projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 1434f66) | patch
Revert r122955. It seems using movups to lower memcpy can cause massive regression...
authorEvan Cheng <evan.cheng@apple.com>
Fri, 7 Jan 2011 19:35:30 +0000 (19:35 +0000)
committerEvan Cheng <evan.cheng@apple.com>
Fri, 7 Jan 2011 19:35:30 +0000 (19:35 +0000)
commita5e1362f968568d66d76ddcdcff4ab98e203a48c
tree53e266c315432b49be8ad6f3a2d2a5873265ab53tree | snapshot
parent1434f66b2e132a707e2c8ccb3350ea13fb5aa051commit | diff
Revert r122955. It seems using movups to lower memcpy can cause massive regression (even on Nehalem) in edge cases. I also didn't see any real performance benefit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@123015 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/X86/X86ISelLowering.cpp diff | blob | history
test/CodeGen/X86/2010-04-08-CoalescerBug.ll diff | blob | history
test/CodeGen/X86/2010-06-25-CoalescerSubRegDefDead.ll diff | blob | history
test/CodeGen/X86/2010-09-17-SideEffectsInChain.ll diff | blob | history
test/CodeGen/X86/memcpy-2.ll diff | blob | history
test/CodeGen/X86/memcpy.ll diff | blob | history
test/CodeGen/X86/memset-2.ll diff | blob | history
test/CodeGen/X86/memset64-on-x86-32.ll diff | blob | history
test/CodeGen/X86/small-byval-memcpy.ll diff | blob | history
test/CodeGen/X86/tlv-1.ll diff | blob | history
test/CodeGen/X86/unaligned-load.ll diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors