[llvm] [AMDGPU] Eliminate unnecessary packing in wider f16 vectors for sdwa/opsel-able instruction (PR #137137)

[Krzysztof Drewniak via llvm-commits]

================
@@ -207,6 +227,52 @@ class SDWADstPreserveOperand : public SDWADstOperand {
 #endif
 };
 
+class SDWAFP16ChainOperand : public SDWAOperand {
+private:
+  SIPeepholeSDWA &Parent;
+  FP16PackCandidate Candidate;
----------------
krzysz00 wrote:

Initial comment: Why does this say "FP16" when this issue isn't limited to f16 instructions?

https://github.com/llvm/llvm-project/pull/137137