projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: a637a8b) | patch
Teach GlobalLoadUsesSimpleEnoughForHeapSRA and the SROA rewriter how to handle
authorChris Lattner <sabre@nondot.org>
Thu, 13 Sep 2007 21:31:36 +0000 (21:31 +0000)
committerChris Lattner <sabre@nondot.org>
Thu, 13 Sep 2007 21:31:36 +0000 (21:31 +0000)
commit309f20fc45a599ca45ad1262525d9f4c332b1911
tree8a1ed0d7f47ef43803aa914e1d37e912d5ccb138tree | snapshot
parenta637a8b1e7ecc9ce06e0463b64be0721bf52255dcommit | diff
Teach GlobalLoadUsesSimpleEnoughForHeapSRA and the SROA rewriter how to handle
a limited form of PHI nodes.  This finally fixes PR1639, speeding 179.art up
from 7.84s to 3.13s on PPC.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@41933 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Transforms/IPO/GlobalOpt.cpp diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors