[PATCH] D96997: [LV] Compute max scalable & fixed VFs up front, then apply them. (WIP)

[Florian Hahn via Phabricator via llvm-commits]

fhahn created this revision.
Herald added a subscriber: hiraditya.
fhahn requested review of this revision.
Herald added a project: LLVM.
Herald added a subscriber: llvm-commits.

This patch updates computeFeasibleMaxVF to operate in 3 stages:

1. Compute the maximum fixed and scalable vectorization factors by optimistically assuming the largest possible VFs and then gradually constrainting them based on legality restrictions.
2. Check if UserVF can be used, given the matching MaxVF
3. Pick feasible VF, clamp it to matching MaxVF (Currently FixedMaxVF).


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D96997

Files:
  llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
  llvm/test/Transforms/LoopVectorize/AArch64/scalable-reductions.ll
  llvm/test/Transforms/LoopVectorize/AArch64/scalable-vf-hint.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D96997.324775.patch
Type: text/x-patch
Size: 12834 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210218/b4a84bc8/attachment.bin>