[llvm] [LLVM][CodeGen][SVE] Improve custom lowering for EXTRACT_SUBVECTOR. (PR #90963)

[Paul Walker via llvm-commits]

================
@@ -13897,23 +13897,27 @@ AArch64TargetLowering::LowerEXTRACT_VECTOR_ELT(SDValue Op,
 
 SDValue AArch64TargetLowering::LowerEXTRACT_SUBVECTOR(SDValue Op,
                                                       SelectionDAG &DAG) const {
-  assert(Op.getValueType().isFixedLengthVector() &&
+  EVT VT = Op.getValueType();
+  assert(VT.isFixedLengthVector() &&
          "Only cases that extract a fixed length vector are supported!");
 
   EVT InVT = Op.getOperand(0).getValueType();
   unsigned Idx = Op.getConstantOperandVal(1);
   unsigned Size = Op.getValueSizeInBits();
 
   // If we don't have legal types yet, do nothing
-  if (!DAG.getTargetLoweringInfo().isTypeLegal(InVT))
+  if (!isTypeLegal(InVT))
     return SDValue();
 
   if (InVT.isScalableVector()) {
     // This will be matched by custom code during ISelDAGToDAG.
-    if (Idx == 0 && isPackedVectorType(InVT, DAG))
----------------
paulwalker-arm wrote:

I'm now less sure about this change because I think we're missing some upstream test coverage.   Please hold of reviewing this PR for now and I'll report back.

https://github.com/llvm/llvm-project/pull/90963