| 1 |
/* { dg-do run { target openacc_nvidia_accel_selected } } */ |
| 2 |
/* { dg-set-target-env-var "GOMP_OPENACC_DIM" ":2:" } */ |
| 3 |
/* { dg-additional-options "-foffload=-fdump-tree-oaccloops" } */ |
| 4 |
/* { dg-set-target-env-var "GOMP_DEBUG" "1" } */ |
| 5 |
|
| 6 |
#include <stdlib.h> |
| 7 |
|
| 8 |
#define N 1024 |
| 9 |
|
| 10 |
unsigned int a[N]; |
| 11 |
unsigned int b[N]; |
| 12 |
unsigned int c[N]; |
| 13 |
unsigned int n = N; |
| 14 |
|
| 15 |
int |
| 16 |
main (void) |
| 17 |
{ |
| 18 |
for (unsigned int i = 0; i < n; ++i) |
| 19 |
{ |
| 20 |
a[i] = i % 3; |
| 21 |
b[i] = i % 5; |
| 22 |
} |
| 23 |
|
| 24 |
#pragma acc parallel vector_length (128) copyin (a,b) copyout (c) |
| 25 |
{ |
| 26 |
#pragma acc loop worker |
| 27 |
for (unsigned int i = 0; i < 4; i++) |
| 28 |
#pragma acc loop vector |
| 29 |
for (unsigned int j = 0; j < n / 4; j++) |
| 30 |
c[(i * N / 4) + j] = a[(i * N / 4) + j] + b[(i * N / 4) + j]; |
| 31 |
} |
| 32 |
|
| 33 |
for (unsigned int i = 0; i < n; ++i) |
| 34 |
if (c[i] != (i % 3) + (i % 5)) |
| 35 |
abort (); |
| 36 |
|
| 37 |
return 0; |
| 38 |
} |
| 39 |
|
| 40 |
/* { dg-final { scan-offload-tree-dump "__attribute__\\(\\(oacc function \\(1, 0, 128\\)" "oaccloops" } } */ |
| 41 |
/* { dg-output "nvptx_exec: kernel main\\\$_omp_fn\\\$0: launch gangs=1, workers=2, vectors=128" } */ |