* FILE: loop10.c
* DESCRIPTION:
*   Demonstrates loop collapse
******************************************************************************/
#include <stdio.h>
#include <sys/time.h>
#include <time.h>
#define	SEED		1995
#define	DIM1		500
#define	DIM2		80
#define	DIM3		4

float a[DIM1][DIM2][DIM3], aa[DIM1][DIM2][DIM3];
float b[DIM1][DIM2][DIM3], bb[DIM1][DIM2][DIM3];
float c[DIM1][DIM2][DIM3], cc[DIM1][DIM2][DIM3];
struct timeval start_time, end_time;

main()
{
   int i, j, k;
   initialize_data();

/*****************************************************************************/
/*				Uncollapsed Loop			     */
/*****************************************************************************/
   gettimeofday(&start_time, (struct timeval*)0);

   for(i=0;i<DIM1;i++)
      for(j=0;j<DIM2;j++)
         for(k=0;k<DIM3;k++)
            a[i][j][k] = a[i][j][k] + b[i][j][k] * c[i][j][k];

   gettimeofday(&end_time, (struct timeval*)0);  /* after time */
   printf("Execution time for the uncollapsed loop:\n");
   print_execution_times();

/*****************************************************************************/
/*				Collapsed Loop				     */
/*****************************************************************************/
   gettimeofday(&start_time, (struct timeval*)0);

   for(i=0;i<(DIM1 * DIM2 * DIM3);i++)
      aa[0][0][i] = aa[0][0][i] + bb[0][0][i] * cc[0][0][i];

   gettimeofday(&end_time, (struct timeval*)0);  /* after time */
   printf("Execution time for the collapsed loop:\n");
   print_execution_times();
   finalize_data();
}

initialize_data()
{
   int i, j, k;
   srandom(SEED);
   for(i=0;i<DIM1;i++)
      for(j=0;j<DIM2;j++)
         for(k=0;k<DIM3;k++)
         {
            a[i][j][k] = aa[i][j][k] = (float) (random() % 100000);
            b[i][j][k] = bb[i][j][k] = (float) (random() % 100000);
            c[i][j][k] = cc[i][j][k] = (float) (random() % 100000);
         }
}

finalize_data()
{
   int i, j, k;
   FILE *fp, *fopen();

   fp = fopen("/dev/null", "w");
   i = random() % DIM1;
   j = random() % DIM2;
   k = random() % DIM3;
   fprintf(fp, "%f\n", a[i][j][k]);
   fprintf(fp, "%f\n", aa[i][j][k]);
   fclose(fp);
}

print_execution_times()
{
   int total_usecs;
   float total_time;
   total_usecs = (end_time.tv_sec-start_time.tv_sec) * 1000000 +
         (end_time.tv_usec-start_time.tv_usec);
   printf("Total time was %.2f mSec\n", ((float) total_usecs) / 1000.0);
   total_time = ((float) total_usecs) / 1000000.0;
}