[PATCH] D116879: [llvm] Allow forced auto-vectorization of sincos() using libmvec

[Tim Schmielau via Phabricator via llvm-commits]

tim.schmielau added a comment.

The failing LLVM.Transforms/LoopVectorize/AArch64::scalable-call.ll <https://reviews.llvm.org/harbormaster/unit/view/5007351/> test makes me wonder what the exact semantics of the vector-function-abi-variant <https://llvm.org/docs/LangRef.html#call-site-attributes> call attribute is in the presence of pointer arguments.

Apparently it does not imply vectorization is safe, because then we wouldn't need to invoke LoopAccessAnalysis at all. But how is LoopAccessAnalysis supposed to work if we don't know the length of the array pointed at? Do we assume pointers always point to a single element?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D116879/new/

https://reviews.llvm.org/D116879