1  /* { dg-do run } */
       2  /* { dg-require-effective-target ssse3 } */
       3  /* { dg-options "-O2 -fno-strict-aliasing -mssse3" } */
       4  
       5  #ifndef CHECK_H
       6  #define CHECK_H "ssse3-check.h"
       7  #endif
       8  
       9  #ifndef TEST
      10  #define TEST ssse3_test
      11  #endif
      12  
      13  #include CHECK_H
      14  
      15  #include "ssse3-vals.h"
      16  
      17  #include <tmmintrin.h>
      18  
      19  /* Test the 64-bit form */
      20  static void
      21  ssse3_test_pmulhrsw (int *i1, int *i2, int *r)
      22  {
      23    __m64 t1 = *(__m64 *) i1;
      24    __m64 t2 = *(__m64 *) i2;
      25    *(__m64 *) r = _mm_mulhrs_pi16 (t1, t2);
      26    _mm_empty ();
      27  }
      28  
      29  /* Test the 128-bit form */
      30  static void
      31  ssse3_test_pmulhrsw128 (int *i1, int *i2, int *r)
      32  {
      33    /* Assumes incoming pointers are 16-byte aligned */
      34    __m128i t1 = *(__m128i *) i1;
      35    __m128i t2 = *(__m128i *) i2;
      36    *(__m128i *) r = _mm_mulhrs_epi16 (t1, t2);
      37  }
      38  
      39  /* Routine to manually compute the results */
      40  static void
      41  compute_correct_result (int *i1, int *i2, int *r)
      42  {
      43    short *s1 = (short *) i1;
      44    short *s2 = (short *) i2;
      45    short *sout = (short *) r;
      46    int t0;
      47    int i;
      48  
      49    for (i = 0; i < 8; i++)
      50      {
      51        t0 = (((int) s1[i] * (int) s2[i]) >> 14) + 1;
      52        sout[i] = (short) (t0 >> 1);
      53      }
      54  }
      55  
      56  static void
      57  TEST (void)
      58  {
      59    int i;
      60    int r [4] __attribute__ ((aligned(16)));
      61    int ck [4];
      62    int fail = 0;
      63  
      64    for (i = 0; i < 256; i += 8)
      65      {
      66        /* Manually compute the result */
      67        compute_correct_result (&vals[i + 0], &vals[i + 4], ck);
      68  
      69        /* Run the 64-bit tests */
      70        ssse3_test_pmulhrsw (&vals[i + 0], &vals[i + 4], &r[0]);
      71        ssse3_test_pmulhrsw (&vals[i + 2], &vals[i + 6], &r[2]);
      72        fail += chk_128 (ck, r);
      73  
      74        /* Run the 128-bit tests */
      75        ssse3_test_pmulhrsw128 (&vals[i + 0], &vals[i + 4], r);
      76        fail += chk_128 (ck, r);
      77      }
      78  
      79    if (fail != 0)
      80      abort ();
      81  }