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)11main (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 kernels copyout (a[0:N]) 22 { 23 #ifdef ACC_LOOP 24 #pragma acc loop 25 #endif 26 for (COUNTERTYPE i = 0; i < N; i++) 27 a[i] = i * 2; 28 } 29 30 #pragma acc kernels copyout (b[0:N]) 31 { 32 #ifdef ACC_LOOP 33 #pragma acc loop 34 #endif 35 for (COUNTERTYPE i = 0; i < N; i++) 36 b[i] = i * 4; 37 } 38 39 #pragma acc kernels copyin (a[0:N], b[0:N]) copyout (c[0:N]) 40 { 41 #ifdef ACC_LOOP 42 #pragma acc loop 43 #endif 44 for (COUNTERTYPE ii = 0; ii < N; ii++) 45 c[ii] = a[ii] + b[ii]; 46 } 47 48 for (COUNTERTYPE i = 0; i < N; i++) 49 if (c[i] != a[i] + b[i]) 50 abort (); 51 52 free (a); 53 free (b); 54 free (c); 55 56 return 0; 57 } 58 59 /* Check that only three loops are analyzed, and that all can be 60 parallelized. */ 61 /* { dg-final { scan-tree-dump-times "SUCCESS: may be parallelized" 3 "parloops1" } } */ 62 /* { dg-final { scan-tree-dump-times "(?n)__attribute__\\(\\(oacc kernels parallelized, oacc function \\(, , \\), oacc kernels, omp target entrypoint\\)\\)" 3 "parloops1" } } */ 63 /* { dg-final { scan-tree-dump-not "FAILED:" "parloops1" } } */ 64 65 /* Check that the loop has been split off into a function. */ 66 /* { dg-final { scan-tree-dump-times "(?n);; Function .*main._omp_fn.0" 1 "optimized" } } */ 67 /* { dg-final { scan-tree-dump-times "(?n);; Function .*main._omp_fn.1" 1 "optimized" } } */ 68 /* { dg-final { scan-tree-dump-times "(?n);; Function .*main._omp_fn.2" 1 "optimized" } } */ 69