slp-33.c

   1 /* { dg-require-effective-target vect_int } */
   2
   3 #include <stdarg.h>
   4 #include "tree-vect.h"
   5
   6 #define N 8
   7
   8 int
   9 main1 ()
  10 {
  11   int i;
  12   unsigned int out[N*8], a0, a1, a2, a3, a4, a5, a6, a7, b1, b0, b2, b3, b4, b5, b6, b7;
  13   unsigned int in[N*8] = {0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63};
  14   float out2[N*8];
  15
  16   /* SLP with unrolling by 4.  */
  17   for (i = 0; i < N; i++)
  18     {
  19       a0 = in[i*7] + 5;
  20       a1 = in[i*7 + 1] + 6;
  21       a2 = in[i*7 + 2] + 7;
  22       a3 = in[i*7 + 3] + 8;
  23       a4 = in[i*7 + 4] + 9;
  24       a5 = in[i*7 + 5] + 10;
  25       a6 = in[i*7 + 6] + 11;
  26
  27       b0 = a0 * 3;
  28       b1 = a1 * 2;
  29       b2 = a2 * 12;
  30       b3 = a3 * 5;
  31       b4 = a4 * 8;
  32       b5 = a5 * 4;
  33       b6 = a6 * 3;
  34
  35       out[i*7] = b0 - 2;
  36       out[i*7 + 1] = b1 - 3;
  37       out[i*7 + 2] = b2 - 2;
  38       out[i*7 + 3] = b3 - 1;
  39       out[i*7 + 4] = b4 - 8;
  40       out[i*7 + 5] = b5 - 7;
  41       out[i*7 + 6] = b6 - 3;
  42     }
  43
  44   /* check results:  */
  45   for (i = 0; i < N; i++)
  46     {
  47       if (out[i*7] !=  (in[i*7] + 5) * 3 - 2
  48          || out[i*7 + 1] != (in[i*7 + 1] + 6) * 2 - 3
  49          || out[i*7 + 2] != (in[i*7 + 2] + 7) * 12 - 2
  50          || out[i*7 + 3] != (in[i*7 + 3] + 8) * 5 - 1
  51          || out[i*7 + 4] != (in[i*7 + 4] + 9) * 8 - 8
  52          || out[i*7 + 5] != (in[i*7 + 5] + 10) * 4 - 7
  53          || out[i*7 + 6] != (in[i*7 + 6] + 11) * 3 - 3)
  54         abort ();
  55     }
  56
  57   /* SLP with unrolling by 4.  */
  58   for (i = 0; i < N*2; i++)
  59     {
  60       out[i*3] = (in[i*3] + 2) * 3;
  61       out[i*3 + 1] = (in[i*3 + 1] + 2) * 7;
  62       out[i*3 + 2] = (in[i*3 + 2] + 7) * 3;
  63     }
  64
  65   /* check results:  */
  66   for (i = 0; i < N*2; i++)
  67     {
  68       if (out[i*3] !=  (in[i*3] + 2) * 3
  69          || out[i*3 + 1] != (in[i*3 + 1] + 2) * 7
  70          || out[i*3 + 2] != (in[i*3 + 2] + 7) * 3)
  71         abort ();
  72     }
  73
  74   /* SLP with unrolling by 4.  */
  75   for (i = 0; i < N*2; i++)
  76     {
  77       out2[i*3] = (float) (in[i*3] * 2 + 5) ;
  78       out2[i*3 + 1] = (float) (in[i*3 + 1] * 3 + 7);
  79       out2[i*3 + 2] = (float) (in[i*3 + 2] * 5 + 4);
  80     }
  81
  82   /* check results:  */
  83   for (i = 0; i < N*2; i++)
  84     {
  85       if (out2[i*3] !=  (float) (in[i*3] * 2 + 5)
  86          || out2[i*3 + 1] != (float) (in[i*3 + 1] * 3 + 7)
  87          || out2[i*3 + 2] != (float) (in[i*3 + 2] * 5 + 4))
  88         abort ();
  89     }
  90
  91
  92   return 0;
  93 }
  94
  95 int main (void)
  96 {
  97   check_vect ();
  98
  99   main1 ();
 100
 101   return 0;
 102 }
 103
 104 /* { dg-final { scan-tree-dump-times "vectorized 3 loops" 1 "vect"  {target {vect_uintfloat_cvt && vect_int_mult} } } } */
 105 /* { dg-final { scan-tree-dump-times "vectorized 2 loops" 1 "vect"  {target {{! { vect_uintfloat_cvt}} && vect_int_mult} } } } */
 106 /* { dg-final { scan-tree-dump-times "vectorized 0 loops" 1 "vect"  {target {{! { vect_uintfloat_cvt}} && {! {vect_int_mult}}} } } } */
 107 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 3 "vect" {target {vect_uintfloat_cvt && vect_int_mult} } } } */
 108 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 2 "vect"  {target {{! { vect_uintfloat_cvt}} && vect_int_mult} } } } */
 109 /* { dg-final { scan-tree-dump-times "vectorizing stmts using SLP" 0 "vect"  {target {{! { vect_uintfloat_cvt}} && {! {vect_int_mult}}} } } } */
 110 /* { dg-final { cleanup-tree-dump "vect" } } */
 111