[clang] [OpenMP][SIMD][FIX] Use conservative "omp simd ordered" lowering (PR #123867)

[via cfe-commits]

================
@@ -0,0 +1,110 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --functions "omp_simd_loop" --check-globals smart --filter "access|\%omp.inner.for.cond.*\!llvm.loop"
----------------
MattPD wrote:

Thanks; missed that, too!

I'm wondering what would be a good home for an OpenMP runtime test?
I think only https://github.com/llvm/llvm-project/blob/ec7167bac7c89b43449dbce81e65fb09e5d27cb5/openmp/runtime/test/parallel/bug54082.c#L34-L36 happens to exercise `#pragma omp simd` but that's still in the context of `#pragma omp parallel` and thus still in "openmp/runtime/test/parallel".
I suppose I could create "openmp/runtime/test/simd".

However, first I'm going to see if I can preserve the salient behavior of this test--i.e., the (non)emission of the llvm.access.group metadata--if I remove the headers and only include the function prototypes. Perhaps that's going to be simpler.


https://github.com/llvm/llvm-project/pull/123867