[llvm] [X86] Generate `vpmuludq` instead of `vpmullq` (PR #121456)

[Simon Pilgrim via llvm-commits]

================
@@ -60832,3 +60832,24 @@ Align X86TargetLowering::getPrefLoopAlignment(MachineLoop *ML) const {
     return Align(1ULL << ExperimentalPrefInnermostLoopAlignment);
   return TargetLowering::getPrefLoopAlignment();
 }
+
+bool X86TargetLowering::shouldSimplifyDemandedVectorElts(
+    SDValue Op, const TargetLoweringOpt &TLO) const {
+  if (Op.getOpcode() == ISD::VECTOR_SHUFFLE) {
----------------
RKSimon wrote:

This isn't going to work in the general case, it will just help shuffles:
https://llvm.godbolt.org/z/no1Gc6fzT

https://github.com/llvm/llvm-project/pull/121456