projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 00666a1) | patch
[ARM] Use AEABI aligned function variants
authorJohn Brawn <john.brawn@arm.com>
Tue, 12 May 2015 13:13:38 +0000 (13:13 +0000)
committerJohn Brawn <john.brawn@arm.com>
Tue, 12 May 2015 13:13:38 +0000 (13:13 +0000)
commite38f45effc7304aebe4b25af03932202c6bae82e
tree49f95b4aa9a96184b919c3019c52dea45b9c82b3tree | snapshot
parent00666a17ff4623d08be1c03cde25898d3c1762ffcommit | diff
[ARM] Use AEABI aligned function variants

AEABI defines aligned variants of memcpy etc. that can be faster than
the default version due to not having to do alignment checks. When
emitting target code for these functions make use of these aligned
variants if possible. Also convert memset to memclr if possible.

Differential Revision: http://reviews.llvm.org/D8060

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@237127 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/ARM/ARMSelectionDAGInfo.cpp diff | blob | history
lib/Target/ARM/ARMSelectionDAGInfo.h diff | blob | history
test/CodeGen/ARM/memfunc.ll diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors