[llvm] [X86] Fold C1 * C2 + Z --> C3 + Z for vpmadd52l/vpmadd52h (PR #156293)

[Simon Pilgrim via llvm-commits]

================
@@ -44954,26 +44954,39 @@ bool X86TargetLowering::SimplifyDemandedBitsForTargetNode(
   }
   case X86ISD::VPMADD52L:
   case X86ISD::VPMADD52H: {
-    KnownBits KnownOp0, KnownOp1;
+    KnownBits Known52BitsOfOp0, Known52BitsOfOp1;
----------------
RKSimon wrote:

This rename seems unnecessary tbh and increases chances of 80col overflow.

https://github.com/llvm/llvm-project/pull/156293