| Issue |
94047
|
| Summary |
OpenMP offload Program with atomic does not finished when build with -O3 of AMD GPUs
|
| Labels |
bug,
openmp,
clang:openmp,
offload
|
| Assignees |
|
| Reporter |
fel-cab
|
On a MI250x
The following program doesn't finish when compiled as:
`clang -fopenmp -O3 --offload-arch=gfx90a `
But it does finish when compiled as:
`clang -fopenmp --offload-arch=gfx90a
`
```#include <omp.h>
#include <stdio.h>
#include <stdlib.h>
#pragma omp requires atomic_default_mem_order(acq_rel)
int main() {
int x = 0, y = 0;
int errors = 0;
#pragma omp target parallel num_threads(2) map(tofrom: x, y, errors)
{
int thrd = omp_get_thread_num();
if (thrd == 0) {
x = 10;
#pragma omp atomic write
y = 1;
}
if (thrd == 1){
int tmp = 0;
while (tmp == 0) {
#pragma omp atomic read
tmp = y;
}
if(x != 10) errors++;
}
}
if(errors)
printf("Test Failed\n");
else
printf("Test Passed\n");
return errors;
}
```
![]()
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
