/******************************************************************************
* FILE: arraypad.c
* DESCRIPTION:
*   Demonstrates effect of array padding
******************************************************************************/
#include <stdio.h>
#include <sys/time.h>
#include <time.h>
#define SIZE	256

double a[SIZE][SIZE], aa[SIZE][SIZE+1];
struct timeval start_time, end_time;

main()
{
   int i, j, k;
   
/*****************************************************************************/
/*				Initialize       		     */
/*****************************************************************************/
   for(i=0;i<SIZE;i++)
     for(j=0;j<SIZE;j++)
       aa[i][j] = a[i][j] = 1.23;

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

   for(j=0;j<SIZE;j++)
     for(i=0;i<SIZE;i++)
       for(k=0;k<SIZE;k++)
         a[i][j] = a[i][j] + a[j][k] - a[k][i];

   gettimeofday(&end_time, (struct timeval*)0);  
   printf("Untuned version took: ");
   print_execution_times();

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

   for(j=0;j<SIZE;j++)
     for(i=0;i<SIZE;i++)
       for(k=0;k<SIZE;k++)
         aa[i][j] = aa[i][j] + aa[j][k] - aa[k][i];

   gettimeofday(&end_time, (struct timeval*)0); 
   printf("Tuned version took: ");
   print_execution_times();
   finalize_data();
}


finalize_data()
{
   int i, j;
   FILE *fp, *fopen();
   fp = fopen("/dev/null", "w");
   for(j=0;j<SIZE;j++) {
     for(i=0;i<SIZE;i++) {
       fprintf(fp, "%f\n", a[i][j]);
       fprintf(fp, "%f\n", aa[i][j]);
       }
     }  
   fclose(fp);
}

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