|
Packit |
67cb25 |
/* blas/source_gemm_r.h
|
|
Packit |
67cb25 |
*
|
|
Packit |
67cb25 |
* Copyright (C) 2001, 2007 Brian Gough
|
|
Packit |
67cb25 |
*
|
|
Packit |
67cb25 |
* This program is free software; you can redistribute it and/or modify
|
|
Packit |
67cb25 |
* it under the terms of the GNU General Public License as published by
|
|
Packit |
67cb25 |
* the Free Software Foundation; either version 3 of the License, or (at
|
|
Packit |
67cb25 |
* your option) any later version.
|
|
Packit |
67cb25 |
*
|
|
Packit |
67cb25 |
* This program is distributed in the hope that it will be useful, but
|
|
Packit |
67cb25 |
* WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
Packit |
67cb25 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
Packit |
67cb25 |
* General Public License for more details.
|
|
Packit |
67cb25 |
*
|
|
Packit |
67cb25 |
* You should have received a copy of the GNU General Public License
|
|
Packit |
67cb25 |
* along with this program; if not, write to the Free Software
|
|
Packit |
67cb25 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
|
|
Packit |
67cb25 |
*/
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
{
|
|
Packit |
67cb25 |
INDEX i, j, k;
|
|
Packit |
67cb25 |
INDEX n1, n2;
|
|
Packit |
67cb25 |
INDEX ldf, ldg;
|
|
Packit |
67cb25 |
int TransF, TransG;
|
|
Packit |
67cb25 |
const BASE *F, *G;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
CHECK_ARGS14(GEMM,Order,TransA,TransB,M,N,K,alpha,A,lda,B,ldb,beta,C,ldc);
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha == 0.0 && beta == 1.0)
|
|
Packit |
67cb25 |
return;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (Order == CblasRowMajor) {
|
|
Packit |
67cb25 |
n1 = M;
|
|
Packit |
67cb25 |
n2 = N;
|
|
Packit |
67cb25 |
F = A;
|
|
Packit |
67cb25 |
ldf = lda;
|
|
Packit |
67cb25 |
TransF = (TransA == CblasConjTrans) ? CblasTrans : TransA;
|
|
Packit |
67cb25 |
G = B;
|
|
Packit |
67cb25 |
ldg = ldb;
|
|
Packit |
67cb25 |
TransG = (TransB == CblasConjTrans) ? CblasTrans : TransB;
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
n1 = N;
|
|
Packit |
67cb25 |
n2 = M;
|
|
Packit |
67cb25 |
F = B;
|
|
Packit |
67cb25 |
ldf = ldb;
|
|
Packit |
67cb25 |
TransF = (TransB == CblasConjTrans) ? CblasTrans : TransB;
|
|
Packit |
67cb25 |
G = A;
|
|
Packit |
67cb25 |
ldg = lda;
|
|
Packit |
67cb25 |
TransG = (TransA == CblasConjTrans) ? CblasTrans : TransA;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form y := beta*y */
|
|
Packit |
67cb25 |
if (beta == 0.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
C[ldc * i + j] = 0.0;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
} else if (beta != 1.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
C[ldc * i + j] *= beta;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha == 0.0)
|
|
Packit |
67cb25 |
return;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (TransF == CblasNoTrans && TransG == CblasNoTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form C := alpha*A*B + C */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
const BASE temp = alpha * F[ldf * i + k];
|
|
Packit |
67cb25 |
if (temp != 0.0) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
C[ldc * i + j] += temp * G[ldg * k + j];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (TransF == CblasNoTrans && TransG == CblasTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form C := alpha*A*B' + C */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
BASE temp = 0.0;
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
temp += F[ldf * i + k] * G[ldg * j + k];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
C[ldc * i + j] += alpha * temp;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (TransF == CblasTrans && TransG == CblasNoTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
const BASE temp = alpha * F[ldf * k + i];
|
|
Packit |
67cb25 |
if (temp != 0.0) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
C[ldc * i + j] += temp * G[ldg * k + j];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (TransF == CblasTrans && TransG == CblasTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
BASE temp = 0.0;
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
temp += F[ldf * k + i] * G[ldg * j + k];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
C[ldc * i + j] += alpha * temp;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
BLAS_ERROR("unrecognized operation");
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|