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 M30 237
      10  #define M01 1322
      11  #define M11 13
      12  #define M21 27271
      13  #define M31 2280
      14  #define M02 74
      15  #define M12 191
      16  #define M22 500
      17  #define M32 111
      18  #define M03 134
      19  #define M13 117
      20  #define M23 11
      21  #define M33 771
      22  
      23  #define N 16
      24  
      25  void foo (unsigned int *__restrict__ pInput, unsigned int *__restrict__ pOutput)
      26  {
      27    unsigned int i, a, b, c, d;
      28  
      29    for (i = 0; i < N / 4; i++)
      30      {
      31         a = *pInput++;
      32         b = *pInput++;
      33         c = *pInput++;
      34         d = *pInput++;
      35  
      36         *pOutput++ = M00 * a + M01 * b + M02 * c + M03 * d;
      37         *pOutput++ = M10 * a + M11 * b + M12 * c + M13 * d;
      38         *pOutput++ = M20 * a + M21 * b + M22 * c + M23 * d;
      39         *pOutput++ = M30 * a + M31 * b + M32 * c + M33 * d;
      40      }
      41  }
      42  
      43  int main (int argc, const char* argv[])
      44  {
      45    unsigned int input[N], output[N], i;
      46    unsigned int check_results[N] = {1872, 746, 28304, 4815, 8392, 2894, 139524, 18411, 14912, 5042, 250744, 32007, 21432, 7190, 361964, 45603};
      47  
      48    check_vect ();
      49  
      50    for (i = 0; i < N; i++)
      51      {
      52        input[i] = i%256;
      53        output[i] = 0;
      54        __asm__ volatile ("");
      55      }
      56  
      57    foo (input, output);
      58  
      59    for (i = 0; i < N; i++)
      60      {
      61        if (output[i] != check_results[i])
      62  	abort ();
      63        __asm__ volatile ("");
      64      }
      65  
      66    return 0;
      67  }
      68  
      69  /* { dg-final { scan-tree-dump-times "vectorized 1 loops" 1 "vect"  { target vect_perm } } } */
      70  /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 1 "vect" { target { vect_perm && {! vect_load_lanes } } } } } */
      71  /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "vect" { target vect_load_lanes } } } */
      72  /* { dg-final { scan-tree-dump "Built SLP cancelled: can use load/store-lanes" "vect" { target { vect_perm && vect_load_lanes } } } } */
      73  /* { dg-final { scan-tree-dump "LOAD_LANES" "vect" { target vect_load_lanes } } } */
      74  /* { dg-final { scan-tree-dump "STORE_LANES" "vect" { target vect_load_lanes } } } */
      75