1  /* { dg-do run } */
       2  /* { dg-require-effective-target arm_neon_hw } */
       3  /* { dg-options "-O -ftree-vectorize" } */
       4  /* { dg-add-options arm_neon } */
       5  
       6  void fill (short *buf) __attribute__ ((noinline));
       7  void fill (short *buf)
       8  {
       9    int i;
      10  
      11    for (i = 0; i < 11 * 8; i++)
      12      buf[i] = i;
      13  }
      14  
      15  void test (unsigned char *dst) __attribute__ ((noinline));
      16  void test (unsigned char *dst)
      17  {
      18    short tmp[11 * 8], *tptr;
      19    int i;
      20  
      21    fill (tmp);
      22  
      23    tptr = tmp;
      24    for (i = 0; i < 8; i++)
      25      {
      26        dst[0] = (-tptr[0] + 9 * tptr[0 + 1] + 9 * tptr[0 + 2] - tptr[0 + 3]) >> 7;
      27        dst[1] = (-tptr[1] + 9 * tptr[1 + 1] + 9 * tptr[1 + 2] - tptr[1 + 3]) >> 7;
      28        dst[2] = (-tptr[2] + 9 * tptr[2 + 1] + 9 * tptr[2 + 2] - tptr[2 + 3]) >> 7;
      29        dst[3] = (-tptr[3] + 9 * tptr[3 + 1] + 9 * tptr[3 + 2] - tptr[3 + 3]) >> 7;
      30        dst[4] = (-tptr[4] + 9 * tptr[4 + 1] + 9 * tptr[4 + 2] - tptr[4 + 3]) >> 7;
      31        dst[5] = (-tptr[5] + 9 * tptr[5 + 1] + 9 * tptr[5 + 2] - tptr[5 + 3]) >> 7;
      32        dst[6] = (-tptr[6] + 9 * tptr[6 + 1] + 9 * tptr[6 + 2] - tptr[6 + 3]) >> 7;
      33        dst[7] = (-tptr[7] + 9 * tptr[7 + 1] + 9 * tptr[7 + 2] - tptr[7 + 3]) >> 7;
      34  
      35        dst += 8;
      36        tptr += 11;
      37      }
      38  }
      39  
      40  int main (void)
      41  {
      42    char buf [8 * 8];
      43  
      44    test (buf);
      45  
      46    return 0;
      47  }
      48