]> granicus.if.org Git - clang/commit
projects / clang / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 9a0c251) | patch
[CUDA] Enable fusing FP ops (-ffp-contract=fast) for CUDA by default.
authorArtem Belevich <tra@google.com>
Thu, 19 May 2016 18:44:45 +0000 (18:44 +0000)
committerArtem Belevich <tra@google.com>
Thu, 19 May 2016 18:44:45 +0000 (18:44 +0000)
commitf159984fe052abd07234c80f7cc15902ba8e297d
treeca379ef47943339c8f7feacdc0c92bfd854ece23tree | snapshot
parent9a0c2517ad81a849ec57558a3ce4283ce78eda77commit | diff
[CUDA] Enable fusing FP ops (-ffp-contract=fast) for CUDA by default.

This matches default nvcc behavior and gives substantial
performance boost on GPU where fmad is much cheaper compared to add+mul.

Differential Revision: http://reviews.llvm.org/D20341

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@270094 91177308-0d34-0410-b5e6-96231b3b80d8
lib/Frontend/CompilerInvocation.cpp diff | blob | history
test/CodeGenCUDA/fp-contract.cu [new file with mode: 0644] blob
Unnamed repository; edit this file 'description' to name the repository.