[PATCH] D32422: LoopVectorizer: let target prefer scalar addressing computations (+ minor improvements in SystemZTTI)

[Jonas Paulsson via Phabricator via llvm-commits]

jonpa updated this revision to Diff 99286.
jonpa added a comment.

ping!

This patch has now been confirmed on benchmarks, and is ready to commit. See previous comment for a summary.

I added a comment like:
// FIXME: Currently widenPHIInstruction() often creates a dead vector
// induction variable when the PHI user is scalarized.

I am not sure if it is motivated or in the right place... It is nothing new with this patch.

(The SystemZ part which was unrelated has been removed from this review)


https://reviews.llvm.org/D32422

Files:
  include/llvm/Analysis/TargetTransformInfo.h
  include/llvm/Analysis/TargetTransformInfoImpl.h
  lib/Analysis/TargetTransformInfo.cpp
  lib/Target/SystemZ/SystemZTargetTransformInfo.h
  lib/Transforms/Vectorize/LoopVectorize.cpp
  test/Transforms/LoopVectorize/SystemZ/addressing.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D32422.99286.patch
Type: text/x-patch
Size: 10207 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170517/b37d0f41/attachment.bin>