_lower, d_S.vec, d_tmp.vec,
CUDA_R_64F,
CUSPARSE_SPSV_ALG_DEFAULT,
spsvDescrL) )
// (b) M_U^-1 tmp => S_aux (triangular solver)
CHECK_CUSPARSE( cusparseSpSV_solve(cusparseHandle,
CUSPARSE_OPERATION_NON_TRANSPOSE,
&one, matM_upper, d_tmp.vec,
d_S_aux.vec, CUDA_R_64F,
CUSPARSE_SPSV_ALG_DEFAULT,
spsvDescrU))
//----------------------------------------------------------------------
// ### 15 ### omega = (A * S_aux, s) / (A * S_aux, A * S_aux)
// (a) T = A * S_aux
CHECK_CUSPARSE( cusparseSpMV(cusparseHandle,
CUSPARSE_OPERATION_NON_TRANSPOSE, &one,
matA, d_S_aux.vec, &zero, d_T.vec,
CUDA_R_64F, CUSPARSE_SPMV_ALG_DEFAULT,
d_bufferMV) )
// (b) omega_num = (A * S_aux, s)
double omega_num, omega_den;
CHECK_CUBLAS( cublasDdot(cublasHandle, m, d_T.ptr, 1, d_S.ptr, 1,
&omega_num) )
// (c) omega_den = (A * S_aux, A * S_aux)
CHECK_CUBLAS( cublasDdot(cublasHandle, m, d_T.ptr, 1, d_T.ptr, 1,
&omega_den) )
// (d) omega = omega_num / omega_den
omega = omega_num / omega_den;
PRINT_INFO(omega)
// ---------------------------------------------------------------------
// ### 16 ### omega = X_i = X_i-1 + alpha * P_aux + omega * S_aux
// (a) X_i has been updated with h = X_i-1 + alpha * P_aux
// X_i = omega * S_aux + X_i
CHECK_CUBLAS( cublasDaxpy(cublasHandle, m, &omega, d_S_aux.ptr, 1,
d_X.ptr, 1) )
// ---------------------------------------------------------------------
// ### 17 ### R_i+1 = S - omega * (A * S_aux)
// (a) copy S in R
CHECK_CUDA( cudaMemcpy(d_R.ptr, d_S.ptr, m * sizeof(double),
cudaMemcpyDeviceToDevice) )
// (a) R_i+1 = -omega * T + R
double minus_omega = -omega;
CHECK_CUBLAS( cublasDaxpy(cublasHandle, m, &minus_omega, d_T.ptr, 1,
d_R.ptr, 1) )
// ---------------------------------------------------------------------
// ### 18 ### check ||R_i|| < threshold
CHECK_CUBLAS( cublasDnrm2(cublasHandle, m, d_R.ptr, 1, &nrm_R) )
PRINT_INFO(nrm_R)
if (nrm_R < threshold)
break;
}
//--------------------------------------------------------------------------
printf("Check Solution\n"); // ||R = b - A * X||
// (a) copy b in R
CHECK_CUDA( cudaMemcpy(d_R.ptr, d_B.ptr, m * sizeof(double),
cudaMemcpyDeviceToDevice) )
// R = -A * X + R
CHECK_CUSPARSE( cusparseSpMV(cusparseHandle,
CUSPARSE_OPERATION_NON_TRANSPOSE, &minus_one,
matA, d_X.vec, &one, d_R.vec, CUDA_R_64F,
CUSPARSE_SPMV_ALG_DEFAULT, d_bufferMV) )
// check ||R||
CHECK_CUBLAS( cublasDnrm2(cublasHandle, m, d_R.ptr, 1, &nrm_R) )
printf("Final error norm = %e\n", nrm_R);
//--------------------------------------------------------------------------
CHECK_CUSPARSE( cusparseSpSV_destroyDescr(spsvDescrL) )
CHECK_CUSPARSE( cusparseSpSV_destroyDescr(spsvDescrU) )
CHECK_CUDA( cudaFree(d_bufferL) )
CHECK_CUDA( cudaFree(d_bufferU) )
return EXIT_SUCCESS;
}
//==============================================================================
//==============================================================================
int main(int argc, char** argv) {
const int maxIterations = 20;
const double tolerance = 0.0000000001;
printf("Usage: bicgstab_example <matrix.mtx>\n");
if (argc != 2) {
printf("Wrong parameter: bicgstab_example <matrix.mtx>\n");
