ger.hh

// Copyright (c) 2017-2023, University of Tennessee. All rights reserved.
// SPDX-License-Identifier: BSD-3-Clause
// This program is free software: you can redistribute it and/or modify it under
// the terms of the BSD 3-Clause license. See the accompanying LICENSE file.
 
#ifndef BLAS_GER_HH
#define BLAS_GER_HH
 
#include "blas/util.hh"
 
#include <limits>
 
namespace blas {
 
// =============================================================================
 
template <typename TA, typename TX, typename TY>
void ger(
    blas::Layout layout,
    int64_t m, int64_t n,
    blas::scalar_type<TA, TX, TY> alpha,
    TX const *x, int64_t incx,
    TY const *y, int64_t incy,
    TA *A, int64_t lda )
{
    typedef blas::scalar_type<TA, TX, TY> scalar_t;
 
    #define A(i_, j_) A[ (i_) + (j_)*lda ]
 
    // constants
    const scalar_t zero = 0;
 
    // check arguments
    blas_error_if( layout != Layout::ColMajor &&
                   layout != Layout::RowMajor );
    blas_error_if( m < 0 );
    blas_error_if( n < 0 );
    blas_error_if( incx == 0 );
    blas_error_if( incy == 0 );
 
    if (layout == Layout::ColMajor)
        blas_error_if( lda < m );
    else
        blas_error_if( lda < n );
 
    // quick return
    if (m == 0 || n == 0 || alpha == zero)
        return;
 
    if (layout == Layout::ColMajor) {
        if (incx == 1 && incy == 1) {
            // unit stride
            for (int64_t j = 0; j < n; ++j) {
                // note: NOT skipping if y[j] is zero, for consistent NAN handling
                scalar_t tmp = alpha * conj( y[j] );
                for (int64_t i = 0; i < m; ++i) {
                    A(i, j) += x[i] * tmp;
                }
            }
        }
        else if (incx == 1) {
            // x unit stride, y non-unit stride
            int64_t jy = (incy > 0 ? 0 : (-n + 1)*incy);
            for (int64_t j = 0; j < n; ++j) {
                scalar_t tmp = alpha * conj( y[jy] );
                for (int64_t i = 0; i < m; ++i) {
                    A(i, j) += x[i] * tmp;
                }
                jy += incy;
            }
        }
        else {
            // x and y non-unit stride
            int64_t kx = (incx > 0 ? 0 : (-m + 1)*incx);
            int64_t jy = (incy > 0 ? 0 : (-n + 1)*incy);
            for (int64_t j = 0; j < n; ++j) {
                scalar_t tmp = alpha * conj( y[jy] );
                int64_t ix = kx;
                for (int64_t i = 0; i < m; ++i) {
                    A(i, j) += x[ix] * tmp;
                    ix += incx;
                }
                jy += incy;
            }
        }
    }
    else {
        // RowMajor
        if (incx == 1 && incy == 1) {
            // unit stride
            for (int64_t i = 0; i < m; ++i) {
                // note: NOT skipping if x[i] is zero, for consistent NAN handling
                scalar_t tmp = alpha * x[i];
                for (int64_t j = 0; j < n; ++j) {
                    A(j, i) += tmp * conj( y[j] );
                }
            }
        }
        else if (incy == 1) {
            // x non-unit stride, y unit stride
            int64_t ix = (incx > 0 ? 0 : (-m + 1)*incx);
            for (int64_t i = 0; i < m; ++i) {
                scalar_t tmp = alpha * x[ix];
                for (int64_t j = 0; j < n; ++j) {
                    A(j, i) += tmp * conj( y[j] );
                }
                ix += incx;
            }
        }
        else {
            // x and y non-unit stride
            int64_t ky = (incy > 0 ? 0 : (-n + 1)*incy);
            int64_t ix = (incx > 0 ? 0 : (-m + 1)*incx);
            for (int64_t i = 0; i < m; ++i) {
                scalar_t tmp = alpha * x[ix];
                int64_t jy = ky;
                for (int64_t j = 0; j < n; ++j) {
                    A(j, i) += tmp * conj( y[jy] );
                    jy += incy;
                }
                ix += incx;
            }
        }
    }
 
    #undef A
}
 
}  // namespace blas
 
#endif        //  #ifndef BLAS_GER_HH