root / src / blas / HPL_dgemv.c @ 9
Historique | Voir | Annoter | Télécharger (12,58 ko)
| 1 |
/*
|
|---|---|
| 2 |
* -- High Performance Computing Linpack Benchmark (HPL)
|
| 3 |
* HPL - 2.0 - September 10, 2008
|
| 4 |
* Antoine P. Petitet
|
| 5 |
* University of Tennessee, Knoxville
|
| 6 |
* Innovative Computing Laboratory
|
| 7 |
* (C) Copyright 2000-2008 All Rights Reserved
|
| 8 |
*
|
| 9 |
* -- Copyright notice and Licensing terms:
|
| 10 |
*
|
| 11 |
* Redistribution and use in source and binary forms, with or without
|
| 12 |
* modification, are permitted provided that the following conditions
|
| 13 |
* are met:
|
| 14 |
*
|
| 15 |
* 1. Redistributions of source code must retain the above copyright
|
| 16 |
* notice, this list of conditions and the following disclaimer.
|
| 17 |
*
|
| 18 |
* 2. Redistributions in binary form must reproduce the above copyright
|
| 19 |
* notice, this list of conditions, and the following disclaimer in the
|
| 20 |
* documentation and/or other materials provided with the distribution.
|
| 21 |
*
|
| 22 |
* 3. All advertising materials mentioning features or use of this
|
| 23 |
* software must display the following acknowledgement:
|
| 24 |
* This product includes software developed at the University of
|
| 25 |
* Tennessee, Knoxville, Innovative Computing Laboratory.
|
| 26 |
*
|
| 27 |
* 4. The name of the University, the name of the Laboratory, or the
|
| 28 |
* names of its contributors may not be used to endorse or promote
|
| 29 |
* products derived from this software without specific written
|
| 30 |
* permission.
|
| 31 |
*
|
| 32 |
* -- Disclaimer:
|
| 33 |
*
|
| 34 |
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
| 35 |
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
| 36 |
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
| 37 |
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY
|
| 38 |
* OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
| 39 |
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
| 40 |
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
| 41 |
* DATA OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
| 42 |
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
| 43 |
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
| 44 |
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
| 45 |
* ---------------------------------------------------------------------
|
| 46 |
*/
|
| 47 |
/*
|
| 48 |
* Include files
|
| 49 |
*/
|
| 50 |
#include "hpl.h" |
| 51 |
|
| 52 |
#ifndef HPL_dgemv
|
| 53 |
|
| 54 |
#ifdef HPL_CALL_VSIPL
|
| 55 |
|
| 56 |
#ifdef STDC_HEADERS
|
| 57 |
static void HPL_dgemv0 |
| 58 |
( |
| 59 |
const enum HPL_TRANS TRANS, |
| 60 |
const int M, |
| 61 |
const int N, |
| 62 |
const double ALPHA, |
| 63 |
const double * A, |
| 64 |
const int LDA, |
| 65 |
const double * X, |
| 66 |
const int INCX, |
| 67 |
const double BETA, |
| 68 |
double * Y,
|
| 69 |
const int INCY |
| 70 |
) |
| 71 |
#else
|
| 72 |
static void HPL_dgemv0( TRANS, M, N, ALPHA, A, LDA, X, INCX, BETA, Y, INCY ) |
| 73 |
const enum HPL_TRANS TRANS; |
| 74 |
const int INCX, INCY, LDA, M, N; |
| 75 |
const double ALPHA, BETA; |
| 76 |
const double * A, * X; |
| 77 |
double * Y;
|
| 78 |
#endif
|
| 79 |
{
|
| 80 |
/*
|
| 81 |
* .. Local Variables ..
|
| 82 |
*/
|
| 83 |
int i, iaij, ix, iy, j, jaj, jx, jy;
|
| 84 |
register double t0; |
| 85 |
/* ..
|
| 86 |
* .. Executable Statements ..
