[llvm] [AMDGPU] Fix implicit operands of VOPD cndmask instructions (PR #87788)

[Matt Arsenault via llvm-commits]

================
@@ -642,6 +642,17 @@ void GCNSubtarget::overrideSchedPolicy(MachineSchedPolicy &Policy,
     Policy.ShouldTrackLaneMasks = true;
 }
 
+void GCNSubtarget::mirFileLoaded(MachineFunction &MF) const {
----------------
arsenm wrote:

I think we should switch to using HwMode on the wavesize to swap out the actual operands on construction 

https://github.com/llvm/llvm-project/pull/87788