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

[Pierre van Houtryve via llvm-commits]

================
@@ -46,6 +54,8 @@ class SDWADstOperand;
 using SDWAOperandsVector = SmallVector<SDWAOperand *, 4>;
 using SDWAOperandsMap = MapVector<MachineInstr *, SDWAOperandsVector>;
 
+struct FP16PackCandidate;
----------------
Pierre-vh wrote:

Forward decl is not needed as you're not using this type name before its definition

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