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
|
/*
-- MAGMA (version 2.9.0) --
Univ. of Tennessee, Knoxville
Univ. of California, Berkeley
Univ. of Colorado, Denver
@date January 2025
@author Stan Tomov
@author Mark Gates
@generated from testing/testing_zgeqr2x_gpu.cpp, normal z -> s, Wed Jan 22 14:40:32 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 "magma_operators.h"
#include "testings.h"
/* ////////////////////////////////////////////////////////////////////////////
-- Testing sgeqrf
*/
int main( int argc, char** argv)
{
TESTING_CHECK( magma_init() );
magma_print_environment();
/* Constants */
float c_zero = MAGMA_S_ZERO;
float c_neg_one = MAGMA_S_NEG_ONE;
float c_one = MAGMA_S_ONE;
float d_one = MAGMA_D_ONE;
float d_neg_one = MAGMA_D_NEG_ONE;
/* Local variables */
real_Double_t gflops, gpu_perf, gpu_time, cpu_perf, cpu_time;
float Anorm, error, error2, diff, terr, rwork[1];
float *h_A, *h_T, *h_R, *tau, *h_work, tmp[1], unused[1];
magmaFloat_ptr d_A, d_T, ddA, dtau;
magmaFloat_ptr dwork;
magma_int_t M, N, lda, ldda, lwork, size, info, min_mn;
magma_int_t ione = 1;
int status = 0;
#define BLOCK_SIZE 64
magma_opts opts;
opts.parse_opts( argc, argv );
float tol = opts.tolerance * lapackf77_slamch("E");
printf("%% version %lld\n", (long long) opts.version );
printf("%% It's okay if |Q - Q_magma| is large; MAGMA and LAPACK\n"
"%% just chose different Householder reflectors, both valid.\n\n");
printf("%% M N CPU Gflop/s (ms) GPU Gflop/s (ms) |R - Q^H*A| |I - Q^H*Q| |T - T_magma| |Q - Q_magma|\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];
if (N > 128) {
printf("%5lld %5lld skipping because sgeqr2x requires N <= 128\n",
(long long) M, (long long) N);
continue;
}
if (M < N) {
printf("%5lld %5lld skipping because sgeqr2x requires M >= N\n",
(long long) M, (long long) N);
continue;
}
min_mn = min( M, N );
lda = M;
ldda = magma_roundup( M, opts.align ); // multiple of 32 by default
// TODO: flops should be GEQRF + LARFT (whatever that is)
gflops = (FLOPS_SGEQRF( M, N ) + FLOPS_SGEQRT( M, N )) / 1e9;
lwork = -1;
lapackf77_sgeqrf( &M, &N, unused, &M, unused, tmp, &lwork, &info );
lwork = (magma_int_t)MAGMA_S_REAL( tmp[0] );
lwork = max( lwork, N*N );
/* Allocate memory for the matrix */
TESTING_CHECK( magma_smalloc_cpu( &tau, min_mn ));
TESTING_CHECK( magma_smalloc_cpu( &h_A, lda*N ));
TESTING_CHECK( magma_smalloc_cpu( &h_T, N*N ));
TESTING_CHECK( magma_smalloc_cpu( &h_work, lwork ));
TESTING_CHECK( magma_smalloc_pinned( &h_R, lda*N ));
TESTING_CHECK( magma_smalloc( &d_A, ldda*N ));
TESTING_CHECK( magma_smalloc( &d_T, N*N ));
TESTING_CHECK( magma_smalloc( &ddA, N*N ));
TESTING_CHECK( magma_smalloc( &dtau, min_mn ));
TESTING_CHECK( magma_smalloc( &dwork, max(5*min_mn, (BLOCK_SIZE*2+2)*min_mn) ));
magmablas_slaset( MagmaFull, N, N, c_zero, c_zero, ddA, N, opts.queue );
magmablas_slaset( MagmaFull, N, N, c_zero, c_zero, d_T, N, opts.queue );
/* Initialize the matrix */
magma_generate_matrix( opts, M, N, h_A, lda );
lapackf77_slacpy( MagmaFullStr, &M, &N, h_A, &lda, h_R, &lda );
magma_ssetmatrix( M, N, h_R, lda, d_A, ldda, opts.queue );
/* ====================================================================
Performs operation using MAGMA
=================================================================== */
gpu_time = magma_sync_wtime( opts.queue );
if (opts.version == 1) {
magma_sgeqr2x_gpu( M, N, d_A, ldda, dtau, d_T, ddA, dwork, &info );
}
else if (opts.version == 2) {
magma_sgeqr2x2_gpu( M, N, d_A, ldda, dtau, d_T, ddA, dwork, &info );
}
else if (opts.version == 3) {
magma_sgeqr2x3_gpu( M, N, d_A, ldda, dtau, d_T, ddA, dwork, &info );
}
else {
/*
Going through NULL stream is faster
Going through any stream is slower
Doing two streams in parallel is slower than doing them sequentially
Queuing happens on the NULL stream - user defined buffers are smaller?
*/
magma_sgeqr2x4_gpu( M, N, d_A, ldda, dtau, d_T, ddA, dwork, opts.queue, &info );
}
gpu_time = magma_sync_wtime( opts.queue ) - gpu_time;
gpu_perf = gflops / gpu_time;
if (info != 0) {
printf("magma_sgeqr2x_gpu version %lld returned error %lld: %s.\n",
(long long) opts.version, (long long) info, magma_strerror( info ));
}
else if ( opts.check ) {
/* =====================================================================
Check the result, following zqrt01 except using the reduced Q.
This works for any M,N (square, tall, wide).
