projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 41b4790) | patch
ARM64: add support for AArch64's addsub_ext.ll
authorTim Northover <tnorthover@apple.com>
Mon, 14 Apr 2014 12:50:50 +0000 (12:50 +0000)
committerTim Northover <tnorthover@apple.com>
Mon, 14 Apr 2014 12:50:50 +0000 (12:50 +0000)
commit3c68c5c55e11315e2f7a9d238f88febed712f5bc
tree74755ffc8eabcd0192461e18c0822258617f8d2btree | snapshot
parent41b47904ba9d7317376a6d0c5b8fcf887ac9ef73commit | diff
ARM64: add support for AArch64's addsub_ext.ll

There was one definite issue in ARM64 (the off-by-1 check for whether
a shift could be folded in) and one difference that is probably
correct: ARM64 didn't fold nodes with multiple uses into the
arithmetic operations unless optimising for code size.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206168 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/ARM64/ARM64ISelDAGToDAG.cpp diff | blob | history
test/CodeGen/AArch64/addsub_ext.ll diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors