#include <emmintrin.h>
#include <math.h>
#include <stdio.h>
#include <sys/time.h>


long qftolsse(float f)
{
  long retval;

  __asm__ volatile
  (
    "cvttss2si %1, %0\n"
    : "=r" (retval)
    : "x" (f)
  );

  return retval;
}



int main(int argc, char** argv)
{
    if (argc != 2)
        return 1;
    int N = atoi(argv[1]);
    float* a;
    posix_memalign((void**)&a, 16,  N * sizeof(float));

    //float* b;
    //posix_memalign((void**)&b, 16,  N * sizeof(float));

    long* res1 = malloc(N*sizeof(long));
    long* res2 = malloc(N*sizeof(long));


    for (int i = 0; i < N; ++i)
       a[i] = i + 0.5;

    {
        struct timeval t1,t2;
        double timeuse;
        gettimeofday(&t1, NULL);

        for (int i = 0; i < N; ++i)
        {
            res1[i] = qftolsse(a[i]);
        }

        gettimeofday(&t2, NULL);
        timeuse = t2.tv_sec - t1.tv_sec + (t2.tv_usec - t1.tv_usec)/1000000.0;
        printf("qftolsse Use Time:%f\n", timeuse);
    }
    printf("res1[0]=%ld, res1[%d]=%ld\n", res1[0], N-1 ,res1[N-1]);


    {
        struct timeval t1,t2;
        double timeuse;
        gettimeofday(&t1, NULL);
        for (int i = 0; i < N; ++i)
        {
            res2[i] = (long)(a[i]);
        }
        gettimeofday(&t2, NULL);
        timeuse = t2.tv_sec - t1.tv_sec + (t2.tv_usec - t1.tv_usec)/1000000.0;
        printf("cast     Use Time:%f\n", timeuse);
    }

    printf("res2[0]=%ld, res2[%d]=%ld\n", res2[0], N-1 ,res2[N-1]);

}