diff options
author | Gheorghe-Teodor Bercea <gheorghe-teod.bercea@ibm.com> | 2017-08-09 15:56:54 +0000 |
---|---|---|
committer | Gheorghe-Teodor Bercea <gheorghe-teod.bercea@ibm.com> | 2017-08-09 15:56:54 +0000 |
commit | 7984a2104f8822c9222dc8ca3b1b30b05b5d61af (patch) | |
tree | 780d96e1655107ebd43a42a4e499c83ef0bb88c9 /lib/Driver/ToolChains/Cuda.cpp | |
parent | 0420738bf55dfd0302ca318299a19905a27f7c18 (diff) |
[OpenMP] Add flag for overwriting default PTX version for OpenMP targets
Summary:
This flag "--fopenmp-ptx=" enables the overwriting of the default PTX version used for GPU offloaded OpenMP target regions: "+ptx42".
Reviewers: arpith-jacob, caomhin, carlo.bertolli, ABataev, Hahnfeld, jlebar, hfinkel, tstellar
Reviewed By: ABataev
Subscribers: rengolin, cfe-commits
Differential Revision: https://reviews.llvm.org/D29660
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@310489 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'lib/Driver/ToolChains/Cuda.cpp')
-rw-r--r-- | lib/Driver/ToolChains/Cuda.cpp | 8 |
1 files changed, 7 insertions, 1 deletions
diff --git a/lib/Driver/ToolChains/Cuda.cpp b/lib/Driver/ToolChains/Cuda.cpp index 1ff41ce66d..b6a7b829f5 100644 --- a/lib/Driver/ToolChains/Cuda.cpp +++ b/lib/Driver/ToolChains/Cuda.cpp @@ -481,7 +481,13 @@ void CudaToolChain::addClangTargetOptions( // than LLVM defaults to. Use PTX4.2 which is the PTX version that // came with CUDA-7.0. CC1Args.push_back("-target-feature"); - CC1Args.push_back("+ptx42"); + + if (DeviceOffloadingKind == Action::OFK_OpenMP) + CC1Args.push_back( + DriverArgs.getLastArgValue(options::OPT_fopenmp_ptx_EQ, + "+ptx42").data()); + else + CC1Args.push_back("+ptx42"); } void CudaToolChain::AddCudaIncludeArgs(const ArgList &DriverArgs, |