[PATCH] D110524: [AArch64ISelLowering] Avoid duplane in some cases when sve enabled

[guopeilin via Phabricator via llvm-commits]

guopeilin added a comment.





================
Comment at: llvm/lib/Target/AArch64/AArch64ISelLowering.cpp:9470
+    auto ExtractedValTye = V.getOperand(0).getValueType();
+    if (ExtractedValTye.getSizeInBits().getValue() <= 128) {
+      Lane += V.getConstantOperandVal(1);
----------------
david-arm wrote:
> This isn't safe for truly scalable vectors where the size is unknown at runtime. I think this should probably be written as:
> 
>   auto ExtractedValType = V.getOperand(0).getValueType();
>   if (ExtractedValType.isFixedLengthVector() && ExtractedValType.getFixedSizeInBits() <= 128) {
>    ...
>   }
> 
> 
>   
Fixed, please review


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D110524/new/

https://reviews.llvm.org/D110524