xref: /petsc/src/benchmarks/streams/OpenMPVersionLikeMPI.c (revision 6c5693054f5123506dab0f5da2d352ed973d0e50)

/*
  A simplification of the Stream benchmark for OpenMP
  The array for each thread is a large distance from the array for the other threads
  Original code developed by John D. McCalpin
*/
#include <stdio.h>
#include <math.h>
#include <limits.h>
#include <float.h>
#include <sys/time.h>
#include <stdlib.h>
#include <omp.h>
#include <petscsys.h>

#define NTIMESINNER 1
#define N           2 * 4 * 20000000
//#define N 1200000
//#define N 120000
#define NTIMES 50
#define OFFSET 0

#if !defined(MIN)
  #define MIN(x, y) ((x) < (y) ? (x) : (y))
#endif
#if !defined(MAX)
  #define MAX(x, y) ((x) > (y) ? (x) : (y))
#endif

static double a[64][10000000], b[64][10000000], c[64][10000000];
static double mintime = FLT_MAX;
static double bytes   = 3 * sizeof(double) * N;

int main()
{
  const static double scalar = 3.0;
#pragma omp threadprivate(scalar)
  double     times[NTIMES], rate;
  int        size;
  static int n;
#pragma omp threadprivate(n)
  char *env;
  FILE *fd;

  env = getenv("OMP_NUM_THREADS");
  if (!env) env = (char *)"1";
  sscanf(env, "%d", &size);

#pragma omp parallel for schedule(static)
  for (int j = 0; j < size; j++) {
    n = (N / size + ((N % size) > omp_get_thread_num()));
    for (int i = 0; i < n; i++) {
      a[j][i] = 1.0;
      b[j][i] = 2.0;
      c[j][i] = 3.0;
    }
  }

  /*  --- MAIN LOOP --- repeat test cases NTIMES times --- */
  for (int k = 0; k < NTIMES; k++) {
    times[k] = MPI_Wtime();
    // https://www.openmp.org/wp-content/uploads/OpenMP-API-Specification-5-2.pdf
// #pragma omp parallel for  (same performance as below)
// #pragma omp parallel for simd schedule(static)  (same performance as below)
#pragma omp parallel for schedule(static)
    for (int j = 0; j < size; j++) {
      n                = (N / size + ((N % size) > omp_get_thread_num()));
      double       *aa = a[j]; // these don't change the timings
      const double *bb = b[j];
      const double *cc = c[j];
      for (int l = 0; l < NTIMESINNER; l++) {
        for (register int i = 0; i < n; i++) aa[i] = bb[i] + scalar * cc[i];
        if (size == 65) printf("never printed %g\n", a[0][11]);
      }
    }
    times[k] = MPI_Wtime() - times[k];
  }
  for (int k = 1; k < NTIMES; k++) { /* note -- skip first iteration */
    mintime = MIN(mintime, times[k]);
  }
  rate = 1.0E-06 * bytes * NTIMESINNER / mintime;

  if (size == 1) {
    printf("%d %11.4f   Rate (MB/s) 1\n", size, rate);
    fd = fopen("flops", "w");
    fprintf(fd, "%g\n", rate);
    fclose(fd);
  } else {
    double prate;
    fd = fopen("flops", "r");
    fscanf(fd, "%lg", &prate);
    fclose(fd);
    printf("%d %11.4f   Rate (MB/s) %g\n", size, rate, rate / prate);
  }
  return 0;
}