#include <unistd.h>
#include <stdio.h>
#include <sys/resource.h>
#include <sys/times.h>
#include <sys/types.h>
#include <time.h>
#include <limits.h>

#define CACHE_MIN (8*128)	/* smallest cache */
#define CACHE_MAX (8*1024*1024)	/* largest cache */
#define	SAMPLE	20		/* to get larger time sample */
//XXX #define	SAMPLE	10		/* to get larger time sample */

int	x[CACHE_MAX];		/* stride thru this array */

double
get_seconds () {		/* routine to read time */
    struct tms rusage;
    times(&rusage);	/* time in clock ticks */
    return (double) (rusage.tms_utime)/sysconf(_SC_CLK_TCK);
}

int
main () {
    int
    register	i, index, stride, limit, temp;
    long	steps, tsteps;
    int		csize;
    double	sec0, sec;	/* timing variables */

    for (csize=CACHE_MIN; csize <= CACHE_MAX; csize=csize*2) {
	for (stride=1; stride <= csize/2; stride=stride*2) {
	    sec = 0;		/* init timer */
	    limit = csize - stride + 1;	/* cache size this loop */
	    steps = 0;
	    do {		/* repeat until collected 1 sec */
		sec0 = get_seconds ();	/* start timer */
		for (i=SAMPLE*stride; i != 0; i=i-1) {	/* larger sample */
		    for (index=0; index < limit; index=index+stride)
			{ x[index] = x[index] + 1; }	/* cache access */
		}
		steps = steps + 1;	/* count while loop iterations */
		sec = sec + (get_seconds() - sec0);	/* end timer */
	    } while (sec < 1.0);

	    /* repeat empty loop to subtract loop overhead */
	    tsteps = 0;		/* used to match # while iterations */
	    do {		/* repeat until same # iterations as above */
		sec0 = get_seconds ();	/* start timer */
		for (i=SAMPLE*stride; i != 0; i=i-1) {	/* larger sample */
		    for (index=0; index < limit; index=index+stride)
			{ temp = temp + index; }	/* dummy code */
		}
		tsteps = tsteps + 1;	/* count while loop iterations */
		sec = sec - (get_seconds() - sec0);	/* end timer */
	    } while (tsteps < steps);

	    printf ("Size:%7d Stride:%7d read+write:%10.3f ns, sec = %6.3f, steps = %6.0f\n",
			csize*sizeof(int), stride*sizeof(int),
			(double) sec*1e9/(steps*SAMPLE*
					stride*((limit-1)/stride+1)),
			sec, (double) steps);
	    fflush(stdout);
	}
	printf ("\n\n");
    }
    return 0;
}