[clang] [llvm] [AMDGPU] Introduce asyncmark/wait intrinsics (PR #173259)

[Sameer Sahasrabuddhe via cfe-commits]

================
@@ -538,6 +538,14 @@ TARGET_BUILTIN(__builtin_amdgcn_ashr_pk_u8_i32, "UsUiUiUi", "nc", "ashr-pk-insts
 TARGET_BUILTIN(__builtin_amdgcn_cvt_scalef32_2xpk16_fp6_f32, "V6UiV16fV16ff", "nc", "gfx950-insts")
 TARGET_BUILTIN(__builtin_amdgcn_cvt_scalef32_2xpk16_bf6_f32, "V6UiV16fV16ff", "nc", "gfx950-insts")
 
+//===----------------------------------------------------------------------===//
+// Async mark builtins.
+//===----------------------------------------------------------------------===//
+
+// FIXME: Not supported on GFX12 yet. Will need a new feature when we do.
+TARGET_BUILTIN(__builtin_amdgcn_asyncmark, "v", "n", "vmem-to-lds-load-insts")
----------------
ssahasra wrote:

Done. Discovered this when I rebased.

https://github.com/llvm/llvm-project/pull/173259