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
|
/*
-- MAGMA (version 2.9.0) --
Univ. of Tennessee, Knoxville
Univ. of California, Berkeley
Univ. of Colorado, Denver
@date January 2025
@generated from testing/testing_zhemv.cpp, normal z -> d, Wed Jan 22 14:40:19 2025
@author Mark Gates
*/
// 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 dsymv
*/
int main(int argc, char **argv)
{
#ifdef MAGMA_HAVE_OPENCL
#define dA(i_, j_) dA, ((i_) + (j_)*ldda)
#define dX(i_) dX, ((i_))
#define dY(i_) dY, ((i_))
#define dwork(i_) dwork, ((i_))
#else
#define dA(i_, j_) (dA + (i_) + (j_)*ldda)
#define dX(i_) (dX + (i_))
#define dY(i_) (dY + (i_))
#define dwork(i_) (dwork + (i_))
#endif
TESTING_CHECK( magma_init() );
magma_print_environment();
const double c_neg_one = MAGMA_D_NEG_ONE;
const magma_int_t ione = 1;
real_Double_t atomics_perf=0, atomics_time=0;
real_Double_t gflops, magma_perf=0, magma_time=0, dev_perf, dev_time, cpu_perf, cpu_time;
double magma_error=0, atomics_error=0, dev_error, work[1];
magma_int_t ISEED[4] = {0,0,0,1};
magma_int_t N, lda, ldda, sizeA, sizeX, sizeY, blocks, ldwork;
magma_int_t incx = 1;
magma_int_t incy = 1;
magma_int_t nb = 64;
double alpha = MAGMA_D_MAKE( 1.5, -2.3 );
double beta = MAGMA_D_MAKE( -0.6, 0.8 );
double *A, *X, *Y, *Yatomics, *Ydev, *Ymagma;
magmaDouble_ptr dA, dX, dY, dwork;
int status = 0;
magma_opts opts;
opts.parse_opts( argc, argv );
// See testing_dgemm about tolerance.
double eps = lapackf77_dlamch("E");
double tol = 3*eps;
printf("%% uplo = %s\n", lapack_uplo_const(opts.uplo) );
#if defined(MAGMA_HAVE_CUDA) || defined(MAGMA_HAVE_HIP)
printf("%% N MAGMA Gflop/s (ms) Atomics Gflop/s %s Gflop/s CPU Gflop/s MAGMA error %s\n",
g_platform_str, g_platform_str );
printf("%%==========================================================================================================\n");
#else
printf("%% N %s Gflop/s CPU Gflop/s MAGMA error %s\n",
g_platform_str, g_platform_str );
printf("%%===============================================================\n");
#endif
for( int itest = 0; itest < opts.ntest; ++itest ) {
for( int iter = 0; iter < opts.niter; ++iter ) {
N = opts.nsize[itest];
lda = N;
ldda = magma_roundup( N, opts.align ); // multiple of 32 by default
sizeA = N*lda;
sizeX = N*incx;
sizeY = N*incy;
gflops = FLOPS_DSYMV( N ) / 1e9;
TESTING_CHECK( magma_dmalloc_cpu( &A, sizeA ));
TESTING_CHECK( magma_dmalloc_cpu( &X, sizeX ));
TESTING_CHECK( magma_dmalloc_cpu( &Y, sizeY ));
TESTING_CHECK( magma_dmalloc_cpu( &Yatomics, sizeY ));
TESTING_CHECK( magma_dmalloc_cpu( &Ydev, sizeY ));
TESTING_CHECK( magma_dmalloc_cpu( &Ymagma, sizeY ));
TESTING_CHECK( magma_dmalloc( &dA, ldda*N ));
TESTING_CHECK( magma_dmalloc( &dX, sizeX ));
TESTING_CHECK( magma_dmalloc( &dY, sizeY ));
blocks = magma_ceildiv( N, nb );
ldwork = ldda*blocks;
TESTING_CHECK( magma_dmalloc( &dwork, ldwork ));
magmablas_dlaset( MagmaFull, ldwork, 1, MAGMA_D_NAN, MAGMA_D_NAN, dwork(0), ldwork, opts.queue );
magmablas_dlaset( MagmaFull, ldda, N, MAGMA_D_NAN, MAGMA_D_NAN, dA(0,0), ldda, opts.queue );
/* Initialize the matrix */
lapackf77_dlarnv( &ione, ISEED, &sizeA, A );
magma_dmake_symmetric( N, A, lda );
// should not use data from the opposite triangle -- fill with NAN to check
magma_int_t N1 = N-1;
if ( opts.uplo == MagmaUpper ) {
lapackf77_dlaset( "Lower", &N1, &N1, &MAGMA_D_NAN, &MAGMA_D_NAN, &A[1], &lda );
}
else {
lapackf77_dlaset( "Upper", &N1, &N1, &MAGMA_D_NAN, &MAGMA_D_NAN, &A[lda], &lda );
}
lapackf77_dlarnv( &ione, ISEED, &sizeX, X );
lapackf77_dlarnv( &ione, ISEED, &sizeY, Y );
// for error checks
double Anorm = safe_lapackf77_dlansy( "F", lapack_uplo_const(opts.uplo), &N, A, &lda, work );
double Xnorm = lapackf77_dlange( "F", &N, &ione, X, &N, work );
double Ynorm = lapackf77_dlange( "F", &N, &ione, Y, &N, work );
/* =====================================================================
Performs operation using cuBLAS / clBLAS
=================================================================== */
magma_dsetmatrix( N, N, A, lda, dA(0,0), ldda, opts.queue );
magma_dsetvector( N, X, incx, dX(0), incx, opts.queue );
magma_dsetvector( N, Y, incy, dY(0), incy, opts.queue );
dev_time = magma_sync_wtime( opts.queue );
magma_dsymv( opts.uplo, N,
alpha, dA(0,0), ldda,
dX(0), incx,
beta, dY(0), incy, opts.queue );
dev_time = magma_sync_wtime( opts.queue ) - dev_time;
dev_perf = gflops / dev_time;
magma_dgetvector( N, dY(0), incy, Ydev, incy, opts.queue );
/* =====================================================================
Performs operation using cuBLAS - using atomics
=================================================================== */
#if defined(MAGMA_HAVE_CUDA) || defined(MAGMA_HAVE_HIP)
cublasSetAtomicsMode( opts.handle, CUBLAS_ATOMICS_ALLOWED );
magma_dsetvector( N, Y, incy, dY(0), incy, opts.queue );
// sync on queue doesn't work -- need device sync or use NULL stream -- bug in CUBLAS?
