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
|
/*
-- MAGMA (version 2.9.0) --
Univ. of Tennessee, Knoxville
Univ. of California, Berkeley
Univ. of Colorado, Denver
@date January 2025
@author Mark gates
@author Ahmad Abdelfattah
@generated from testing/testing_zgbsv_batched.cpp, normal z -> s, Wed Jan 22 14:40:43 2025
*/
// includes, system
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
// includes, project
#include "flops.h"
#include "magma_v2.h"
#include "magma_lapack.h"
#include "testings.h"
#if defined(_OPENMP)
#include <omp.h>
#include "../control/magma_threadsetting.h" // internal header
#endif
#define PRECISION_s
/* ////////////////////////////////////////////////////////////////////////////
-- Testing sgesv_batched
*/
int main(int argc, char **argv)
{
TESTING_CHECK( magma_init() );
magma_print_environment();
real_Double_t gflops, cpu_perf, cpu_time, gpu_perf, gpu_time = 0;
float error, Rnorm, Anorm, Xnorm, *work;
float c_one = MAGMA_S_ONE;
float c_neg_one = MAGMA_S_NEG_ONE;
float *h_A, *h_B, *h_X;
magmaFloat_ptr d_A, d_B;
magma_int_t *dipiv, *dinfo_array;
magma_int_t *ipiv, *cpu_info;
magma_int_t N, Nband, KL, KU, KV, nrhs, lda, ldb, ldda, lddb, info = 0, sizeA, sizeB;
magma_int_t ione = 1;
magma_int_t ISEED[4] = {0,0,0,1};
int status = 0;
magma_int_t batchCount;
float **dA_array = NULL;
float **dB_array = NULL;
magma_int_t **dipiv_array = NULL;
magma_opts opts( MagmaOptsBatched );
opts.parse_opts( argc, argv );
float tol = opts.tolerance * lapackf77_slamch("E");
nrhs = opts.nrhs;
KL = opts.kl;
KU = opts.ku;
KV = KL + KU;
batchCount = opts.batchcount;
printf("%% ## INFO ##: Gflop/s calculation is not available\n");
printf("%% Lower bandwidth (KL) = %lld\n", (long long)KL);
printf("%% Upper bandwidth (KU) = %lld\n", (long long)KU);
printf("%% BatchCount N NRHS CPU Gflop/s (ms) GPU Gflop/s (ms) ||B - AX|| / N*||A||*||X||\n");
printf("%%============================================================================================\n");
for( int itest = 0; itest < opts.ntest; ++itest ) {
for( int iter = 0; iter < opts.niter; ++iter ) {
N = opts.nsize[itest];
Nband = KL + 1 + KV; // need extra KL superdiagonals for the upper factor
lda = Nband;
ldb = N;
ldda = magma_roundup( lda, opts.align ); // multiple of 32 by default
lddb = magma_roundup( ldb, opts.align );
gflops = 0.; // TODO: gflop formula for gbsv?
