[llvm] [X86] Fold chained i32 CLMUL operands (PR #202250)

[Simon Pilgrim via llvm-commits]

================
@@ -788,6 +788,34 @@ SDValue TargetLowering::SimplifyMultipleUseDemandedBits(
 
     break;
   }
+  // On little-endian targets, a scalar-to-vector round trip from lane zero
+  // preserves the demanded low bits when the source and destination widths
+  // match.
+  case ISD::SCALAR_TO_VECTOR: {
----------------
RKSimon wrote:

This needs to be its own patch

https://github.com/llvm/llvm-project/pull/202250