|
Packit |
67cb25 |
/* blas/source_her2.h
|
|
Packit |
67cb25 |
*
|
|
Packit |
67cb25 |
* Copyright (C) 1996, 1997, 1998, 1999, 2000 Gerard Jungman
|
|
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;
|
|
Packit |
67cb25 |
const int conj = (order == CblasColMajor) ? -1 : 1;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
CHECK_ARGS10(CZ_HER2,order,Uplo,N,alpha,X,incX,Y,incY,A,lda);
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
{
|
|
Packit |
67cb25 |
const BASE alpha_real = CONST_REAL0(alpha);
|
|
Packit |
67cb25 |
const BASE alpha_imag = CONST_IMAG0(alpha);
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if (alpha_real == 0.0 && alpha_imag == 0.0)
|
|
Packit |
67cb25 |
return;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
if ((order == CblasRowMajor && Uplo == CblasUpper)
|
|
Packit |
67cb25 |
|| (order == CblasColMajor && Uplo == CblasLower)) {
|
|
Packit |
67cb25 |
INDEX ix = OFFSET(N, incX);
|
|
Packit |
67cb25 |
INDEX iy = OFFSET(N, incY);
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
const BASE Xi_real = CONST_REAL(X, ix);
|
|
Packit |
67cb25 |
const BASE Xi_imag = CONST_IMAG(X, ix);
|
|
Packit |
67cb25 |
/* tmp1 = alpha Xi */
|
|
Packit |
67cb25 |
const BASE tmp1_real = alpha_real * Xi_real - alpha_imag * Xi_imag;
|
|
Packit |
67cb25 |
const BASE tmp1_imag = alpha_imag * Xi_real + alpha_real * Xi_imag;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
const BASE Yi_real = CONST_REAL(Y, iy);
|
|
Packit |
67cb25 |
const BASE Yi_imag = CONST_IMAG(Y, iy);
|
|
Packit |
67cb25 |
/* tmp2 = conj(alpha) Yi */
|
|
Packit |
67cb25 |
const BASE tmp2_real = alpha_real * Yi_real + alpha_imag * Yi_imag;
|
|
Packit |
67cb25 |
const BASE tmp2_imag = -alpha_imag * Yi_real + alpha_real * Yi_imag;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
INDEX jx = ix + incX;
|
|
Packit |
67cb25 |
INDEX jy = iy + incY;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* Aij = alpha*Xi*conj(Yj) + conj(alpha)*Yi*conj(Xj) */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
REAL(A, lda * i + i) += 2 * (tmp1_real * Yi_real + tmp1_imag * Yi_imag);
|
|
Packit |
67cb25 |
IMAG(A, lda * i + i) = 0;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (j = i + 1; j < N; j++) {
|
|
Packit |
67cb25 |
const BASE Xj_real = CONST_REAL(X, jx);
|
|
Packit |
67cb25 |
const BASE Xj_imag = CONST_IMAG(X, jx);
|
|
Packit |
67cb25 |
const BASE Yj_real = CONST_REAL(Y, jy);
|
|
Packit |
67cb25 |
const BASE Yj_imag = CONST_IMAG(Y, jy);
|
|
Packit |
67cb25 |
REAL(A, lda * i + j) += ((tmp1_real * Yj_real + tmp1_imag * Yj_imag)
|
|
Packit |
67cb25 |
+ (tmp2_real * Xj_real + tmp2_imag * Xj_imag));
|
|
Packit |
67cb25 |
IMAG(A, lda * i + j) +=
|
|
Packit |
67cb25 |
conj * ((tmp1_imag * Yj_real - tmp1_real * Yj_imag) +
|
|
Packit |
67cb25 |
(tmp2_imag * Xj_real - tmp2_real * Xj_imag));
|
|
Packit |
67cb25 |
jx += incX;
|
|
Packit |
67cb25 |
jy += incY;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
ix += incX;
|
|
Packit |
67cb25 |
iy += incY;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
} else if ((order == CblasRowMajor && Uplo == CblasLower)
|
|
Packit |
67cb25 |
|| (order == CblasColMajor && Uplo == CblasUpper)) {
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
INDEX ix = OFFSET(N, incX);
|
|
Packit |
67cb25 |
INDEX iy = OFFSET(N, incY);
|
|
Packit |
67cb25 |
for (i = 0; i < N; i++) {
|
|
Packit |
67cb25 |
const BASE Xi_real = CONST_REAL(X, ix);
|
|
Packit |
67cb25 |
const BASE Xi_imag = CONST_IMAG(X, ix);
|
|
Packit |
67cb25 |
const BASE tmp1_real = alpha_real * Xi_real - alpha_imag * Xi_imag;
|
|
Packit |
67cb25 |
const BASE tmp1_imag = alpha_imag * Xi_real + alpha_real * Xi_imag;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
const BASE Yi_real = CONST_REAL(Y, iy);
|
|
Packit |
67cb25 |
const BASE Yi_imag = CONST_IMAG(Y, iy);
|
|
Packit |
67cb25 |
const BASE tmp2_real = alpha_real * Yi_real + alpha_imag * Yi_imag;
|
|
Packit |
67cb25 |
const BASE tmp2_imag = -alpha_imag * Yi_real + alpha_real * Yi_imag;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
INDEX jx = OFFSET(N, incX);
|
|
Packit |
67cb25 |
INDEX jy = OFFSET(N, incY);
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
/* Aij = alpha*Xi*conj(Yj) + conj(alpha)*Yi*conj(Xj) */
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
for (j = 0; j < i; j++) {
|
|
Packit |
67cb25 |
const BASE Xj_real = CONST_REAL(X, jx);
|
|
Packit |
67cb25 |
const BASE Xj_imag = CONST_IMAG(X, jx);
|
|
Packit |
67cb25 |
const BASE Yj_real = CONST_REAL(Y, jy);
|
|
Packit |
67cb25 |
const BASE Yj_imag = CONST_IMAG(Y, jy);
|
|
Packit |
67cb25 |
REAL(A, lda * i + j) += ((tmp1_real * Yj_real + tmp1_imag * Yj_imag)
|
|
Packit |
67cb25 |
+ (tmp2_real * Xj_real + tmp2_imag * Xj_imag));
|
|
Packit |
67cb25 |
IMAG(A, lda * i + j) +=
|
|
Packit |
67cb25 |
conj * ((tmp1_imag * Yj_real - tmp1_real * Yj_imag) +
|
|
Packit |
67cb25 |
(tmp2_imag * Xj_real - tmp2_real * Xj_imag));
|
|
Packit |
67cb25 |
jx += incX;
|
|
Packit |
67cb25 |
jy += incY;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
REAL(A, lda * i + i) += 2 * (tmp1_real * Yi_real + tmp1_imag * Yi_imag);
|
|
Packit |
67cb25 |
IMAG(A, lda * i + i) = 0;
|
|
Packit |
67cb25 |
|
|
Packit |
67cb25 |
ix += incX;
|
|
Packit |
67cb25 |
iy += incY;
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
} else {
|
|
Packit |
67cb25 |
BLAS_ERROR("unrecognized operation");
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|
|
Packit |
67cb25 |
}
|