[PATCH] D76514: [ARM] Avoid pointless vrev of element-wise vmov

[Dave Green via Phabricator via llvm-commits]

dmgreen added a comment.

I see. Because we are just swapping around the same values anyway. Makes sense.

Would it make sense to do this as a DAG fold? That way we would get MVE for free too.



================
Comment at: llvm/test/CodeGen/ARM/vmov.ll:1
-; RUN: llc -mtriple=arm-eabi -mattr=+neon %s -o - | FileCheck %s
+; RUN: llc -mtriple=arm-eabi -mattr=+neon,+fullfp16 %s -o - | FileCheck %s
+; RUN: llc -mtriple=armeb-eabi -mattr=+neon,+fullfp16 %s -o - | FileCheck %s
----------------
Can you please run the update script on this with old codegen, pre-commit the result and then show the diffs against new codegen here.

It might help to make them hard-float too.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D76514/new/

https://reviews.llvm.org/D76514