OpenMP

/*
Test of C code speed
compile with: gcc -O3 -mavx mactest.c
or for openmp use: gcc -Ofast -fopenmp mactest.c
Modified version of https://gist.github.com/culurciello/8389426
*/

#include <stdio.h>
#include <stdlib.h>
#include <fcntl.h>
#include <sys/time.h>
#include <math.h>
#include <omp.h>

int main()
{
  int i,j,k;
  // nb of operations:
  int dsize = 524288/2; // 1024*512
  float a[2][dsize], b[2][dsize];
  int opsMAC = 2; // operations per MAC

  int vcrz = 8; // vectorize by #
  int nthreads = 2;
  int nbOfAverages = 1024*10;
  struct timeval start,end;
  gettimeofday(&start, NULL);


#pragma omp parallel for private (i,j,k)
  for (k=0; k<nthreads; k++) {
    // printf("Hello from thread %d, nthreads %d\n", omp_get_thread_num(), omp_get_num_threads());
    for(i=0;i<nbOfAverages;i++) {
      for (j = 0; j < dsize/vcrz; j=j+4) {
#      pragma simd
#      pragma vector aligned
    a[k][j] = a[k][j] * b[k][j]; // MAC operations
    a[k][j+1] = a[k][j+1] * b[k][j+1];
    a[k][j+2] = a[k][j+2] * b[k][j+2];
    a[k][j+3] = a[k][j+3] * b[k][j+3];
    a[k][j+4] = a[k][j+4] * b[k][j+4];
    a[k][j+5] = a[k][j+5] * b[k][j+5];
    a[k][j+6] = a[k][j+6] * b[k][j+6];
    a[k][j+7] = a[k][j+7] * b[k][j+7];
      }
    }
  }
  gettimeofday(&end, NULL);
  double t = ((double) (end.tv_sec - start.tv_sec))
    + ((double) (end.tv_usec - start.tv_usec)) / 1e6; //reports time in [s] - verified!

  // report performance:
  //printf("\nclockstart, clockedn, CLOCKS_PER_SEC: %ld,%ld,%d", start, end, CLOCKS_PER_SEC);
  printf("\nTotal ops = %d, # of treads = %d", opsMAC*dsize, nthreads);
  printf("\nTime in s: %lf:", t);
  printf("\nTest performance [G OP/s] %lf:", nthreads * opsMAC * nbOfAverages * dsize/t/1e9);
  printf("\n");
  return(0);
}