|
Packit |
67cb25 |
/* blas/source_syr2k_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 |
int uplo, trans;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
CHECK_ARGS13(SYR2K,Order,Uplo,Trans,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 |
uplo = Uplo;
|
|
Packit |
67cb25 |
trans = (Trans == CblasConjTrans) ? CblasTrans : Trans;
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
uplo = (Uplo == CblasUpper) ? CblasLower : CblasUpper;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (Trans == CblasTrans || Trans == CblasConjTrans) {
|
|
Packit |
67cb25 |
trans = CblasNoTrans;
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
trans = CblasTrans;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form C := beta*C */
|
|
Packit |
67cb25 |
if (beta == 0.0) {
|
|
Packit |
67cb25 |
if (uplo == CblasUpper) {
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
for (j = i; j < N; j++) {
|
|
Packit |
67cb25 |
C[ldc * i + j] = 0.0;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j <= i; j++) {
|
|
Packit |
67cb25 |
C[ldc * i + j] = 0.0;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
} else if (beta != 1.0) {
|
|
Packit |
67cb25 |
if (uplo == CblasUpper) {
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
for (j = i; j < N; j++) {
|
|
Packit |
67cb25 |
C[ldc * i + j] *= beta;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j <= i; j++) {
|
|
Packit |
67cb25 |
C[ldc * i + j] *= beta;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha == 0.0)
|
|
Packit |
67cb25 |
return;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (uplo == CblasUpper && trans == CblasNoTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
for (j = i; j < N; j++) {
|
|
Packit |
67cb25 |
BASE temp = 0.0;
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
temp += (A[i * lda + k] * B[j * ldb + k]
|
|
Packit |
67cb25 |
+ B[i * ldb + k] * A[j * lda + k]);
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
C[i * ldc + j] += alpha * temp;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (uplo == CblasUpper && trans == CblasTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
BASE temp1 = alpha * A[k * lda + i];
|
|
Packit |
67cb25 |
BASE temp2 = alpha * B[k * ldb + i];
|
|
Packit |
67cb25 |
for (j = i; j < N; j++) {
|
|
Packit |
67cb25 |
C[i * lda + j] += temp1 * B[k * ldb + j] + temp2 * A[k * lda + j];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (uplo == CblasLower && trans == CblasNoTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j <= i; j++) {
|
|
Packit |
67cb25 |
BASE temp = 0.0;
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
temp += (A[i * lda + k] * B[j * ldb + k]
|
|
Packit |
67cb25 |
+ B[i * ldb + k] * A[j * lda + k]);
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
C[i * ldc + j] += alpha * temp;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (uplo == CblasLower && trans == CblasTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
BASE temp1 = alpha * A[k * lda + i];
|
|
Packit |
67cb25 |
BASE temp2 = alpha * B[k * ldb + i];
|
|
Packit |
67cb25 |
for (j = 0; j <= i; j++) {
|
|
Packit |
67cb25 |
C[i * lda + j] += temp1 * B[k * ldb + j] + temp2 * A[k * lda + j];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
BLAS_ERROR("unrecognized operation");
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|