projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 12f35c5) | patch
Add a DAGCombine xform to fold away redundant float->double->float conversions around...
authorOwen Anderson <resistor@mac.com>
Sun, 18 Jul 2010 08:47:54 +0000 (08:47 +0000)
committerOwen Anderson <resistor@mac.com>
Sun, 18 Jul 2010 08:47:54 +0000 (08:47 +0000)
commit3ecdfafe3b06844ab0412b48f6daedf3e13e7037
tree174bd64b9e3a9186d57fdf2de7a6c2dc5babe300tree | snapshot
parent12f35c52a533da0c2c4c3e0a04f83355514992f9commit | diff
Add a DAGCombine xform to fold away redundant float->double->float conversions around sqrt instructions.
I am assured by people more knowledgeable than me that there are no rounding issues in eliminating this.

This fixed <rdar://problem/8197504>.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@108639 91177308-0d34-0410-b5e6-96231b3b80d8
lib/CodeGen/SelectionDAG/DAGCombiner.cpp diff | blob | history
C/C++ LLVM-based compilers that forbids OOTA behaviors