|
| 87 |
*/
|
| 88 |
if( ( M == 0 ) || ( N == 0 ) || |
| 89 |
( ( ALPHA == HPL_rzero ) && ( BETA == HPL_rone ) ) ) return;
|
| 90 |
|
| 91 |
if( ALPHA == HPL_rzero ) { HPL_dscal( M, BETA, Y, INCY ); return; } |
| 92 |
|
| 93 |
if( TRANS == HplNoTrans )
|
| 94 |
{
|
| 95 |
HPL_dscal( M, BETA, Y, INCY ); |
| 96 |
for( j = 0, jaj = 0, jx = 0; j < N; j++, jaj += LDA, jx += INCX ) |
| 97 |
{
|
| 98 |
t0 = ALPHA * X[jx]; |
| 99 |
for( i = 0, iaij = jaj, iy = 0; i < M; i++, iaij += 1, iy += INCY ) |
| 100 |
{ Y[iy] += A[iaij] * t0; }
|
| 101 |
} |
| 102 |
} |
| 103 |
else
|
| 104 |
{
|
| 105 |
for( j = 0, jaj = 0, jy = 0; j < N; j++, jaj += LDA, jy += INCY ) |
| 106 |
{
|
| 107 |
t0 = HPL_rzero; |
| 108 |
for( i = 0, iaij = jaj, ix = 0; i < M; i++, iaij += 1, ix += INCX ) |
| 109 |
{ t0 += A[iaij] * X[ix]; }
|
| 110 |
if( BETA == HPL_rzero ) Y[jy] = ALPHA * t0;
|
| 111 |
else Y[jy] = BETA * Y[jy] + ALPHA * t0;
|
| 112 |
} |
| 113 |
} |
| 114 |
} |
| 115 |
#endif
|
| 116 |
|
| 117 |
#ifdef STDC_HEADERS
|
| 118 |
void HPL_dgemv
|
| 119 |
( |
| 120 |
const enum HPL_ORDER ORDER, |
| 121 |
const enum HPL_TRANS TRANS, |
| 122 |
const int M, |
| 123 |
const int N, |
| 124 |
const double ALPHA, |
| 125 |
const double * A, |
| 126 |
const int LDA, |
| 127 |
const double * X, |
| 128 |
const int INCX, |
| 129 |
const double BETA, |
| 130 |
double * Y,
|
| 131 |
const int INCY |
| 132 |
) |
| 133 |
#else
|
| 134 |
void HPL_dgemv
|
| 135 |
( ORDER, TRANS, M, N, ALPHA, A, LDA, X, INCX, BETA, Y, INCY ) |
| 136 |
const enum HPL_ORDER ORDER; |
| 137 |
const enum HPL_TRANS TRANS; |
| 138 |
const int M; |
| 139 |
const int N; |
| 140 |
const double ALPHA; |
| 141 |
const double * A; |
| 142 |
const int LDA; |
| 143 |
const double * X; |
| 144 |
const int INCX; |
| 145 |
const double BETA; |
| 146 |
double * Y;
|
| 147 |
const int INCY; |
| 148 |
#endif
|
| 149 |
{
|
| 150 |
/*
|
| 151 |
* Purpose
|
| 152 |
* =======
|
| 153 |
*
|
| 154 |
* HPL_dgemv performs one of the matrix-vector operations
|
| 155 |
*
|
| 156 |
* y := alpha * op( A ) * x + beta * y,
|
| 157 |
*
|
| 158 |
* where op( X ) is one of
|
| 159 |
*
|
| 160 |
* op( X ) = X or op( X ) = X^T.
|
| 161 |
*
|
| 162 |
* where alpha and beta are scalars, x and y are vectors and A is an m
|
| 163 |
* by n matrix.
|
| 164 |
*
|
| 165 |
* Arguments
|
| 166 |
* =========
|
| 167 |
*
|
| 168 |
* ORDER (local input) const enum HPL_ORDER
|
| 169 |
* On entry, ORDER specifies the storage format of the operands
|
| 170 |
* as follows:
|
| 171 |
* ORDER = HplRowMajor,
|
| 172 |
* ORDER = HplColumnMajor.
|
| 173 |
*
|
| 174 |
* TRANS (local input) const enum HPL_TRANS
|
| 175 |
* On entry, TRANS specifies the operation to be performed as
|
| 176 |
* follows:
|
| 177 |
* TRANS = HplNoTrans y := alpha*A *x + beta*y,
|
| 178 |
* TRANS = HplTrans y := alpha*A^T*x + beta*y.
|
| 179 |
*
|
| 180 |
* M (local input) const int
|
| 181 |
* On entry, M specifies the number of rows of the matrix A.
|
| 182 |
* M must be at least zero.
|
| 183 |
*
|
| 184 |
* N (local input) const int
|
| 185 |
* On entry, N specifies the number of columns of the matrix A.
|
| 186 |
* N must be at least zero.
|
| 187 |
*
|
| 188 |
* ALPHA (local input) const double
|
| 189 |
* On entry, ALPHA specifies the scalar alpha. When ALPHA is
|
| 190 |
* supplied as zero then A and X need not be set on input.
|
| 191 |
*
|
| 192 |
* A (local input) const double *
|
| 193 |
* On entry, A points to an array of size equal to or greater
|
| 194 |
* than LDA * n. Before entry, the leading m by n part of the
|
| 195 |
* array A must contain the matrix coefficients.
|
| 196 |
*
|
| 197 |
* LDA (local input) const int
|
| 198 |
* On entry, LDA specifies the leading dimension of A as
|
| 199 |
* declared in the calling (sub) program. LDA must be at
|
| 200 |
* least MAX(1,m).
|
| 201 |
*
|
| 202 |
* X (local input) const double *
|
| 203 |
* On entry, X is an incremented array of dimension at least
|
| 204 |
* ( 1 + ( n - 1 ) * abs( INCX ) ) that contains the vector x.
|
| 205 |
*
|
| 206 |
* INCX (local input) const int
|
| 207 |
* On entry, INCX specifies the increment for the elements of X.
|
| 208 |
* INCX must not be zero.
|
| 209 |
*
|
| 210 |
* BETA (local input) const double
|
| 211 |
* On entry, BETA specifies the scalar beta. When ALPHA is
|
| 212 |
* supplied as zero then Y need not be set on input.
|
| 213 |
*
|
| 214 |
* Y (local input/output) double *
|
| 215 |
* On entry, Y is an incremented array of dimension at least
|
| 216 |
* ( 1 + ( n - 1 ) * abs( INCY ) ) that contains the vector y.
|
| 217 |
* Before entry with BETA non-zero, the incremented array Y must
|
| 218 |
* contain the vector y. On exit, Y is overwritten by the
|
| 219 |
* updated vector y.
|
| 220 |
*
|
| 221 |
* INCY (local input) const int
|
| 222 |
* On entry, INCY specifies the increment for the elements of Y.
|
| 223 |
* INCY must not be zero.
|
| 224 |
*
|
| 225 |
* ---------------------------------------------------------------------
|
| 226 |
*/
|
| 227 |
#ifdef HPL_CALL_CBLAS
|
| 228 |
cblas_dgemv( ORDER, TRANS, M, N, ALPHA, A, LDA, X, INCX, BETA, Y, INCY ); |
| 229 |
#endif
|
| 230 |
#ifdef HPL_CALL_VSIPL
|
| 231 |
if( ORDER == HplColumnMajor )
|
| 232 |
{
|
| 233 |
HPL_dgemv0( TRANS, M, N, ALPHA, A, LDA, X, INCX, BETA, Y, INCY ); |
| 234 |
} |
| 235 |
else
|
| 236 |
{
|
| 237 |
HPL_dgemv0( ( TRANS == HplNoTrans ? HplTrans : HplNoTrans ), |
| 238 |
N, M, ALPHA, A, LDA, X, INCX, BETA, Y, INCY ); |
| 239 |
} |
| 240 |
#endif
|
| 241 |
|
| 242 |
|
| 243 |
#ifdef HPL_CALL_FBLAS
|
| 244 |
double alpha = ALPHA, beta = BETA;
|
| 245 |
#ifdef StringSunStyle
|
| 246 |
#ifdef HPL_USE_F77_INTEGER_DEF
|
| 247 |
F77_INTEGER IONE = 1;
|
| 248 |
#else
|
| 249 |
int IONE = 1; |
| 250 |
#endif
|
| 251 |
#endif
|
| 252 |
#ifdef StringStructVal
|
| 253 |
F77_CHAR ftran; |
| 254 |
#endif
|
| 255 |
#ifdef StringStructPtr
|
| 256 |
F77_CHAR ftran; |
| 257 |
#endif
|
| 258 |
#ifdef StringCrayStyle
|
| 259 |
F77_CHAR ftran; |
| 260 |
#endif
|
| 261 |
|
| 262 |
#ifdef HPL_USE_F77_INTEGER_DEF
|
| 263 |
const F77_INTEGER F77M = M, F77N = N,
|
| 264 |
F77lda = LDA, F77incx = INCX, F77incy = INCY; |
| 265 |
#else
|
| 266 |
#define F77M M
|
| 267 |
#define F77N N
|
| 268 |
