/* Last edited on 2013-10-02 03:50:32 by stolfilocal */

#include <stdio.h>
#include <stdint.h>
#include <stdlib.h>
#include <assert.h>

#include <jstime.h>
#include <jsmath.h>

#define TYPE int32_t

void test(int m, int n);

#define int64_dec_fmt "ld"

void test(int m, int n)
{
  int j, i;
  double start;
  TYPE x = 31, y = 17;
  int64_t xx, yy;
  
  start = user_cpu_time_usec();
  for (j = 0; j < n; j++)
    { for (i = 0; i < m; i++) 
        { /* Ops to avoid the loop being optimized out of existence: */
          x = y + i; y = x + j;
        }
    }
  double tare = user_cpu_time_usec() - start; /* Time for "empty" loop (us). */
  xx = x; yy = y; fprintf(stderr, "x = %ld y = %ld\n", xx, yy);

  start = user_cpu_time_usec();
  for (j = 0; j < n; j++)
    { for (i = 0; i < m; i++)
        { /* Ops to avoid the loop being optimized out of existence: */
          x = y + i; y = x + j;
          /* Ops being timed: */
          x = y + i; y = x + j;
          x = y + i; y = x + i;
          x = y + x; y = x + j;
          x = y + j; y = x + y;
          x = y + i; y = x + i;
          x = y + j; y = x + y;
        }
    }
  double toto = user_cpu_time_usec() - start; /* Time for loop with timed ops (us). */
  xx = x; yy = y; fprintf(stderr, "x = %" int64_dec_fmt " y = %" int64_dec_fmt "\n", xx, yy);

  int nops = 12;                               /* Extra ops in inner loop 2. */

  double ctop = ((double)n)*((double)m)*nops; /* Total number of ops executed. */
  double tmop = (toto - tare)/ctop;           /* Time per op (us) */
  double vcpu = 1.0/tmop;                     /* Million ops per second. */
  
  fprintf(stderr, "tare = %16.12f sec\n", tare/1e6);
  fprintf(stderr, "toto = %16.12f sec\n", toto/1e6);
  fprintf(stderr, "tmop = %16.12f usec/op\n", tmop);
  fprintf(stderr, "vcpu = %16.1f Mops/sec\n", vcpu);
}

int main(int argc, char **argv)
{
  int m = 10000000;
  int n = 1000;
  fprintf(stderr, "start  m = %d n = %d\n", m, n);
  test(m, n);
  fprintf(stderr, "stop\n");
  return 0;
}