return EXIT_FAILURE;
}
int base = 0;
int num_rows, num_cols, nnz, num_lines, is_symmetric;
mtx_header(argv[1], &num_lines, &num_rows, &num_cols, &nnz, &is_symmetric);
printf("\nmatrix name: %s\n"
"num. rows: %d\n"
"num. cols: %d\n"
"nnz: %d\n"
"structure: %s\n\n",
argv[1], num_rows, num_cols, nnz,
(is_symmetric) ? "symmetric" : "unsymmetric");
if (num_rows != num_cols) {
printf("the input matrix must be square\n");
return EXIT_FAILURE;
}
if (!is_symmetric) {
printf("the input matrix must be symmetric\n");
return EXIT_FAILURE;
}
int m = num_rows;
int num_offsets = m + 1;
int* h_A_rows = (int*) malloc(num_offsets * sizeof(int));
int* h_A_columns = (int*) malloc(nnz * sizeof(int));
double* h_A_values = (double*) malloc(nnz * sizeof(double));
double* h_X = (double*) malloc(m * sizeof(double));
printf("Matrix parsing...\n");
mtx_parsing(argv[1], num_lines, num_rows, nnz, h_A_rows,
h_A_columns, h_A_values, base);
printf("Testing BiCGStab\n");
for (int i = 0; i < num_rows; i++)
h_X[i] = 1.0;
//--------------------------------------------------------------------------
// ### Device memory management ###
int* d_A_rows, *d_A_columns;
double* d_A_values, *d_M_values;
Vec d_B, d_X, d_R, d_R0, d_P, d_P_aux, d_S, d_S_aux, d_V, d_T, d_tmp;
// allocate device memory for CSR matrices
CHECK_CUDA( cudaMalloc((void**) &d_A_rows, num_offsets * sizeof(int)) )
CHECK_CUDA( cudaMalloc((void**) &d_A_columns, nnz * sizeof(int)) )
CHECK_CUDA( cudaMalloc((void**) &d_A_values, nnz * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_M_values, nnz * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_B.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_X.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_R.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_R0.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_P.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_P_aux.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_S.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_S_aux.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_V.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_T.ptr, m * sizeof(double)) )
CHECK_CUDA( cudaMalloc((void**) &d_tmp.ptr, m * sizeof(double)) )
// copy the CSR matrices and vectors into device memory
CHECK_CUDA( cudaMemcpy(d_A_rows, h_A_rows, num_offsets * sizeof(int),
cudaMemcpyHostToDevice) )
CHECK_CUDA( cudaMemcpy(d_A_columns, h_A_columns, nnz * sizeof(int),
cudaMemcpyHostToDevice) )
CHECK_CUDA( cudaMemcpy(d_A_values, h_A_values, nnz * sizeof(double),
cudaMemcpyHostToDevice) )
CHECK_CUDA( cudaMemcpy(d_M_values, h_A_values, nnz * sizeof(double),
cudaMemcpyHostToDevice) )
CHECK_CUDA( cudaMemcpy(d_X.ptr, h_X, m * sizeof(double),
cudaMemcpyHostToDevice) )
//--------------------------------------------------------------------------
// ### cuSPARSE Handle and descriptors initialization ###
// create the test matrix on the host
cublasHandle_t cublasHandle = NULL;
cusparseHandle_t cusparseHandle = NULL;
CHECK_CUBLAS( cublasCreate(&cublasHandle) )
CHECK_CUSPARSE( cusparseCreate(&cusparseHandle) )
// Create dense vectors
