| Issue |
184345
|
| Summary |
parsing problem #pragma omp target unroll partial ignores unroll partial on -O2, while nesting with brackets #pragma omp target {#pragma omp unroll ...} works in clang 22
|
| Labels |
clang
|
| Assignees |
|
| Reporter |
bschulz81
|
compile with
/usr/lib/llvm/22/bin/clang++ -fopenmp -fopenmp-targets=nvptx64-nvidia-cuda -O2 main.cpp
```
int
main ()
{
int a[200];
#pragma omp target enter data map (to: a[0:200])
#pragma omp target unroll partial
for (int i = 0; i < 200; i++)
{
a[i] = 2*i + 1;
}
return 0;
}
```
creates a warning
> CMakeFiles/gpu_compiler_test.dir/main.cpp.o.d -o CMakeFiles/gpu_compiler_test.dir/main.cpp.o -c /home/benni/projects/openmptestnew/openmpoffloatest/main.cpp
> /home/benni/projects/openmptestnew/openmpoffloatest/main.cpp:8:24: warning: extra tokens at the end of '#pragma omp target' are ignored [-Wextra-tokens]
> 8 | #pragma omp target unroll partial
> | ^
> 1 warning generated.
> /home/benni/projects/openmptestnew/openmpoffloatest/main.cpp:8:24: warning: extra tokens at the end of '#pragma omp target' are ignored [-Wextra-tokens]
That it ignores unroll partial,
The code below, which merely puts a bracket between the openmp pragmas, compiles without the warning (it appears the partial unroll is then compiled)
```
int
main ()
{
int a[200];
#pragma omp target enter data map (to: a[0:200])
#pragma omp target
{
#pragma omp unroll partial
for (int i = 0; i < 200; i++)
{
a[i] = 2*i + 1;
}
}
return 0;
}
```
![]()
_______________________________________________
llvm-bugs mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
