[PATCH] D44627: [CallSiteSplitting] Only record conditions up to the IDom(call site).

[Florian Hahn via Phabricator via llvm-commits]

fhahn updated this revision to Diff 173155.
fhahn added a comment.
Herald added subscribers: dexonsmith, steven_wu.

Resurrecting the patch. The cases we loose (propagating %a == null in test_cond_no_effect) is now handled by ipsccp, which is a better place IMO. @junbuml what do you think?

This patch makes CallSiteSplitting stop backtracking once it is not necessary any more and avoid splitting unnecessarily.


https://reviews.llvm.org/D44627

Files:
  lib/Transforms/Scalar/CallSiteSplitting.cpp
  test/Other/new-pm-lto-defaults.ll
  test/Other/opt-O3-pipeline.ll
  test/Transforms/CallSiteSplitting/callsite-split-or-phi.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D44627.173155.patch
Type: text/x-patch
Size: 11793 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20181108/9c955500/attachment.bin>