Blame cblas/source_symm_r.h

Packit 67cb25
/* blas/source_symm_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
  int uplo, side;
Packit 67cb25
Packit 67cb25
  CHECK_ARGS13(SYMM,Order,Side,Uplo,M,N,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
    uplo = Uplo;
Packit 67cb25
    side = Side;
Packit 67cb25
  } else {
Packit 67cb25
    n1 = N;
Packit 67cb25
    n2 = M;
Packit 67cb25
    uplo = (Uplo == CblasUpper) ? CblasLower : CblasUpper;
Packit 67cb25
    side = (Side == CblasLeft) ? CblasRight : CblasLeft;
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 (side == CblasLeft && uplo == CblasUpper) {
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
        const BASE temp1 = alpha * B[ldb * i + j];
Packit 67cb25
        BASE temp2 = 0.0;
Packit 67cb25
        C[i * ldc + j] += temp1 * A[i * lda + i];
Packit 67cb25
        for (k = i + 1; k < n1; k++) {
Packit 67cb25
          const BASE Aik = A[i * lda + k];
Packit 67cb25
          C[k * ldc + j] += Aik * temp1;
Packit 67cb25
          temp2 += Aik * B[ldb * k + j];
Packit 67cb25
        }
Packit 67cb25
        C[i * ldc + j] += alpha * temp2;
Packit 67cb25
      }
Packit 67cb25
    }
Packit 67cb25
Packit 67cb25
  } else if (side == CblasLeft && uplo == CblasLower) {
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
        const BASE temp1 = alpha * B[ldb * i + j];
Packit 67cb25
        BASE temp2 = 0.0;
Packit 67cb25
        for (k = 0; k < i; k++) {
Packit 67cb25
          const BASE Aik = A[i * lda + k];
Packit 67cb25
          C[k * ldc + j] += Aik * temp1;
Packit 67cb25
          temp2 += Aik * B[ldb * k + j];
Packit 67cb25
        }
Packit 67cb25
        C[i * ldc + j] += temp1 * A[i * lda + i] + alpha * temp2;
Packit 67cb25
      }
Packit 67cb25
    }
Packit 67cb25
Packit 67cb25
  } else if (side == CblasRight && uplo == CblasUpper) {
Packit 67cb25
Packit 67cb25
    /* form  C := alpha*B*A + C */
Packit 67cb25
Packit 67cb25
    for (i = 0; i < n1; i++) {
Packit 67cb25
      for (j = 0; j < n2; j++) {
Packit 67cb25
        const BASE temp1 = alpha * B[ldb * i + j];
Packit 67cb25
        BASE temp2 = 0.0;
Packit 67cb25
        C[i * ldc + j] += temp1 * A[j * lda + j];
Packit 67cb25
        for (k = j + 1; k < n2; k++) {
Packit 67cb25
          const BASE Ajk = A[j * lda + k];
Packit 67cb25
          C[i * ldc + k] += temp1 * Ajk;
Packit 67cb25
          temp2 += B[ldb * i + k] * Ajk;
Packit 67cb25
        }
Packit 67cb25
        C[i * ldc + j] += alpha * temp2;
Packit 67cb25
      }
Packit 67cb25
    }
Packit 67cb25
Packit 67cb25
  } else if (side == CblasRight && uplo == CblasLower) {
Packit 67cb25
Packit 67cb25
    /* form  C := alpha*B*A + C */
Packit 67cb25
Packit 67cb25
    for (i = 0; i < n1; i++) {
Packit 67cb25
      for (j = 0; j < n2; j++) {
Packit 67cb25
        const BASE temp1 = alpha * B[ldb * i + j];
Packit 67cb25
        BASE temp2 = 0.0;
Packit 67cb25
        for (k = 0; k < j; k++) {
Packit 67cb25
          const BASE Ajk = A[j * lda + k];
Packit 67cb25
          C[i * ldc + k] += temp1 * Ajk;
Packit 67cb25
          temp2 += B[ldb * i + k] * Ajk;
Packit 67cb25
        }
Packit 67cb25
        C[i * ldc + j] += temp1 * A[j * lda + j] + alpha * temp2;
Packit 67cb25
      }
Packit 67cb25
    }
Packit 67cb25
Packit 67cb25
  } else {
Packit 67cb25
    BLAS_ERROR("unrecognized operation");
Packit 67cb25
  }
Packit 67cb25
}