CHECK_CUSPARSE( cusparseCreateDnVec(&d_B.vec, m, d_B.ptr, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_X.vec, m, d_X.ptr, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_R.vec, m, d_R.ptr, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_R0.vec, m, d_R0.ptr, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_P.vec, m, d_P.ptr, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_P_aux.vec, m, d_P_aux.ptr,
CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_S.vec, m, d_S.ptr, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_S_aux.vec, m, d_S_aux.ptr,
CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_V.vec, m, d_V.ptr, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_T.vec, m, d_T.ptr, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseCreateDnVec(&d_tmp.vec, m, d_tmp.ptr, CUDA_R_64F) )
cusparseIndexBase_t baseIdx = CUSPARSE_INDEX_BASE_ZERO;
// IMPORTANT: Upper/Lower triangular decompositions of A
// (matM_lower, matM_upper) must use two distinct descriptors
cusparseSpMatDescr_t matA, matM_lower, matM_upper;
cusparseMatDescr_t matLU;
int* d_M_rows = d_A_rows;
int* d_M_columns = d_A_columns;
cusparseFillMode_t fill_lower = CUSPARSE_FILL_MODE_LOWER;
cusparseDiagType_t diag_unit = CUSPARSE_DIAG_TYPE_UNIT;
cusparseFillMode_t fill_upper = CUSPARSE_FILL_MODE_UPPER;
cusparseDiagType_t diag_non_unit = CUSPARSE_DIAG_TYPE_NON_UNIT;
// A
CHECK_CUSPARSE( cusparseCreateCsr(&matA, m, m, nnz, d_A_rows,
d_A_columns, d_A_values,
CUSPARSE_INDEX_32I, CUSPARSE_INDEX_32I,
baseIdx, CUDA_R_64F) )
// M_lower
CHECK_CUSPARSE( cusparseCreateCsr(&matM_lower, m, m, nnz, d_M_rows,
d_M_columns, d_M_values,
CUSPARSE_INDEX_32I, CUSPARSE_INDEX_32I,
baseIdx, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseSpMatSetAttribute(matM_lower,
CUSPARSE_SPMAT_FILL_MODE,
&fill_lower, sizeof(fill_lower)) )
CHECK_CUSPARSE( cusparseSpMatSetAttribute(matM_lower,
CUSPARSE_SPMAT_DIAG_TYPE,
&diag_unit, sizeof(diag_unit)) )
// M_upper
CHECK_CUSPARSE( cusparseCreateCsr(&matM_upper, m, m, nnz, d_M_rows,
d_M_columns, d_M_values,
CUSPARSE_INDEX_32I, CUSPARSE_INDEX_32I,
baseIdx, CUDA_R_64F) )
CHECK_CUSPARSE( cusparseSpMatSetAttribute(matM_upper,
CUSPARSE_SPMAT_FILL_MODE,
&fill_upper, sizeof(fill_upper)) )
CHECK_CUSPARSE( cusparseSpMatSetAttribute(matM_upper,
CUSPARSE_SPMAT_DIAG_TYPE,
&diag_non_unit,
sizeof(diag_non_unit)) )
//--------------------------------------------------------------------------
// ### Preparation ### b = A * X
const double alpha = 0.75;
size_t bufferSizeMV;
void* d_bufferMV;
double beta = 0.0;
CHECK_CUSPARSE( cusparseSpMV_bufferSize(
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
&alpha, matA, d_X.vec, &beta, d_B.vec, CUDA_R_64F,
CUSPARSE_SPMV_ALG_DEFAULT, &bufferSizeMV) )
CHECK_CUDA( cudaMalloc(&d_bufferMV, bufferSizeMV) )
CHECK_CUSPARSE( cusparseSpMV(
cusparseHandle, CUSPARSE_OPERATION_NON_TRANSPOSE,
&alpha, matA, d_X.vec, &beta, d_B.vec, CUDA_R_64F,
CUSPARSE_SPMV_ALG_DEFAULT, d_bufferMV) )
// X0 = 0
CHECK_CUDA( cudaMemset(d_X.ptr, 0x0, m * sizeof(double)) )
//--------------------------------------------------------------------------
// Perform Incomplete-LU factorization of A (csrilu0) -> M_lower, M_upper
