[X86] combineX86ShuffleChain - prefer combining to X86ISD::SHUF128 if PERMQ operands are splittable (#133900)

If the 512-bit unary shuffle is a concatenation of 128/256-bit subvectors then we're better off using a X86ISD::SHUF128 node so we can fold the concatenation into the shuffle as well.

Author: RKSimon

llvm/lib/Target/X86/X86ISelLowering.cpp

@@ -39797,7 +39797,8 @@ static SDValue combineX86ShuffleChain(
     // FIXME: Is there a better way to do this? is256BitLaneRepeatedShuffleMask
     // doesn't work because our mask is for 128 bits and we don't have an MVT
     // to match that.
-    bool PreferPERMQ = UnaryShuffle && isUndefOrInRange(ScaledMask[0], 0, 2) &&
+    bool PreferPERMQ = UnaryShuffle && !isFreeToSplitVector(V1, DAG) &&
+                       isUndefOrInRange(ScaledMask[0], 0, 2) &&
                        isUndefOrInRange(ScaledMask[1], 0, 2) &&
                        isUndefOrInRange(ScaledMask[2], 2, 4) &&
                        isUndefOrInRange(ScaledMask[3], 2, 4) &&