=================================================================== */
magma_sgetmatrix( M, N, d_A, ldda, h_R, lda, opts.queue );
magma_sgetmatrix( N, N, ddA, N, h_T, N, opts.queue );
magma_sgetmatrix( min_mn, 1, dtau, min_mn, tau, min_mn, opts.queue );
// Restore the upper triangular part of A before the check
lapackf77_slacpy( "Upper", &N, &N, h_T, &N, h_R, &lda );
magma_int_t ldq = M;
magma_int_t ldr = min_mn;
float *Q, *R;
float *work;
TESTING_CHECK( magma_smalloc_cpu( &Q, ldq*min_mn )); // M by K
TESTING_CHECK( magma_smalloc_cpu( &R, ldr*N )); // K by N
TESTING_CHECK( magma_smalloc_cpu( &work, min_mn ));
// generate M by K matrix Q, where K = min(M,N)
lapackf77_slacpy( "Lower", &M, &min_mn, h_R, &lda, Q, &ldq );
lapackf77_sorgqr( &M, &min_mn, &min_mn, Q, &ldq, tau, h_work, &lwork, &info );
assert( info == 0 );
// copy K by N matrix R
lapackf77_slaset( "Lower", &min_mn, &N, &c_zero, &c_zero, R, &ldr );
lapackf77_slacpy( "Upper", &min_mn, &N, h_R, &lda, R, &ldr );
// error = || R - Q^H*A || / (N * ||A||)
blasf77_sgemm( "Conj", "NoTrans", &min_mn, &N, &M,
&c_neg_one, Q, &ldq, h_A, &lda, &c_one, R, &ldr );
Anorm = lapackf77_slange( "1", &M, &N, h_A, &lda, work );
error = lapackf77_slange( "1", &min_mn, &N, R, &ldr, work );
if ( N > 0 && Anorm > 0 )
error /= (N*Anorm);
// set R = I (K by K identity), then R = I - Q^H*Q
// error = || I - Q^H*Q || / N
lapackf77_slaset( "Upper", &min_mn, &min_mn, &c_zero, &c_one, R, &ldr );
blasf77_ssyrk( "Upper", "Conj", &min_mn, &M, &d_neg_one, Q, &ldq, &d_one, R, &ldr );
error2 = safe_lapackf77_slansy( "1", "Upper", &min_mn, R, &ldr, work );
if ( N > 0 )
error2 /= N;
magma_free_cpu( Q ); Q = NULL;
magma_free_cpu( R ); R = NULL;
magma_free_cpu( work ); work = NULL;
/* =====================================================================
Performs operation using LAPACK
=================================================================== */
cpu_time = magma_wtime();
lapackf77_sgeqrf( &M, &N, h_A, &lda, tau, h_work, &lwork, &info );
lapackf77_slarft( MagmaForwardStr, MagmaColumnwiseStr,
&M, &N, h_A, &lda, tau, h_work, &N );
cpu_time = magma_wtime() - cpu_time;
cpu_perf = gflops / cpu_time;
if (info != 0) {
printf("lapackf77_sgeqrf returned error %lld: %s.\n",
(long long) info, magma_strerror( info ));
}
/* =====================================================================
Check the result compared to LAPACK
Okay if these are different -- just chose different Householder reflectors
=================================================================== */
size = lda*N;
blasf77_saxpy( &size, &c_neg_one, h_A, &ione, h_R, &ione );
Anorm = lapackf77_slange( "M", &M, &N, h_A, &lda, rwork );
diff = lapackf77_slange( "M", &M, &N, h_R, &lda, rwork );
if ( Anorm > 0) {
diff /= (N*Anorm);
}
/* =====================================================================
Check if T is correct
=================================================================== */
// Recompute T in h_work for d_A (magma), in case it is different than h_A (lapack)
magma_sgetmatrix( M, N, d_A, ldda, h_R, lda, opts.queue );
magma_sgetmatrix( min_mn, 1, dtau, min_mn, tau, min_mn, opts.queue );
lapackf77_slarft( MagmaForwardStr, MagmaColumnwiseStr,
&M, &N, h_R, &lda, tau, h_work, &N );
magma_sgetmatrix( N, N, d_T, N, h_T, N, opts.queue );
size = N*N;
blasf77_saxpy( &size, &c_neg_one, h_work, &ione, h_T, &ione );
Anorm = lapackf77_slantr( "F", "U", "N", &N, &N, h_work, &N, rwork );
terr = lapackf77_slantr( "F", "U", "N", &N, &N, h_T, &N, rwork );
if (Anorm > 0) {
terr /= Anorm;
}
bool okay = (error < tol) && (error2 < tol) && (terr < tol);
status += ! okay;
printf("%5lld %5lld %7.2f (%7.2f) %7.2f (%7.2f) %8.2e %8.2e %8.2e %8.2e %s\n",
(long long) M, (long long) N, cpu_perf, 1000.*cpu_time, gpu_perf, 1000.*gpu_time,
error, error2, terr, diff,
(okay ? "ok" : "failed"));
}
else {
printf("%5lld %5lld --- ( --- ) %7.2f (%7.2f) --- \n",
(long long) M, (long long) N, gpu_perf, 1000.*gpu_time);
}
magma_free_cpu( tau );
magma_free_cpu( h_A );
magma_free_cpu( h_T );
magma_free_cpu( h_work );
magma_free_pinned( h_R );
magma_free( d_A );
magma_free( d_T );
magma_free( ddA );
magma_free( dtau );
magma_free( dwork );
fflush( stdout );
}
if ( opts.niter > 1 ) {
printf( "\n" );
}
}
opts.cleanup();
TESTING_CHECK( magma_finalize() );
return status;
}
|
