https://github.com/TIFitis commented:

The teams changes look reasonable overall, but I found an optimized AMDGPU 
compilation failure that also affects the new teams path.

The reduced example below omits `teams` because the same failure reproduces on 
the base PRs. With `-fclangir -O2` targeting `gfx1100`, device compilation 
aborts at `StackToShared.cpp:100`:

```text
unsupported non-default address space alloca with multiple uses
```

```c
void use(int);
void test(void) {
#pragma omp target
  {
    int x = 1;
#pragma omp parallel
    use(x);
  }
}
```

https://github.com/llvm/llvm-project/pull/231982
_______________________________________________
llvm-branch-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits

Reply via email to