Blame cblas/source_gbmv_r.h

Packit 67cb25
/* blas/source_gbmv_r.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
  INDEX lenX, lenY, L, U;
Packit 67cb25
Packit 67cb25
  const int Trans = (TransA != CblasConjTrans) ? TransA : CblasTrans;
Packit 67cb25
Packit 67cb25
  CHECK_ARGS14(GBMV,order,TransA,M,N,KL,KU,alpha,A,lda,X,incX,beta,Y,incY);
Packit 67cb25
Packit 67cb25
  if (M == 0 || N == 0)
Packit 67cb25
    return;
Packit 67cb25
Packit 67cb25
  if (alpha == 0.0 && beta == 1.0)
Packit 67cb25
    return;
Packit 67cb25
Packit 67cb25
  if (Trans == CblasNoTrans) {
Packit 67cb25
    lenX = N;
Packit 67cb25
    lenY = M;
Packit 67cb25
    L = KL;
Packit 67cb25
    U = KU;
Packit 67cb25
  } else {
Packit 67cb25
    lenX = M;
Packit 67cb25
    lenY = N;
Packit 67cb25
    L = KU;
Packit 67cb25
    U = KL;
Packit 67cb25
  }
Packit 67cb25
Packit 67cb25
  /* form  y := beta*y */
Packit 67cb25
  if (beta == 0.0) {
Packit 67cb25
    INDEX iy = OFFSET(lenY, incY);
Packit 67cb25
    for (i = 0; i < lenY; i++) {
Packit 67cb25
      Y[iy] = 0;
Packit 67cb25
      iy += incY;
Packit 67cb25
    }
Packit 67cb25
  } else if (beta != 1.0) {
Packit 67cb25
    INDEX iy = OFFSET(lenY, incY);
Packit 67cb25
    for (i = 0; i < lenY; i++) {
Packit 67cb25
      Y[iy] *= beta;
Packit 67cb25
      iy += incY;
Packit 67cb25
    }
Packit 67cb25
  }
Packit 67cb25
Packit 67cb25
  if (alpha == 0.0)
Packit 67cb25
    return;
Packit 67cb25
Packit 67cb25
  if ((order == CblasRowMajor && Trans == CblasNoTrans)
Packit 67cb25
      || (order == CblasColMajor && Trans == CblasTrans)) {
Packit 67cb25
    /* form  y := alpha*A*x + y */
Packit 67cb25
    INDEX iy = OFFSET(lenY, incY);
Packit 67cb25
    for (i = 0; i < lenY; i++) {
Packit 67cb25
      BASE temp = 0.0;
Packit 67cb25
      const INDEX j_min = (i > L ? i - L : 0);
Packit 67cb25
      const INDEX j_max = GSL_MIN(lenX, i + U + 1);
Packit 67cb25
      INDEX jx = OFFSET(lenX, incX) + j_min * incX;
Packit 67cb25
      for (j = j_min; j < j_max; j++) {
Packit 67cb25
        temp += X[jx] * A[(L - i + j) + i * lda];
Packit 67cb25
        jx += incX;
Packit 67cb25
      }
Packit 67cb25
      Y[iy] += alpha * temp;
Packit 67cb25
      iy += incY;
Packit 67cb25
    }
Packit 67cb25
  } else if ((order == CblasRowMajor && Trans == CblasTrans)
Packit 67cb25
             || (order == CblasColMajor && Trans == CblasNoTrans)) {
Packit 67cb25
    /* form  y := alpha*A'*x + y */
Packit 67cb25
    INDEX jx = OFFSET(lenX, incX);
Packit 67cb25
    for (j = 0; j < lenX; j++) {
Packit 67cb25
      const BASE temp = alpha * X[jx];
Packit 67cb25
      if (temp != 0.0) {
Packit 67cb25
        const INDEX i_min = (j > U ? j - U : 0);
Packit 67cb25
        const INDEX i_max = GSL_MIN(lenY, j + L + 1);
Packit 67cb25
        INDEX iy = OFFSET(lenY, incY) + i_min * incY;
Packit 67cb25
        for (i = i_min; i < i_max; i++) {
Packit 67cb25
          Y[iy] += temp * A[lda * j + (U + i - j)];
Packit 67cb25
          iy += incY;
Packit 67cb25
        }
Packit 67cb25
      }
Packit 67cb25
      jx += incX;
Packit 67cb25
    }
Packit 67cb25
  } else {
Packit 67cb25
    BLAS_ERROR("unrecognized operation");
Packit 67cb25
  }
Packit 67cb25
}