[llvm] [AMDGPU][True16][CodeGen] optimize codengen for mad-mix in true16 (PR #124995)

[Matt Arsenault via llvm-commits]

================
@@ -3651,6 +3651,10 @@ bool AMDGPUDAGToDAGISel::SelectVOP3PMadMixModsImpl(SDValue In, SDValue &Src,
       // TODO: Should we try to look for neg/abs here?
     }
 
+    // Prevent unnecessary subreg COPY to VGPR_16
+    if (Subtarget->useRealTrue16Insts() && Src.getOpcode() == ISD::TRUNCATE) {
+      Src = Src.getOperand(0);
+    }
----------------
arsenm wrote:

Can we just do this unconditionally? The truncate is synthetic in the fake 16 case anyway 

https://github.com/llvm/llvm-project/pull/124995