[llvm] [AMDGPU] Add waterfall intrinsics (PR #192409)

[Jay Foad via llvm-commits]

================
@@ -143,6 +143,9 @@ static std::optional<Instruction *> modifyIntrinsicCall(
   NewCall->copyMetadata(OldIntr);
   if (isa<FPMathOperator>(NewCall))
     NewCall->copyFastMathFlags(&OldIntr);
+  // Copy attributes
----------------
jayfoad wrote:

I'd prefer to put this in a separate PR but I won't insist. You should be able to merge this first, if you can find a test case that exercises this code and shows that the call site attrs are not preserved.

https://github.com/llvm/llvm-project/pull/192409