[PATCH] D63815: [DAGCombine] visitEXTRACT_SUBVECTOR - 'little to big' extract_subvector(bitcast()) support

[Eli Friedman via Phabricator via llvm-commits]

efriedma added a comment.

It looks like the arm64-neon-2velem.ll regressions are a shuffle lowering issue, yes; we're creating a DUPLANE where the operand is an extract_subvector, and it doesn't simplify.

The failure to match fcvtl2 probably is just a matter of fixing the pattern in the .td file: it explicitly checks for an operation where the operand is an extract_subvector.

Not sure what the right resolution is for allowsMisalignedMemoryAccesses; we currently only use FeatureSlowMisaligned128Store for certain Exynos targets.  Probably not too important.


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D63815/new/

https://reviews.llvm.org/D63815