| 1 |
/* { dg-do run { target openacc_nvidia_accel_selected } } */ |
| 2 |
/* { dg-additional-options "-lcuda" } */ |
| 3 |
/* { dg-require-effective-target openacc_cuda } */ |
| 4 |
|
| 5 |
#include <openacc.h> |
| 6 |
#include <stdlib.h> |
| 7 |
#include "cuda.h" |
| 8 |
|
| 9 |
#include <stdio.h> |
| 10 |
|
| 11 |
#define n 128 |
| 12 |
|
| 13 |
int |
| 14 |
main (void) |
| 15 |
{ |
| 16 |
CUresult r; |
| 17 |
CUstream stream1; |
| 18 |
int N = n; |
| 19 |
int a[n]; |
| 20 |
int c[n]; |
| 21 |
|
| 22 |
acc_init (acc_device_nvidia); |
| 23 |
|
| 24 |
r = cuStreamCreate (&stream1, CU_STREAM_NON_BLOCKING); |
| 25 |
if (r != CUDA_SUCCESS) |
| 26 |
{ |
| 27 |
fprintf (stderr, "cuStreamCreate failed: %d\n", r); |
| 28 |
abort (); |
| 29 |
} |
| 30 |
|
| 31 |
acc_set_cuda_stream (1, stream1); |
| 32 |
|
| 33 |
for (int i = 0; i < n; i++) |
| 34 |
{ |
| 35 |
a[i] = 3; |
| 36 |
c[i] = 0; |
| 37 |
} |
| 38 |
|
| 39 |
#pragma acc data copy (a, c) copyin (N) |
| 40 |
{ |
| 41 |
#pragma acc parallel async (1) |
| 42 |
; |
| 43 |
|
| 44 |
#pragma acc parallel async (1) num_gangs (320) |
| 45 |
#pragma acc loop gang |
| 46 |
for (int ii = 0; ii < N; ii++) |
| 47 |
c[ii] = (a[ii] + a[N - ii - 1]); |
| 48 |
|
| 49 |
#pragma acc parallel async (1) |
| 50 |
#pragma acc loop seq |
| 51 |
for (int ii = 0; ii < n; ii++) |
| 52 |
a[ii] = 6; |
| 53 |
|
| 54 |
#pragma acc wait (1) |
| 55 |
} |
| 56 |
|
| 57 |
for (int i = 0; i < n; i++) |
| 58 |
if (c[i] != 6) |
| 59 |
abort (); |
| 60 |
|
| 61 |
return 0; |
| 62 |
} |