|
Packit |
67cb25 |
/* blas/source_trsm_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 |
|
|
Packit |
67cb25 |
const int nonunit = (Diag == CblasNonUnit);
|
|
Packit |
67cb25 |
int side, uplo, trans;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
CHECK_ARGS12(TRSM,Order,Side,Uplo,TransA,Diag,M,N,alpha,A,lda,B,ldb);
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (Order == CblasRowMajor) {
|
|
Packit |
67cb25 |
n1 = M;
|
|
Packit |
67cb25 |
n2 = N;
|
|
Packit |
67cb25 |
side = Side;
|
|
Packit |
67cb25 |
uplo = Uplo;
|
|
Packit |
67cb25 |
trans = (TransA == CblasConjTrans) ? CblasTrans : TransA;
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
n1 = N;
|
|
Packit |
67cb25 |
n2 = M;
|
|
Packit |
67cb25 |
side = (Side == CblasLeft) ? CblasRight : CblasLeft;
|
|
Packit |
67cb25 |
uplo = (Uplo == CblasUpper) ? CblasLower : CblasUpper;
|
|
Packit |
67cb25 |
trans = (TransA == CblasConjTrans) ? CblasTrans : TransA;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (side == CblasLeft && uplo == CblasUpper && trans == CblasNoTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form B := alpha * inv(TriU(A)) *B */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha != 1.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] *= alpha;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = n1; i > 0 && i--;) {
|
|
Packit |
67cb25 |
if (nonunit) {
|
|
Packit |
67cb25 |
BASE Aii = A[lda * i + i];
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] /= Aii;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (k = 0; k < i; k++) {
|
|
Packit |
67cb25 |
const BASE Aki = A[k * lda + i];
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * k + j] -= Aki * B[ldb * i + j];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (side == CblasLeft && uplo == CblasUpper && trans == CblasTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form B := alpha * inv(TriU(A))' *B */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha != 1.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] *= alpha;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
if (nonunit) {
|
|
Packit |
67cb25 |
BASE Aii = A[lda * i + i];
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] /= Aii;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (k = i + 1; k < n1; k++) {
|
|
Packit |
67cb25 |
const BASE Aik = A[i * lda + k];
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * k + j] -= Aik * B[ldb * i + j];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (side == CblasLeft && uplo == CblasLower && trans == CblasNoTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form B := alpha * inv(TriL(A))*B */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha != 1.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] *= alpha;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
if (nonunit) {
|
|
Packit |
67cb25 |
BASE Aii = A[lda * i + i];
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] /= Aii;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (k = i + 1; k < n1; k++) {
|
|
Packit |
67cb25 |
const BASE Aki = A[k * lda + i];
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * k + j] -= Aki * B[ldb * i + j];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (side == CblasLeft && uplo == CblasLower && trans == CblasTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form B := alpha * TriL(A)' *B */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha != 1.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] *= alpha;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = n1; i > 0 && i--;) {
|
|
Packit |
67cb25 |
if (nonunit) {
|
|
Packit |
67cb25 |
BASE Aii = A[lda * i + i];
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] /= Aii;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (k = 0; k < i; k++) {
|
|
Packit |
67cb25 |
const BASE Aik = A[i * lda + k];
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * k + j] -= Aik * B[ldb * i + j];
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (side == CblasRight && uplo == CblasUpper && trans == CblasNoTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form B := alpha * B * inv(TriU(A)) */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha != 1.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] *= alpha;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
if (nonunit) {
|
|
Packit |
67cb25 |
BASE Ajj = A[lda * j + j];
|
|
Packit |
67cb25 |
B[ldb * i + j] /= Ajj;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
{
|
|
Packit |
67cb25 |
BASE Bij = B[ldb * i + j];
|
|
Packit |
67cb25 |
for (k = j + 1; k < n2; k++) {
|
|
Packit |
67cb25 |
B[ldb * i + k] -= A[j * lda + k] * Bij;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (side == CblasRight && uplo == CblasUpper && trans == CblasTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form B := alpha * B * inv(TriU(A))' */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha != 1.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] *= alpha;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = n2; j > 0 && j--;) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (nonunit) {
|
|
Packit |
67cb25 |
BASE Ajj = A[lda * j + j];
|
|
Packit |
67cb25 |
B[ldb * i + j] /= Ajj;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
{
|
|
Packit |
67cb25 |
BASE Bij = B[ldb * i + j];
|
|
Packit |
67cb25 |
for (k = 0; k < j; k++) {
|
|
Packit |
67cb25 |
B[ldb * i + k] -= A[k * lda + j] * Bij;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (side == CblasRight && uplo == CblasLower && trans == CblasNoTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form B := alpha * B * inv(TriL(A)) */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha != 1.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] *= alpha;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = n2; j > 0 && j--;) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (nonunit) {
|
|
Packit |
67cb25 |
BASE Ajj = A[lda * j + j];
|
|
Packit |
67cb25 |
B[ldb * i + j] /= Ajj;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
{
|
|
Packit |
67cb25 |
BASE Bij = B[ldb * i + j];
|
|
Packit |
67cb25 |
for (k = 0; k < j; k++) {
|
|
Packit |
67cb25 |
B[ldb * i + k] -= A[j * lda + k] * Bij;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else if (side == CblasRight && uplo == CblasLower && trans == CblasTrans) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* form B := alpha * B * inv(TriL(A))' */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha != 1.0) {
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
B[ldb * i + j] *= alpha;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (i = 0; i < n1; i++) {
|
|
Packit |
67cb25 |
for (j = 0; j < n2; j++) {
|
|
Packit |
67cb25 |
if (nonunit) {
|
|
Packit |
67cb25 |
BASE Ajj = A[lda * j + j];
|
|
Packit |
67cb25 |
B[ldb * i + j] /= Ajj;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
{
|
|
Packit |
67cb25 |
BASE Bij = B[ldb * i + j];
|
|
Packit |
67cb25 |
for (k = j + 1; k < n2; k++) {
|
|
Packit |
67cb25 |
B[ldb * i + k] -= A[k * lda + j] * Bij;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
BLAS_ERROR("unrecognized operation");
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|