1  /* { dg-require-effective-target vect_int } */
       2  
       3  #include <stdarg.h>
       4  #include "tree-vect.h"
       5  
       6  #define M00 100
       7  #define M10 216
       8  #define M20 23
       9  #define M01 1322
      10  #define M11 13
      11  #define M21 27271
      12  #define M02 74
      13  #define M12 191
      14  #define M22 500
      15  
      16  #if VECTOR_BITS > 128
      17  #define N (VECTOR_BITS * 3 / 32 + 4)
      18  #else
      19  #define N 16
      20  #endif
      21  
      22  void foo (unsigned int *__restrict__ pInput, unsigned int *__restrict__ pOutput)
      23  {
      24    unsigned int i, a, b, c;
      25  
      26    for (i = 0; i < N / 3; i++)
      27      {
      28         a = *pInput++;
      29         b = *pInput++;
      30         c = *pInput++;
      31  
      32         *pOutput++ = M00 * a + M01 * b + M02 * c;
      33         *pOutput++ = M10 * a + M11 * b + M12 * c;
      34         *pOutput++ = M20 * a + M21 * b + M22 * c;
      35      }
      36  }
      37  
      38  int main (int argc, const char* argv[])
      39  {
      40    unsigned int input[N], output[N], i;
      41  
      42    check_vect ();
      43  
      44    for (i = 0; i < N; i++)
      45      {
      46        input[i] = i%256;
      47        output[i] = 0;
      48        __asm__ volatile ("");
      49      }
      50  
      51  #if N == 16
      52    unsigned int check_results[N] = {1470, 395, 28271, 5958, 1655, 111653, 10446, 2915, 195035, 14934, 4175, 278417, 19422, 5435, 361799, 0};
      53  #else
      54    volatile unsigned int check_results[N] = {};
      55  
      56    for (unsigned int i = 0; i < N / 3; i++)
      57      {
      58        unsigned int a = input[i * 3];
      59        unsigned int b = input[i * 3 + 1];
      60        unsigned int c = input[i * 3 + 2];
      61  
      62        check_results[i * 3] = M00 * a + M01 * b + M02 * c;
      63        check_results[i * 3 + 1] = M10 * a + M11 * b + M12 * c;
      64        check_results[i * 3 + 2] = M20 * a + M21 * b + M22 * c;
      65  
      66        asm volatile ("" ::: "memory");
      67      }
      68  #endif
      69  
      70    foo (input, output);
      71  
      72    for (i = 0; i < N; i++)
      73      {
      74        if (output[i] != check_results[i])
      75  	abort ();
      76        __asm__ volatile ("");
      77      }
      78  
      79    return 0;
      80  }
      81  
      82  /* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect"  { target vect_perm } } } */
      83  /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "vect" { target { vect_perm3_int && {! vect_load_lanes } } } } } */
      84  /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "vect" { target vect_load_lanes } } } */
      85  /* { dg-final { scan-tree-dump "Built SLP cancelled: can use load/store-lanes" "vect" { target { vect_perm3_int && vect_load_lanes } } } } */
      86  /* { dg-final { scan-tree-dump "LOAD_LANES" "vect" { target vect_load_lanes } } } */
      87  /* { dg-final { scan-tree-dump "STORE_LANES" "vect" { target vect_load_lanes } } } */
      88