/* Include polybench common header. */
#include "polybench.h"

/* Include benchmark-specific header. */
/* Default data type is double, default size is 4000. */
#include "syrk.h"

/* Main computational kernel. The whole function will be timed,
   including the call and return. */
void kernel_syrk(int ni, int nj,
		 DATA_TYPE alpha,
		 DATA_TYPE beta,
		 DATA_TYPE POLYBENCH_2D(C,NI,NI,ni,ni),
		 DATA_TYPE POLYBENCH_2D(A,NI,NJ,ni,nj))
{
  int i, j, k;

  /*  C := alpha*A*A' + beta*C */
#if 0
  for (i = 0; i < _PB_NI; i++)
    for (j = 0; j < _PB_NI; j++)
      C[i][j] *= beta;
#endif
  for (i = 0; i < _PB_NI; i++)
    for (j = 0; j < _PB_NI; j++)
      for (k = 0; k < _PB_NJ; k++)
	  C[i][j] += alpha * A[i][k] * A[j][k];
}