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

[Pierre van Houtryve via llvm-commits]

================
@@ -89,6 +100,16 @@ class SIPeepholeSDWALegacy : public MachineFunctionPass {
 
 using namespace AMDGPU::SDWA;
 
+struct FP16PackCandidate {
----------------
Pierre-vh wrote:

Please add comments to tell what each field does

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