[llvm] [X86][FP16][BF16] Improve vectorization of fcmp (PR #116153)

[Simon Pilgrim via llvm-commits]

================
@@ -23483,14 +23485,40 @@ static SDValue LowerVSETCC(SDValue Op, const X86Subtarget &Subtarget,
   SDLoc dl(Op);
 
   if (isFP) {
-    MVT EltVT = Op0.getSimpleValueType().getVectorElementType();
+    MVT SVT = Op0.getSimpleValueType();
----------------
RKSimon wrote:

We usually use SVT for scalar VTs - maybe OpVT? The getSimpleValueType can also be merged with the isFP check above

https://github.com/llvm/llvm-project/pull/116153