/************************************************************************************
*                    Micro Benchmark   mbcsb
*                 -----------------------------   
* File name: mbcsb.c
* Author   : Yuanhua Yu
* Date     : 20/12/1999.
*
*-------------------------------------------------------------------------------------
*
*  Function:  Measuring the subblock size of Cache Level-I and Level-II
*
*   Methods:  Sequential array read access
*
* Condition:  Running under Linux Operating System
*
*   History:  First version : 05/10/1999
*             Second version: 20/12/1999
*
*Parameters:  STRIDE_START ---- intinial value of the stride : 1      (*4bytes)      
*             STRIDE_END   ---- max value of the stride      : 1024   (*4bytes)
*             TOTAL_ACCESS ---- number of the accesses done  : (32768*1024*16)
*
*  Variable:  cache_size   ---- capacity of the Level-I or Level-II  cache in KB
*             arr_size     ---- size of the array accessed;
*             arr_item     ---- number of elements in the array accessed;
*             stride       ---- increment of stride
*
*************************************************************************************/

#include <stdlib.h>
#include <stdio.h>
#include <math.h>
#include <time.h>
#include <malloc.h>
#include <string.h>

#define STRIDE_START      1                 //*4 bytes
#define STRIDE_END        1024              //*4 bytes
#define TOTAL_ACCESS      (1024*1024*32/4)  //operations
#define CACHE_SIZE_I      (1024*16)         //bytes
#define CACHE_SIZE_II     (1024*512)        //bytes
#define ATYPE long

/*------------------------------------------------------------------------------------*/

int main(char **av, int ac)
{
    volatile ATYPE  *va;
    register ATYPE  s1;
    register long	i;
    unsigned long start_time, stop_time;
    long          arr_size, arr_item, stride;
    double	     time_cost, cost_per_op;

    FILE *fp;

    fp = fopen("Subblocksize.dat","a+");
    fprintf(fp,"--------- Measurement of Cache Subblock size ---------\n");
    fprintf(fp,"Array_size(bytes)  Access_T(ns)  \tTotal_T(s)\n");
    printf("Step_size(byte)       Access_T(ns)  \tTotal_T(s)\n");

	arr_size = 4*CACHE_SIZE_II;
	arr_item = arr_size /  sizeof(ATYPE);
	va = (ATYPE *)calloc(arr_item,sizeof(ATYPE));

	for (stride=STRIDE_START; stride<STRIDE_END; stride*=2)
	{
          for(i=0; i<arr_item; i+=stride)
              va[i] = i + stride;
	      va[i-stride] = 0;

	      s1 = 0;
	      start_time = clock();  /* ---------- start clock tickle */
	      for (i=0; i<TOTAL_ACCESS; i+=8) 
	      {	
	           s1=va[s1]; s1=va[s1]; s1=va[s1]; s1=va[s1];    
	           s1=va[s1]; s1=va[s1]; s1=va[s1]; s1=va[s1];    
	      }
	      stop_time = clock();   /* ----------- stop clock ticklw */	

	      time_cost = ((double)(stop_time-start_time))/((double)CLOCKS_PER_SEC);
	      cost_per_op  = (1000000000.0 * time_cost)/TOTAL_ACCESS;

          printf("%10d\t\t%6.3f\t\t%6.3f %d\n",
       	          stride*sizeof(ATYPE),cost_per_op,time_cost,s1);
	      fprintf(fp,"%10d\t\t%6.3f\t\t%6.3f\n",
	 	          stride*sizeof(ATYPE),cost_per_op,time_cost);
       }

       fclose(fp);
       return 0;
}