clflushopt.c

/*
             clflush 4278.300100[MiB/s]
          clflushopt 44717.389375[MiB/s]
              memset 30407.684046[MiB/s]
*/

#include <stdio.h>
#include <x86intrin.h>
#include <sys/time.h>
#include <string.h>
#include <time.h>

double
sec(void)
{
    struct timespec ts;
    clock_gettime(CLOCK_MONOTONIC, &ts);

    return ts.tv_sec + (ts.tv_nsec / 1000000000.0);
}

unsigned char *data;

void __attribute__((noinline,noclone))
clflush(size_t nline)
{
    int li;
    for (li=0; li<nline; li++) {
        _mm_clflush(data + li*64);
    }

    asm volatile ("" ::: "memory");
}

void __attribute__((noinline,noclone))
clflush_opt(size_t nline)
{
    int li;
    for (li=0; li<nline; li++) {
        _mm_clflushopt(data + li*64);
    }

    asm volatile ("" ::: "memory");
}

void __attribute__((noinline,noclone))
run_memset(size_t nline)
{
    int li;
    memset(data, 0, nline * 64);
    asm volatile ("" ::: "memory");
}

int
main(int argc, char **argv)
{
    double t0, t1, total;
    size_t size = 128 * 1024*1024;
    void *p;
    int li;
    int nloop = 32;

    if (argc > 1) {
        size = atoi(argv[1]) * 1024*1024;
    }

    posix_memalign(&p, 4096, size);
    data = p;

    clflush(size/64);
    clflush_opt(size/64);

    t0 = sec();
    for (li=0; li<nloop; li++) {
        clflush(size/64);
    }
    t1 = sec();

    total = size * (double)nloop;
    printf("%20s %f[MiB/s]\n", "clflush", total / ((t1-t0) * 1024*1024));

    t0 = sec();
    for (li=0; li<nloop; li++) {
        clflush_opt(size/64);
    }
    t1 = sec();

    total = size * (double)nloop;
    printf("%20s %f[MiB/s]\n", "clflushopt", total / ((t1-t0) * 1024*1024));


    t0 = sec();
    for (li=0; li<nloop; li++) {
        run_memset(size/64);
    }
    t1 = sec();

    total = size * (double)nloop;
    printf("%20s %f[MiB/s]\n", "memset", total / ((t1-t0) * 1024*1024));
    
}