slp-perm-4.c

   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 M40 437
  11
  12 #define M01 1322
  13 #define M11 13
  14 #define M21 27271
  15 #define M31 2280
  16 #define M41 284
  17
  18 #define M02 74
  19 #define M12 191
  20 #define M22 500
  21 #define M32 111
  22 #define M42 1114
  23
  24 #define M03 134
  25 #define M13 117
  26 #define M23 11
  27 #define M33 771
  28 #define M43 71
  29
  30 #define M04 334
  31 #define M14 147
  32 #define M24 115
  33 #define M34 7716
  34 #define M44 16
  35
  36 #define N 16
  37
  38 void foo (unsigned int *__restrict__ pInput, unsigned int *__restrict__ pOutput)
  39 {
  40   unsigned int i, a, b, c, d, e;
  41
  42   for (i = 0; i < N / 5; i++)
  43     {
  44        a = *pInput++;
  45        b = *pInput++;
  46        c = *pInput++;
  47        d = *pInput++;
  48        e = *pInput++;
  49
  50        *pOutput++ = M00 * a + M01 * b + M02 * c + M03 * d + M04 * e;
  51        *pOutput++ = M10 * a + M11 * b + M12 * c + M13 * d + M14 * e;
  52        *pOutput++ = M20 * a + M21 * b + M22 * c + M23 * d + M24 * e;
  53        *pOutput++ = M30 * a + M31 * b + M32 * c + M33 * d + M34 * e;
  54        *pOutput++ = M40 * a + M41 * b + M42 * c + M43 * d + M44 * e;
  55     }
  56 }
  57
  58 int main (int argc, const char* argv[])
  59 {
  60   unsigned int input[N], output[N], i;
  61   unsigned int check_results[N] = {3208, 1334, 28764, 35679, 2789, 13028, 4754, 168364, 91254, 12399, 22848, 8174, 307964, 146829, 22009, 0};
  62
  63   check_vect ();
  64
  65   for (i = 0; i < N; i++)
  66     {
  67       input[i] = i%256;
  68       if (input[i] > 200)
  69         abort();
  70       output[i] = 0;
  71       __asm__ volatile ("");
  72     }
  73
  74   foo (input, output);
  75
  76   for (i = 0; i < N; i++)
  77     if (output[i] != check_results[i])
  78       abort ();
  79
  80   return 0;
  81 }
  82
  83 /* { dg-final { scan-tree-dump-times "vectorized 1 loops" 0 "vect"  } } */
  84 /* { dg-final { scan-tree-dump-times "permutation requires at least three vectors" 1 "vect" { target vect_perm } } } */
  85 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "vect"  } } */
  86 /* { dg-final { cleanup-tree-dump "vect" } } */
  87