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 (1024 * 512)
8 #define COUNTERTYPE unsigned int
9
10 int
main(void)11 main (void)
12 {
13 unsigned int *__restrict a;
14 unsigned int *__restrict b;
15 unsigned int *__restrict c;
16
17 a = (unsigned int *)malloc (N * sizeof (unsigned int));
18 b = (unsigned int *)malloc (N * sizeof (unsigned int));
19 c = (unsigned int *)malloc (N * sizeof (unsigned int));
20
21 #pragma acc data copyout (a[0:N], b[0:N], c[0:N])
22 {
23 #pragma acc kernels present (a[0:N])
24 {
25 for (COUNTERTYPE i = 0; i < N; i++)
26 a[i] = i * 2;
27 }
28
29 #pragma acc kernels present (b[0:N])
30 {
31 for (COUNTERTYPE i = 0; i < N; i++)
32 b[i] = i * 4;
33 }
34
35 #pragma acc kernels present (a[0:N], b[0:N], c[0:N])
36 {
37 for (COUNTERTYPE ii = 0; ii < N; ii++)
38 c[ii] = a[ii] + b[ii];
39 }
40 }
41
42 for (COUNTERTYPE i = 0; i < N; i++)
43 if (c[i] != a[i] + b[i])
44 abort ();
45
46 free (a);
47 free (b);
48 free (c);
49
50 return 0;
51 }
52
53 /* Check that only three loops are analyzed, and that all can be
54 parallelized. */
55 /* { dg-final { scan-tree-dump-times "SUCCESS: may be parallelized" 3 "parloops1" } } */
56 /* { dg-final { scan-tree-dump-times "(?n)__attribute__\\(\\(oacc kernels parallelized, oacc function \\(, , \\), oacc kernels, omp target entrypoint\\)\\)" 3 "parloops1" } } */
57 /* { dg-final { scan-tree-dump-not "FAILED:" "parloops1" } } */
58
59 /* Check that the loop has been split off into a function. */
60 /* { dg-final { scan-tree-dump-times "(?n);; Function .*main._omp_fn.0" 1 "optimized" } } */
61 /* { dg-final { scan-tree-dump-times "(?n);; Function .*main._omp_fn.1" 1 "optimized" } } */
62 /* { dg-final { scan-tree-dump-times "(?n);; Function .*main._omp_fn.2" 1 "optimized" } } */
63