#define F77lda LDA
|
| 269 |
#define F77incx INCX
|
| 270 |
#define F77incy INCY
|
| 271 |
#endif
|
| 272 |
char ctran;
|
| 273 |
|
| 274 |
if( ORDER == HplColumnMajor )
|
| 275 |
{
|
| 276 |
ctran = ( TRANS == HplNoTrans ? 'N' : 'T' ); |
| 277 |
|
| 278 |
#ifdef StringSunStyle
|
| 279 |
F77dgemv( &ctran, &F77M, &F77N, &alpha, A, &F77lda, X, &F77incx, |
| 280 |
&beta, Y, &F77incy, IONE ); |
| 281 |
#endif
|
| 282 |
#ifdef StringCrayStyle
|
| 283 |
ftran = HPL_C2F_CHAR( ctran ); |
| 284 |
F77dgemv( ftran, &F77M, &F77N, &alpha, A, &F77lda, X, &F77incx, |
| 285 |
&beta, Y, &F77incy ); |
| 286 |
#endif
|
| 287 |
#ifdef StringStructVal
|
| 288 |
ftran.len = 1; ftran.cp = &ctran;
|
| 289 |
F77dgemv( ftran, &F77M, &F77N, &alpha, A, &F77lda, X, &F77incx, |
| 290 |
&beta, Y, &F77incy ); |
| 291 |
#endif
|
| 292 |
#ifdef StringStructPtr
|
| 293 |
ftran.len = 1; ftran.cp = &ctran;
|
| 294 |
F77dgemv( &ftran, &F77M, &F77N, &alpha, A, &F77lda, X, &F77incx, |
| 295 |
&beta, Y, &F77incy ); |
| 296 |
#endif
|
| 297 |
} |
| 298 |
else
|
| 299 |
{
|
| 300 |
ctran = ( TRANS == HplNoTrans ? 'T' : 'N' ); |
| 301 |
#ifdef StringSunStyle
|
| 302 |
F77dgemv( &ctran, &F77N, &F77M, &alpha, A, &F77lda, X, &F77incx, |
| 303 |
&beta, Y, &F77incy, IONE ); |
| 304 |
#endif
|
| 305 |
#ifdef StringCrayStyle
|
| 306 |
ftran = HPL_C2F_CHAR( ctran ); |
| 307 |
F77dgemv( ftran, &F77N, &F77M, &alpha, A, &F77lda, X, &F77incx, |
| 308 |
&beta, Y, &F77incy ); |
| 309 |
#endif
|
| 310 |
#ifdef StringStructVal
|
| 311 |
ftran.len = 1; ftran.cp = &ctran;
|
| 312 |
F77dgemv( ftran, &F77N, &F77M, &alpha, A, &F77lda, X, &F77incx, |
| 313 |
&beta, Y, &F77incy ); |
| 314 |
#endif
|
| 315 |
#ifdef StringStructPtr
|
| 316 |
ftran.len = 1; ftran.cp = &ctran;
|
| 317 |
F77dgemv( &ftran, &F77N, &F77M, &alpha, A, &F77lda, X, &F77incx, |
| 318 |
&beta, Y, &F77incy ); |
| 319 |
#endif
|
| 320 |
} |
| 321 |
|
| 322 |
#endif
|
| 323 |
|
| 324 |
#ifdef HPL_CALL_CUBLAS
|
| 325 |
double alpha = ALPHA, beta = BETA;
|
| 326 |
|
| 327 |
int IONE = 1; |
| 328 |
|
| 329 |
#define CUBLASM M
|
| 330 |
#define CUBLASN N
|
| 331 |
#define CUBLASlda LDA
|
| 332 |
#define CUBLASincx INCX
|
| 333 |
#define CUBLASincy INCY
|
| 334 |
|
| 335 |
char ctran;
|
| 336 |
|
| 337 |
if( ORDER == HplColumnMajor )
|
| 338 |
{
|
| 339 |
ctran = ( TRANS == HplNoTrans ? 'N' : 'T' ); |
| 340 |
|
| 341 |
CUBLAS_DGEMV( &ctran, &CUBLASM, &CUBLASN, |
| 342 |
&alpha, A, &CUBLASlda, X, &CUBLASincx, |
| 343 |
&beta, Y, &CUBLASincy, IONE ); |
| 344 |
} |
| 345 |
else
|
| 346 |
{
|
| 347 |
ctran = ( TRANS == HplNoTrans ? 'T' : 'N' ); |
| 348 |
|
| 349 |
CUBLAS_DGEMV( &ctran, &CUBLASN, &CUBLASM, |
| 350 |
&alpha, A, &CUBLASlda, X, &CUBLASincx, |
| 351 |
&beta, Y, &CUBLASincy, IONE ); |
| 352 |
} |
| 353 |
|
| 354 |
#endif
|
| 355 |
/*
|
| 356 |
* End of HPL_dgemv
|
| 357 |
*/
|
| 358 |
} |
| 359 |
|
| 360 |
#endif
|