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
|
/*
-- MAGMA (version 2.9.0) --
Univ. of Tennessee, Knoxville
Univ. of California, Berkeley
Univ. of Colorado, Denver
@date January 2025
@generated from testing/testing_zgebrd.cpp, normal z -> d, Wed Jan 22 14:40:40 2025
*/
// includes, system
#include <stdlib.h>
#include <stdio.h>
#include <string.h>
#include <math.h>
// includes, project
#include "flops.h"
#include "magma_v2.h"
#include "magma_lapack.h"
#include "testings.h"
#define REAL
/* ////////////////////////////////////////////////////////////////////////////
-- Testing dgebrd
*/
int main( int argc, char** argv)
{
TESTING_CHECK( magma_init() );
magma_print_environment();
real_Double_t gflops, gpu_perf, gpu_time, cpu_perf, cpu_time;
double *h_A, *h_Q, *h_PT, *h_work;
double *taup, *tauq;
double *diag, *offdiag;
double result[3] = {0., 0., 0.};
magma_int_t M, N, lda, lhwork, info, minmn, nb;
magma_int_t ione = 1;
int status = 0;
magma_opts opts;
opts.parse_opts( argc, argv );
double tol = opts.tolerance * lapackf77_dlamch("E");
double eps = lapackf77_dlamch( "E" );
printf("%% M N CPU Gflop/s (sec) GPU Gflop/s (sec) |A-QBP^H|/N|A| |I-QQ^H|/N |I-PP^H|/N\n");
printf("%%=============================================================================================\n");
for( int itest = 0; itest < opts.ntest; ++itest ) {
for( int iter = 0; iter < opts.niter; ++iter ) {
M = opts.msize[itest];
N = opts.nsize[itest];
minmn = min(M, N);
nb = magma_get_dgebrd_nb( M, N );
lda = M;
lhwork = (M + N)*nb;
gflops = FLOPS_DGEBRD( M, N ) / 1e9;
TESTING_CHECK( magma_dmalloc_cpu( &h_A, lda*N ));
TESTING_CHECK( magma_dmalloc_cpu( &tauq, minmn ));
TESTING_CHECK( magma_dmalloc_cpu( &taup, minmn ));
TESTING_CHECK( magma_dmalloc_cpu( &diag, minmn ));
TESTING_CHECK( magma_dmalloc_cpu( &offdiag, minmn-1 ));
TESTING_CHECK( magma_dmalloc_pinned( &h_Q, lda*N ));
TESTING_CHECK( magma_dmalloc_pinned( &h_work, lhwork ));
/* Initialize the matrices */
magma_generate_matrix( opts, M, N, h_A, lda );
lapackf77_dlacpy( MagmaFullStr, &M, &N, h_A, &lda, h_Q, &lda );
/* ====================================================================
Performs operation using MAGMA
=================================================================== */
gpu_time = magma_wtime();
magma_dgebrd( M, N, h_Q, lda,
diag, offdiag, tauq, taup,
h_work, lhwork, &info );
gpu_time = magma_wtime() - gpu_time;
gpu_perf = gflops / gpu_time;
if (info != 0) {
printf("magma_dgebrd returned error %lld: %s.\n",
(long long) info, magma_strerror( info ));
}
/* =====================================================================
Check the factorization
=================================================================== */
if ( opts.check ) {
// dorgbr prefers minmn*NB
// dbdt01 needs M+N
// dort01 prefers minmn*(minmn+1) to check Q and P
magma_int_t lwork_err;
double *h_work_err;
lwork_err = max( minmn * nb, M+N );
lwork_err = max( lwork_err, minmn*(minmn+1) );
TESTING_CHECK( magma_dmalloc_cpu( &h_PT, lda*N ));
TESTING_CHECK( magma_dmalloc_cpu( &h_work_err, lwork_err ));
// dbdt01 needs M
// dort01 needs minmn
#ifdef COMPLEX
double *rwork_err;
TESTING_CHECK( magma_dmalloc_cpu( &rwork_err, M ));
#endif
lapackf77_dlacpy( MagmaFullStr, &M, &N, h_Q, &lda, h_PT, &lda );
// generate Q & P^H
lapackf77_dorgbr("Q", &M, &minmn, &N, h_Q, &lda, tauq, h_work_err, &lwork_err, &info);
if (info != 0) {
printf("lapackf77_dorgbr #1 returned error %lld: %s.\n",
(long long) info, magma_strerror( info ));
}
lapackf77_dorgbr("P", &minmn, &N, &M, h_PT, &lda, taup, h_work_err, &lwork_err, &info);
if (info != 0) {
printf("lapackf77_dorgbr #2 returned error %lld: %s.\n",
(long long) info, magma_strerror( info ));
}
// Test 1: Check the decomposition A := Q * B * PT
// 2: Check the orthogonality of Q
// 3: Check the orthogonality of PT
lapackf77_dbdt01( &M, &N, &ione,
h_A, &lda, h_Q, &lda,
diag, offdiag, h_PT, &lda,
h_work_err,
#ifdef COMPLEX
rwork_err,
#endif
&result[0] );
// LAPACK normalizes by N*|A|, but that fails for very tall matrices,
// so normalize by max(M*N)*|A|. TODO: is there justification for that change?
result[0] = N*result[0] / max(M,N);
lapackf77_dort01( "Columns", &M, &minmn, h_Q, &lda, h_work_err, &lwork_err,
#ifdef COMPLEX
rwork_err,
#endif
&result[1]);
lapackf77_dort01( "Rows", &minmn, &N, h_PT, &lda, h_work_err, &lwork_err,
#ifdef COMPLEX
rwork_err,
#endif
&result[2]);
magma_free_cpu( h_PT );
magma_free_cpu( h_work_err );
#ifdef COMPLEX
magma_free_cpu( rwork_err );
#endif
// lapack normalizes by eps
result[0] *= eps;
result[1] *= eps;
result[2] *= eps;
}
/* =====================================================================
Performs operation using LAPACK
=================================================================== */
if ( opts.lapack ) {
cpu_time = magma_wtime();
lapackf77_dgebrd( &M, &N, h_A, &lda,
diag, offdiag, tauq, taup,
h_work, &lhwork, &info );
cpu_time = magma_wtime() - cpu_time;
cpu_perf = gflops / cpu_time;
if (info != 0) {
printf("lapackf77_dgebrd returned error %lld: %s.\n",
(long long) info, magma_strerror( info ));
}
}
/* =====================================================================
Print performance and error.
=================================================================== */
if ( opts.lapack ) {
printf("%5lld %5lld %7.2f (%7.2f) %7.2f (%7.2f)",
(long long) M, (long long) N, cpu_perf, cpu_time, gpu_perf, gpu_time );
}
else {
printf("%5lld %5lld --- ( --- ) %7.2f (%7.2f)",
(long long) M, (long long) N, gpu_perf, gpu_time );
}
if ( opts.check ) {
bool okay = (result[0] < tol) && (result[1] < tol) && (result[2] < tol);
status += ! okay;
printf(" %8.2e %8.2e %8.2e %s\n",
result[0], result[1], result[2],
(okay ? "ok" : "failed") );
} else {
printf(" --- --- \n");
}
magma_free_cpu( h_A );
magma_free_cpu( tauq );
magma_free_cpu( taup );
magma_free_cpu( diag );
magma_free_cpu( offdiag );
magma_free_pinned( h_Q );
magma_free_pinned( h_work );
fflush( stdout );
}
if ( opts.niter > 1 ) {
printf( "\n" );
}
}
opts.cleanup();
TESTING_CHECK( magma_finalize() );
return status;
}
|
