[llvm] [CodeGen] [AMDGPU] Attempt DAGCombine for fmul with select to ldexp (PR #111109)

[Matt Arsenault via llvm-commits]

================
@@ -14526,9 +14522,13 @@ SDValue SITargetLowering::performFMulCombine(SDNode *N,
         else
           return SDValue();
 
-        SDValue ExtNode = DAG.getNode(ExtOp, SL, MVT::i32, RHS.getOperand(0));
+        EVT ExtVT = VT.isVector()
+                        ? EVT::getVectorVT(*DAG.getContext(), MVT::i32,
+                                           VT.getVectorNumElements())
+                        : MVT::i32;
----------------
arsenm wrote:

changeElementType 

https://github.com/llvm/llvm-project/pull/111109