[llvm] [AMDGPU] Waterfall loop codegen improvement in SIInstrInfo (PR #192415)

[Jay Foad via llvm-commits]

================
@@ -27,10 +27,13 @@ class LaneMaskConstants {
   const unsigned AndN2Opc;
   const unsigned AndN2SaveExecOpc;
   const unsigned AndN2TermOpc;
+  const unsigned AndN2WRExecOpc;
----------------
jayfoad wrote:

Also, _maybe_ add a comment here saying that it's only valid on GFX10+? Although we haven't done that for any of the other opcodes.

https://github.com/llvm/llvm-project/pull/192415