projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 1abb7bc) | patch
[NVPTX] Use approximate FP ops when unsafe-fp-math is used, and append
authorJustin Holewinski <jholewinski@nvidia.com>
Mon, 22 Jul 2013 12:18:04 +0000 (12:18 +0000)
committerJustin Holewinski <jholewinski@nvidia.com>
Mon, 22 Jul 2013 12:18:04 +0000 (12:18 +0000)
commit3a8ee4ffd783bd0cf2d83089edb43ec546b49d0d
treebb08127142c401b1737718f4ac00320393e5e0fbtree | snapshot
parent1abb7bc7e917771a80cd7788ee37ba7dab98f183commit | diff
[NVPTX] Use approximate FP ops when unsafe-fp-math is used, and append
.ftz to instructions if the nvptx-f32ftz attribute is set to "true"

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@186820 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Target/NVPTX/NVPTXISelDAGToDAG.cpp diff | blob | history
lib/Target/NVPTX/NVPTXISelDAGToDAG.h diff | blob | history
lib/Target/NVPTX/NVPTXInstrInfo.td diff | blob | history
test/CodeGen/NVPTX/fast-math.ll [new file with mode: 0644] blob
C/C++ LLVM-based compilers that forbids OOTA behaviors