atomics_time = magma_sync_wtime( NULL /*opts.queue*/ );
magma_dsymv( opts.uplo, N,
alpha, dA(0,0), ldda,
dX(0), incx,
beta, dY(0), incy, opts.queue );
atomics_time = magma_sync_wtime( NULL /*opts.queue*/ ) - atomics_time;
atomics_perf = gflops / atomics_time;
magma_dgetvector( N, dY(0), incy, Yatomics, incy, opts.queue );
cublasSetAtomicsMode( opts.handle, CUBLAS_ATOMICS_NOT_ALLOWED );
#endif
/* =====================================================================
Performs operation using MAGMABLAS (only with CUDA)
=================================================================== */
#if defined(MAGMA_HAVE_CUDA) || defined(MAGMA_HAVE_HIP)
magma_dsetvector( N, Y, incy, dY(0), incy, opts.queue );
magma_time = magma_sync_wtime( opts.queue );
if ( opts.version == 1 ) {
magmablas_dsymv_work( opts.uplo, N,
alpha, dA(0,0), ldda,
dX(0), incx,
beta, dY(0), incy,
dwork(0), ldwork, opts.queue );
}
else {
// non-work interface (has added overhead)
magmablas_dsymv( opts.uplo, N,
alpha, dA(0,0), ldda,
dX(0), incx,
beta, dY(0), incy, opts.queue );
}
magma_time = magma_sync_wtime( opts.queue ) - magma_time;
magma_perf = gflops / magma_time;
magma_dgetvector( N, dY(0), incy, Ymagma, incy, opts.queue );
#endif
/* =====================================================================
Performs operation using CPU BLAS
=================================================================== */
cpu_time = magma_wtime();
blasf77_dsymv( lapack_uplo_const(opts.uplo), &N, &alpha, A, &lda, X, &incx, &beta, Y, &incy );
cpu_time = magma_wtime() - cpu_time;
cpu_perf = gflops / cpu_time;
/* =====================================================================
Check the result
=================================================================== */
// See testing_dgemm for formula. Here K = N.
blasf77_daxpy( &N, &c_neg_one, Y, &incy, Ydev, &incy );
dev_error = lapackf77_dlange( "M", &N, &ione, Ydev, &N, work )
/ (sqrt(double(N+2))*fabs(alpha)*Anorm*Xnorm + 2*fabs(beta)*Ynorm);
#if defined(MAGMA_HAVE_CUDA) || defined(MAGMA_HAVE_HIP)
blasf77_daxpy( &N, &c_neg_one, Y, &incy, Yatomics, &incy );
atomics_error = lapackf77_dlange( "M", &N, &ione, Yatomics, &N, work )
/ (sqrt(double(N+2))*fabs(alpha)*Anorm*Xnorm + 2*fabs(beta)*Ynorm);
blasf77_daxpy( &N, &c_neg_one, Y, &incy, Ymagma, &incy );
magma_error = lapackf77_dlange( "M", &N, &ione, Ymagma, &N, work )
/ (sqrt(double(N+2))*fabs(alpha)*Anorm*Xnorm + 2*fabs(beta)*Ynorm);
#endif
bool okay = (magma_error < tol && dev_error < tol && atomics_error < tol);
status += ! okay;
printf("%5lld %7.2f (%7.2f) %7.2f (%7.2f) %7.2f (%7.2f) %7.2f (%7.2f) %8.2e %8.2e %8.2e %s\n",
(long long) N,
magma_perf, 1000.*magma_time,
atomics_perf, 1000.*atomics_time,
dev_perf, 1000.*dev_time,
cpu_perf, 1000.*cpu_time,
magma_error, dev_error, atomics_error,
(okay ? "ok" : "failed"));
magma_free_cpu( A );
magma_free_cpu( X );
magma_free_cpu( Y );
magma_free_cpu( Ydev );
magma_free_cpu( Yatomics );
magma_free_cpu( Ymagma );
magma_free( dA );
magma_free( dX );
magma_free( dY );
magma_free( dwork );
fflush( stdout );
}
if ( opts.niter > 1 ) {
printf( "\n" );
}
}
opts.cleanup();
TESTING_CHECK( magma_finalize() );
return status;
}
|
