[PATCH] D69510: [ARM] MVE reverse shuffles.

[Sam Parker via Phabricator via llvm-commits]

samparker added inline comments.


================
Comment at: llvm/lib/Target/ARM/ARMISelLowering.cpp:7659
+  EVT VT = Op.getValueType();
 
+  SDValue OpLHS = DAG.getNode(ARMISD::VREV64, DL, VT, Op.getOperand(0));
----------------
I think keeping some kind of assert is a good idea here.


================
Comment at: llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp:642
+          {ISD::VECTOR_SHUFFLE, MVT::v4f32, 5},
+          {ISD::VECTOR_SHUFFLE, MVT::v8i16, 5},
+          {ISD::VECTOR_SHUFFLE, MVT::v8f16, 5},
----------------
should we be considering v4i16 and v8i8 too?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69510/new/

https://reviews.llvm.org/D69510