|
Packit |
67cb25 |
/* blas/source_herk.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_ARGS11(HERK,Order,Uplo,Trans,N,K,alpha,A,lda,beta,C,ldc);
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (beta == 1.0 && (alpha == 0.0 || K == 0))
|
|
Packit |
67cb25 |
return;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (Order == CblasRowMajor) {
|
|
Packit |
67cb25 |
uplo = Uplo;
|
|
Packit |
67cb25 |
trans = Trans;
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
uplo = (Uplo == CblasUpper) ? CblasLower : CblasUpper;
|
|
Packit |
67cb25 |
trans = (Trans == CblasNoTrans) ? CblasConjTrans : CblasNoTrans;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form y := beta*y */
|
|
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 |
REAL(C, ldc * i + j) = 0.0;
|
|
Packit |
67cb25 |
IMAG(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 |
REAL(C, ldc * i + j) = 0.0;
|
|
Packit |
67cb25 |
IMAG(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 |
REAL(C, ldc * i + i) *= beta;
|
|
Packit |
67cb25 |
IMAG(C, ldc * i + i) = 0;
|
|
Packit |
67cb25 |
for (j = i + 1; j < N; j++) {
|
|
Packit |
67cb25 |
REAL(C, ldc * i + j) *= beta;
|
|
Packit |
67cb25 |
IMAG(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 |
REAL(C, ldc * i + j) *= beta;
|
|
Packit |
67cb25 |
IMAG(C, ldc * i + j) *= beta;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
REAL(C, ldc * i + i) *= beta;
|
|
Packit |
67cb25 |
IMAG(C, ldc * i + i) = 0;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
/* set imaginary part of Aii to zero */
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
IMAG(C, ldc * i + i) = 0.0;
|
|
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_real = 0.0;
|
|
Packit |
67cb25 |
BASE temp_imag = 0.0;
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
const BASE Aik_real = CONST_REAL(A, i * lda + k);
|
|
Packit |
67cb25 |
const BASE Aik_imag = CONST_IMAG(A, i * lda + k);
|
|
Packit |
67cb25 |
const BASE Ajk_real = CONST_REAL(A, j * lda + k);
|
|
Packit |
67cb25 |
const BASE Ajk_imag = -CONST_IMAG(A, j * lda + k);
|
|
Packit |
67cb25 |
temp_real += Aik_real * Ajk_real - Aik_imag * Ajk_imag;
|
|
Packit |
67cb25 |
temp_imag += Aik_real * Ajk_imag + Aik_imag * Ajk_real;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
REAL(C, i * ldc + j) += alpha * temp_real;
|
|
Packit |
67cb25 |
IMAG(C, i * ldc + j) += alpha * temp_imag;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (uplo == CblasUpper && trans == CblasConjTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
for (j = i; j < N; j++) {
|
|
Packit |
67cb25 |
BASE temp_real = 0.0;
|
|
Packit |
67cb25 |
BASE temp_imag = 0.0;
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
const BASE Aki_real = CONST_REAL(A, k * lda + i);
|
|
Packit |
67cb25 |
const BASE Aki_imag = -CONST_IMAG(A, k * lda + i);
|
|
Packit |
67cb25 |
const BASE Akj_real = CONST_REAL(A, k * lda + j);
|
|
Packit |
67cb25 |
const BASE Akj_imag = CONST_IMAG(A, k * lda + j);
|
|
Packit |
67cb25 |
temp_real += Aki_real * Akj_real - Aki_imag * Akj_imag;
|
|
Packit |
67cb25 |
temp_imag += Aki_real * Akj_imag + Aki_imag * Akj_real;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
REAL(C, i * ldc + j) += alpha * temp_real;
|
|
Packit |
67cb25 |
IMAG(C, i * ldc + j) += alpha * temp_imag;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (uplo == CblasLower && trans == CblasNoTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j <= i; j++) {
|
|
Packit |
67cb25 |
BASE temp_real = 0.0;
|
|
Packit |
67cb25 |
BASE temp_imag = 0.0;
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
const BASE Aik_real = CONST_REAL(A, i * lda + k);
|
|
Packit |
67cb25 |
const BASE Aik_imag = CONST_IMAG(A, i * lda + k);
|
|
Packit |
67cb25 |
const BASE Ajk_real = CONST_REAL(A, j * lda + k);
|
|
Packit |
67cb25 |
const BASE Ajk_imag = -CONST_IMAG(A, j * lda + k);
|
|
Packit |
67cb25 |
temp_real += Aik_real * Ajk_real - Aik_imag * Ajk_imag;
|
|
Packit |
67cb25 |
temp_imag += Aik_real * Ajk_imag + Aik_imag * Ajk_real;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
REAL(C, i * ldc + j) += alpha * temp_real;
|
|
Packit |
67cb25 |
IMAG(C, i * ldc + j) += alpha * temp_imag;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (uplo == CblasLower && trans == CblasConjTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j <= i; j++) {
|
|
Packit |
67cb25 |
BASE temp_real = 0.0;
|
|
Packit |
67cb25 |
BASE temp_imag = 0.0;
|
|
Packit |
67cb25 |
for (k = 0; k < K; k++) {
|
|
Packit |
67cb25 |
const BASE Aki_real = CONST_REAL(A, k * lda + i);
|
|
Packit |
67cb25 |
const BASE Aki_imag = -CONST_IMAG(A, k * lda + i);
|
|
Packit |
67cb25 |
const BASE Akj_real = CONST_REAL(A, k * lda + j);
|
|
Packit |
67cb25 |
const BASE Akj_imag = CONST_IMAG(A, k * lda + j);
|
|
Packit |
67cb25 |
temp_real += Aki_real * Akj_real - Aki_imag * Akj_imag;
|
|
Packit |
67cb25 |
temp_imag += Aki_real * Akj_imag + Aki_imag * Akj_real;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
REAL(C, i * ldc + j) += alpha * temp_real;
|
|
Packit |
67cb25 |
IMAG(C, i * ldc + j) += alpha * temp_imag;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
BLAS_ERROR("unrecognized operation");
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|