sizeA = lda*N*batchCount;
sizeB = ldb*nrhs*batchCount;
TESTING_CHECK( magma_smalloc_cpu( &h_A, sizeA ));
TESTING_CHECK( magma_smalloc_cpu( &h_B, sizeB ));
TESTING_CHECK( magma_smalloc_cpu( &h_X, sizeB ));
TESTING_CHECK( magma_smalloc_cpu( &work, N ));
TESTING_CHECK( magma_imalloc_cpu( &ipiv, batchCount*N ));
TESTING_CHECK( magma_imalloc_cpu( &cpu_info, batchCount ));
TESTING_CHECK( magma_smalloc( &d_A, ldda*N*batchCount ));
TESTING_CHECK( magma_smalloc( &d_B, lddb*nrhs*batchCount ));
TESTING_CHECK( magma_imalloc( &dipiv, N * batchCount ));
TESTING_CHECK( magma_imalloc( &dinfo_array, batchCount ));
TESTING_CHECK( magma_malloc( (void**) &dA_array, batchCount * sizeof(float*) ));
TESTING_CHECK( magma_malloc( (void**) &dB_array, batchCount * sizeof(float*) ));
TESTING_CHECK( magma_malloc( (void**) &dipiv_array, batchCount * sizeof(magma_int_t*) ));
/* Initialize the matrices */
lapackf77_slarnv( &ione, ISEED, &sizeA, h_A );
lapackf77_slarnv( &ione, ISEED, &sizeB, h_B );
// random initialization of h_A seems to produce
// some matrices that are singular, the additive statements below
// seem to avoid that
#pragma omp parallel for schedule(dynamic)
for(int s = 0; s < batchCount; s++) {
float* hA = h_A + s*lda*N;
for(int j = 0; j < lda*N; j++) {
MAGMA_S_REAL( hA[j] ) += 20.;
#if defined(PRECISION_c) || defined(PRECISION_z)
MAGMA_S_IMAG( hA[j] ) += 20.;
#endif
}
}
magma_ssetmatrix( Nband, N*batchCount, h_A, lda, d_A, ldda, opts.queue );
magma_ssetmatrix( N, nrhs*batchCount, h_B, ldb, d_B, lddb, opts.queue );
/* ====================================================================
Performs operation using MAGMA
=================================================================== */
magma_sset_pointer( dA_array, d_A, ldda, 0, 0, ldda*N, batchCount, opts.queue );
magma_sset_pointer( dB_array, d_B, lddb, 0, 0, lddb*nrhs, batchCount, opts.queue );
magma_iset_pointer( dipiv_array, dipiv, 1, 0, 0, N, batchCount, opts.queue );
if(opts.version == 1) {
// synchronous api with ptr array
gpu_time = magma_sync_wtime( opts.queue );
info = magma_sgbsv_batched(
N, KL, KU, nrhs,
dA_array, ldda, dipiv_array,
dB_array, lddb, dinfo_array,
batchCount, opts.queue);
gpu_time = magma_sync_wtime( opts.queue ) - gpu_time;
}
else if(opts.version == 2) {
// synchronous api with stride
gpu_time = magma_sync_wtime( opts.queue );
info = magma_sgbsv_batched_strided(
N, KL, KU, nrhs,
d_A, ldda, ldda*N,
dipiv, N,
d_B, lddb, lddb*nrhs,
dinfo_array, batchCount, opts.queue);
gpu_time = magma_sync_wtime( opts.queue ) - gpu_time;
}
else if(opts.version == 3) {
// async api with ptr array
// query workspace
magma_int_t lwork[1] = {-1};
magma_sgbsv_batched_work(
N, KL, KU, nrhs,
NULL, ldda, NULL,
NULL, lddb,
NULL, NULL, lwork, batchCount, opts.queue);
// allocate workspace
void* device_work = NULL;
TESTING_CHECK( magma_malloc(&device_work, lwork[0]) );
// time the async interface only
gpu_time = magma_sync_wtime( opts.queue );
info = magma_sgbsv_batched_work(
N, KL, KU, nrhs,
dA_array, ldda, dipiv_array,
dB_array, lddb,
dinfo_array, device_work, lwork, batchCount, opts.queue);
gpu_time = magma_sync_wtime( opts.queue ) - gpu_time;
// free workspace
magma_free( device_work );
}
else if(opts.version == 4) {
// async api with stride
// query workspace
magma_int_t lwork[1] = {-1};
magma_sgbsv_batched_strided_work(
N, KL, KU, nrhs,
NULL, ldda, ldda*N, NULL, N,
NULL, lddb, lddb*nrhs, NULL, NULL, lwork,
batchCount, opts.queue);
// allocate workspace
void* device_work = NULL;
TESTING_CHECK( magma_malloc(&device_work, lwork[0]) );
// time the async interface only
