[clang] [llvm] [AMDGPU] Support loop unrolled asyncmark (PR #212929)

[Matt Arsenault via cfe-commits]

================
@@ -0,0 +1,18 @@
+; RUN: not llc -global-isel=0 -mtriple=amdgpu12.50 -filetype=null %s 2>&1 | FileCheck %s
+; RUN: not llc -global-isel=1 -mtriple=amdgpu12.50 -filetype=null %s 2>&1 | FileCheck %s -check-prefix=CHECK-GISEL
+
+; The wait_asyncmark operand indexes compiler-tracked async mark state, so it
+; must fold to a constant by instruction selection.
----------------
arsenm wrote:

Yes. This would need to be some kind of constexpr frontend time unrolling 

https://github.com/llvm/llvm-project/pull/212929