[clang] [llvm] [AMDGPU] Add intrinsics and builtins for v_wave_match_b32 (PR #226971)

[Matt Arsenault via llvm-commits]

================
@@ -5946,6 +5946,78 @@ define void @ds_bpermute_uniform_lane(ptr addrspace(1) %out, i32 %lanearg, i32 %
   ret void
 }
 
+; --------------------------------------------------------------------
----------------
arsenm wrote:

I'd probably move this to its own file, especially to avoid all the attribute renumbering noise 

https://github.com/llvm/llvm-project/pull/226971