| 1 |
/* { dg-do run { target openacc_nvidia_accel_selected } } |
| 2 |
{ dg-skip-if "" { *-*-* } { "*" } { "-O2" } } */ |
| 3 |
/* { dg-additional-options "-foffload=-fdump-rtl-mach" } */ |
| 4 |
|
| 5 |
#define n 1024 |
| 6 |
|
| 7 |
int |
| 8 |
main (void) |
| 9 |
{ |
| 10 |
#pragma acc parallel |
| 11 |
{ |
| 12 |
#pragma acc loop worker |
| 13 |
for (int i = 0; i < n; i++) |
| 14 |
; |
| 15 |
|
| 16 |
#pragma acc loop worker |
| 17 |
for (int i = 0; i < n; i++) |
| 18 |
; |
| 19 |
} |
| 20 |
|
| 21 |
return 0; |
| 22 |
} |
| 23 |
|
| 24 |
/* Atm, %ntid.y is broadcast from one loop to the next, so there are 2 bar.syncs |
| 25 |
for that (the other two are there for the same reason as in pr85381-2.c). |
| 26 |
Todo: Recompute %ntid.y instead of broadcasting it. */ |
| 27 |
/* { dg-final { scan-offload-rtl-dump-times "nvptx_barsync" 4 "mach" } } */ |