#include <iostream>
#include <cstdio>
#include <cstdlib>
#include <string.h>
using namespace std;

#include "MOA.h"
#include <emmintrin.h>

int main(int argc, char **argv)
{
  int print;
  int simd;
  int bytes;
  int i;

  uint16_t *buf1;  

  __m128i a, b, x, *c;


  /* usage: odd_shorts_time bytes SIMD(Y|N|none) seed print(Y|N) */

  if (argc != 5 || (strcmp(argv[2], "Y") != 0 && strcmp(argv[2], "N") != 0 &&
                    strcmp(argv[2], "none") != 0)) {
    fprintf(stderr, "usage: odd_shorts_time bytes SIMD(Y|N|none) seed print(Y|N)\n");
    exit(1);
  }

  /* Set the variables from the command line */

  bytes = atoi(argv[1]);
  if (bytes % 16 != 0) { fprintf(stderr, "bytes mod 16 must = 0\n"); exit(1); }
  simd = argv[2][0];
  print = (argv[4][0] == 'Y');
  MOA_Seed(atoi(argv[3]));
  
  /* Allocate the buffer. */

  buf1 = (uint16_t *) malloc(bytes);

  /* Use MOA.c to fill buf1 and buf2 with random bytes */

  MOA_Fill_Random_Region (buf1, bytes);

  if (print) {
    for (i = 0; i < bytes/2; i++) {
      printf(" %04x", buf1[i]);
      if (i % 8 == 7) printf("\n");
    }
    printf("\n");
  }

  /* This lets you access the buffer a SIMD vector. */

  c = (__m128i *) buf1;

  /* Use SIMD to isolate the odd shorts. */

  if (simd == 'Y') {
    for (i = 0; i < bytes; i += 16) {
      b = *c;
      x = _mm_set1_epi16(1);
      a = _mm_and_si128(x, b);
      x = _mm_setzero_si128();
      a = _mm_sub_epi16(x, a);
      b = _mm_and_si128(a, b);
      *c = b;
      c++;
    }

  /* Use regular code to isolate the odd shorts. */

  } else if (simd == 'N') {
    for (i = 0; i < bytes/2; i++) {
      if (buf1[i] % 2 == 0) buf1[i] = 0;
    }

  }

  /* Print everything out. */

  if (print) {
    for (i = 0; i < bytes/2; i++) {
      printf(" %04x", buf1[i]);
      if (i % 8 == 7) printf("\n");
    }
  }
  return 0;
}