gcc/testsuite/c-c++-common/goacc/kernels-reduction.c

   1 /* { dg-additional-options "-O2" } */
   2 /* { dg-additional-options "-fdump-tree-parloops1-all" } */
   3 /* { dg-additional-options "-fdump-tree-optimized" } */
   4
   5 #include <stdlib.h>
   6
   7 #define n 10000
   8
   9 unsigned int a[n];
  10
  11 void  __attribute__((noinline,noclone))
  12 foo (void)
  13 {
  14   int i;
  15   unsigned int sum = 1;
  16
  17 #pragma acc kernels copyin (a[0:n]) copy (sum)
  18   {
  19     for (i = 0; i < n; ++i)
  20       sum += a[i];
  21   }
  22
  23   if (sum != 5001)
  24     abort ();
  25 }
  26
  27 /* Check that only one loop is analyzed, and that it can be parallelized.  */
  28 /* { dg-final { scan-tree-dump-times "SUCCESS: may be parallelized" 1 "parloops1" } } */
  29 /* { dg-final { scan-tree-dump-times "(?n)__attribute__\\(\\(oacc kernels parallelized, oacc function \\(, , \\), oacc kernels, omp target entrypoint\\)\\)" 1 "parloops1" } } */
  30 /* { dg-final { scan-tree-dump-not "FAILED:" "parloops1" } } */
  31
  32 /* Check that the loop has been split off into a function.  */
  33 /* { dg-final { scan-tree-dump-times "(?n);; Function .*foo.*._omp_fn.0" 1 "optimized" } } */