[llvm] [AMDGPU] Change isSingleLaneExecution to account for WWM enabling lanes even if there's only one workitem (PR #188316)

[Jay Foad via llvm-commits]

================
@@ -108,6 +108,9 @@ intrinsicToAttrMask(Intrinsic::ID ID, bool &NonKernelOnly, bool &NeedsImplicit,
     // offsets.
     return CodeObjectVersion >= AMDGPU::AMDHSA_COV5 ? IMPLICIT_ARG_PTR
                                                     : QUEUE_PTR;
+  case Intrinsic::amdgcn_wwm:
----------------
jayfoad wrote:

That sounds fine as-is.

I was getting muddled up with the difference between the "wqm" and "softwqm" intrinsics.

https://github.com/llvm/llvm-project/pull/188316