有效的方式来计算矩阵与GSL直积 [英] Efficient way to compute Kronecker product of matrices with GSL

问题描述

我的算法的瓶颈是我的函数克罗内克产品称为KPro：

The bottleneck of my algorithm is my function Kronecker Product called KPro:

gsl_matrix *KPro(gsl_matrix *a, gsl_matrix *b) {
    int i, j, k, l;
    int m, p, n, q;
    m = a->size1;
    p = a->size2;
    n = b->size1;
    q = b->size2;

    gsl_matrix *c = gsl_matrix_alloc(m*n, p*q);
    double da, db;

     for (i = 0; i < m; i++)    {
          for (j = 0; j < p; j++)   {
              da = gsl_matrix_get (a, i, j);
              for (k = 0; k < n; k++)   {
                  for (l = 0; l < q; l++)   {
                      db = gsl_matrix_get (b, k, l);
                      gsl_matrix_set (c, n*i+k, q*j+l, da * db);                
                  }
              }
          }
      }

    return c;
}

你知道使用GSL高效的实现？我无法找到一个合适的程序。

Do you know an efficient implementation using GSL? I can't find a suitable routine.

推荐答案

您可以显著提高通过堵，更有效地利用高速缓存的性能。

You can significantly improve the performance by 'blocking' and utilizing cache memory more effectively.

看看这个纸。是有伪code，我想你就可以很容易地变成C $ C $℃。它也有一种算法来找出最佳块大小给定的高速缓存大小和矩阵参数

Take a look at this paper. Is has pseudo code that I think you will be able to easily turn into C code. It also has an algorithm to figure out the optimum block size given cache size and matrix parameters.

这篇关于有效的方式来计算矩阵与GSL直积的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！