gpu_time = magma_sync_wtime( opts.queue );
info = magma_sgbsv_batched_strided_work(
N, KL, KU, nrhs,
d_A, ldda, ldda*N, dipiv, N,
d_B, lddb, lddb*nrhs,
dinfo_array, device_work, lwork,
batchCount, opts.queue);
gpu_time = magma_sync_wtime( opts.queue ) - gpu_time;
// free workspace
magma_free( device_work );
}
gpu_perf = gflops / gpu_time;
// check correctness of results throught "dinfo_magma" and correctness of argument throught "info"
magma_getvector( batchCount, sizeof(magma_int_t), dinfo_array, 1, cpu_info, 1, opts.queue );
if (info != 0) {
printf("magma_sgbsv_batched returned argument error %lld: %s.\n",
(long long) info, magma_strerror( info ));
}
else {
for (int i=0; i < batchCount; i++) {
if (cpu_info[i] != 0 ) {
printf("magma_sgbsv_batched matrix %lld returned internal error %lld\n",
(long long) i, (long long) cpu_info[i] );
}
}
}
//=====================================================================
// Residual
//=====================================================================
magma_sgetmatrix( N, nrhs*batchCount, d_B, lddb, h_X, ldb, opts.queue );
error = 0;
for (magma_int_t s=0; s < batchCount; s++) {
float* hA = h_A + s * lda * N + KL;
float* hX = h_X + s * ldb * nrhs;
float* hB = h_B + s * ldb * nrhs;
Anorm = lapackf77_slangb("I", &N, &KL, &KU, hA, &lda, work);
Xnorm = lapackf77_slange("I", &N, &nrhs, hX, &ldb, work);
for(magma_int_t j = 0; j < nrhs; j++) {
blasf77_sgbmv( MagmaNoTransStr, &N, &N, &KL, &KU,
&c_one, hA , &lda,
hX + j * ldb, &ione,
&c_neg_one, hB + j * ldb, &ione);
}
Rnorm = lapackf77_slange("I", &N, &nrhs, hB, &ldb, work);
float err = Rnorm/(N*Anorm*Xnorm);
if (std::isnan(err) || std::isinf(err)) {
error = err;
break;
}
error = max( err, error );
}
bool okay = (error < tol);
status += ! okay;
/* ====================================================================
Performs operation using LAPACK
=================================================================== */
if ( opts.lapack ) {
cpu_time = magma_wtime();
// #define BATCHED_DISABLE_PARCPU
#if !defined (BATCHED_DISABLE_PARCPU) && defined(_OPENMP)
magma_int_t nthreads = magma_get_lapack_numthreads();
magma_set_lapack_numthreads(1);
magma_set_omp_numthreads(nthreads);
#pragma omp parallel for schedule(dynamic)
#endif
for (magma_int_t s=0; s < batchCount; s++)
{
magma_int_t locinfo;
lapackf77_sgbsv(
&N, &KL, &KU, &nrhs,
h_A + s*lda*N, &lda, ipiv + s*N,
h_B + s*ldb*nrhs, &ldb, &locinfo );
if (locinfo != 0) {
printf("lapackf77_sgesv matrix %lld returned error %lld: %s.\n",
(long long) s, (long long) locinfo, magma_strerror( locinfo ));
}
}
#if !defined (BATCHED_DISABLE_PARCPU) && defined(_OPENMP)
magma_set_lapack_numthreads(nthreads);
#endif
cpu_time = magma_wtime() - cpu_time;
cpu_perf = gflops / cpu_time;
printf( "%10lld %5lld %5lld %7.2f (%7.2f) %7.2f (%7.2f) %8.2e %s\n",
(long long) batchCount, (long long) N, (long long) nrhs,
cpu_perf, 1000.*cpu_time, gpu_perf, 1000.*gpu_time,
error, (okay ? "ok" : "failed"));
}
else {
printf( "%10lld %5lld %5lld --- ( --- ) %7.2f (%7.2f) %8.2e %s\n",
(long long) batchCount, (long long) N, (long long) nrhs,
gpu_perf, 1000.*gpu_time,
error, (okay ? "ok" : "failed"));
}
magma_free_cpu( h_A );
magma_free_cpu( h_B );
magma_free_cpu( h_X );
magma_free_cpu( work );
magma_free_cpu( ipiv );
magma_free_cpu( cpu_info );
magma_free( d_A );
magma_free( d_B );
magma_free( dipiv );
magma_free( dinfo_array );
magma_free( dA_array );
magma_free( dB_array );
magma_free( dipiv_array );
fflush( stdout );
}
if ( opts.niter > 1 ) {
printf( "\n" );
}
}
opts.cleanup();
TESTING_CHECK( magma_finalize() );
return status;
}
|