csrilu02Info_t infoM = NULL;
int bufferSizeLU = 0;
void* d_bufferLU;
CHECK_CUSPARSE( cusparseCreateMatDescr(&matLU) )
CHECK_CUSPARSE( cusparseSetMatType(matLU, CUSPARSE_MATRIX_TYPE_GENERAL) )
CHECK_CUSPARSE( cusparseSetMatIndexBase(matLU, baseIdx) )
CHECK_CUSPARSE( cusparseCreateCsrilu02Info(&infoM) )
CHECK_CUSPARSE( cusparseDcsrilu02_bufferSize(
cusparseHandle, m, nnz, matLU, d_M_values,
d_A_rows, d_A_columns, infoM, &bufferSizeLU) )
CHECK_CUDA( cudaMalloc(&d_bufferLU, bufferSizeLU) )
CHECK_CUSPARSE( cusparseDcsrilu02_analysis(
cusparseHandle, m, nnz, matLU, d_M_values,
d_A_rows, d_A_columns, infoM,
CUSPARSE_SOLVE_POLICY_USE_LEVEL, d_bufferLU) )
int structural_zero;
CHECK_CUSPARSE( cusparseXcsrilu02_zeroPivot(cusparseHandle, infoM,
&structural_zero) )
// M = L * U
CHECK_CUSPARSE( cusparseDcsrilu02(
cusparseHandle, m, nnz, matLU, d_M_values,
d_A_rows, d_A_columns, infoM,
CUSPARSE_SOLVE_POLICY_USE_LEVEL, d_bufferLU) )
// Find numerical zero
int numerical_zero;
CHECK_CUSPARSE( cusparseXcsrilu02_zeroPivot(cusparseHandle, infoM,
&numerical_zero) )
CHECK_CUSPARSE( cusparseDestroyCsrilu02Info(infoM) )
CHECK_CUSPARSE( cusparseDestroyMatDescr(matLU) )
CHECK_CUDA( cudaFree(d_bufferLU) )
//--------------------------------------------------------------------------
// ### Run BiCGStab computation ###
printf("BiCGStab loop:\n");
gpu_BiCGStab(cublasHandle, cusparseHandle, m,
matA, matM_lower, matM_upper,
d_B, d_X, d_R0, d_R, d_P, d_P_aux, d_S, d_S_aux, d_V, d_T,
d_tmp, d_bufferMV, maxIterations, tolerance);
//--------------------------------------------------------------------------
// ### Free resources ###
CHECK_CUSPARSE( cusparseDestroyDnVec(d_B.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_X.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_R.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_R0.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_P.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_P_aux.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_S.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_S_aux.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_V.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_T.vec) )
CHECK_CUSPARSE( cusparseDestroyDnVec(d_tmp.vec) )
CHECK_CUSPARSE( cusparseDestroySpMat(matA) )
CHECK_CUSPARSE( cusparseDestroySpMat(matM_lower) )
CHECK_CUSPARSE( cusparseDestroySpMat(matM_upper) )
CHECK_CUSPARSE( cusparseDestroy(cusparseHandle) )
CHECK_CUBLAS( cublasDestroy(cublasHandle) )
free(h_A_rows);
free(h_A_columns);
free(h_A_values);
free(h_X);
CHECK_CUDA( cudaFree(d_X.ptr) )
CHECK_CUDA( cudaFree(d_B.ptr) )
CHECK_CUDA( cudaFree(d_R.ptr) )
CHECK_CUDA( cudaFree(d_R0.ptr) )
CHECK_CUDA( cudaFree(d_P.ptr) )
CHECK_CUDA( cudaFree(d_P_aux.ptr) )
CHECK_CUDA( cudaFree(d_S.ptr) )
CHECK_CUDA( cudaFree(d_S_aux.ptr) )
CHECK_CUDA( cudaFree(d_V.ptr) )
CHECK_CUDA( cudaFree(d_T.ptr) )
CHECK_CUDA( cudaFree(d_tmp.ptr) )
CHECK_CUDA( cudaFree(d_A_values) )
CHECK_CUDA( cudaFree(d_A_columns) )
CHECK_CUDA( cudaFree(d_A_rows) )
CHECK_CUDA( cudaFree(d_M_values) )
CHECK_CUDA( cudaFree(d_bufferMV) )
return EXIT_SUCCESS;
}