[Mlir-commits] [llvm] [mlir] [llvm][OpenMP] Handle complex types in atomic read (PR #111377)

[Tom Eccles]

tblah wrote:

Thanks for your work on this.

One of our tests segfaults at runtime. Here is a minimal reproducer (that should loop forever)
```
program complex_read
  complex :: x = 0
  complex v
  !$omp parallel
  do
    !$omp atomic read
      v = x
    if (v /= x) exit
  end do
  !$omp end parallel
end program
```

I am testing on aarch64 (aws graviton 3) and using the atomic implementation from compiler-rt.

https://github.com/llvm/llvm-project/pull/111377