Using Clang Tools on CUDA Programs

Project Infrastructure LLVM Dev List Archives
1

Hello,

I’m not sure if this is the right place to post this question. If not, I apologize and would appreciate if you redirect me.

I recently built and installed llvm/clang in order to build a CUDA source-to-source translator using Clang. I followed these instructions: http://clang.llvm.org/docs/LibASTMatchersTutorial.html.

The built-in tools (like clang-check) work fine on C/C++ programs, but they don’t work on CUDA. If I execute the following command (regardless of what kernel.cu contains) I get the error shown:

$ clang-check kernel.cu –

error: unable to handle compilation, expected exactly one compiler job in ’ “/usr/local/bin/clang-check” “-cc1” “-triple” “nvptx64-nvidia-cuda” “-aux-triple” “x86_64-unknown-linux-gnu” “-fcuda-target-overloads” “-fcuda-disable-target-call-checks” “-fsyntax-only” “-disable-free” “-main-file-name” “kernel.cu” “-mrelocation-model” “static” “-mthread-model” “posix” “-mdisable-fp-elim” “-fmath-errno” “-no-integrated-as” “-mconstructor-aliases” “-fcuda-is-device” “-mlink-cuda-bitcode” “/usr/local/cuda/nvvm/libdevice/libdevice.compute_20.10.bc” “-target-feature” “+ptx42” “-target-cpu” “sm_20” “-dwarf-column-info” “-resource-dir” “/usr/local/bin/…/lib/clang/3.8.0” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/c++/4.8” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/x86_64-linux-gnu/c++/4.8” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/x86_64-linux-gnu/c++/4.8” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/c++/4.8/backward” “-internal-isystem” “/usr/local/include” “-internal-isystem” “/usr/local/bin/…/lib/clang/3.8.0/include” “-internal-externc-isystem” “/include” “-internal-externc-isystem” “/usr/include” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/c++/4.8” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/x86_64-linux-gnu/c++/4.8” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/x86_64-linux-gnu/c++/4.8” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/c++/4.8/backward” “-internal-isystem” “/usr/local/cuda/include” “-include” “__clang_cuda_runtime_wrapper.h” “-fdeprecated-macro” “-fno-dwarf-directory-asm” “-fdebug-compilation-dir” “//test” “-ferror-limit” “19” “-fmessage-length” “192” “-fobjc-runtime=gcc” “-fcxx-exceptions” “-fexceptions” “-fdiagnostics-show-option” “-fcolor-diagnostics” “-x” “cuda” “//test/kernel.cu”; “/usr/local/bin/clang-check” “-cc1” “-triple” “x86_64-unknown-linux-gnu” “-aux-triple” “nvptx64-nvidia-cuda” “-fcuda-target-overloads” “-fcuda-disable-target-call-checks” “-fsyntax-only” “-disable-free” “-main-file-name” “kernel.cu” “-mrelocation-model” “static” “-mthread-model” “posix” “-mdisable-fp-elim” “-fmath-errno” “-masm-verbose” “-mconstructor-aliases” “-munwind-tables” “-fuse-init-array” “-target-cpu” “x86-64” “-dwarf-column-info” “-resource-dir” “/usr/local/bin/…/lib/clang/3.8.0” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/c++/4.8” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/x86_64-linux-gnu/c++/4.8” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/x86_64-linux-gnu/c++/4.8” “-internal-isystem” “/usr/lib/gcc/x86_64-linux-gnu/4.8/…/…/…/…/include/c++/4.8/backward” “-internal-isystem” “/usr/local/include” “-internal-isystem” “/usr/local/bin/…/lib/clang/3.8.0/include” “-internal-externc-isystem” “/usr/include/x86_64-linux-gnu” “-internal-externc-isystem” “/include” “-internal-externc-isystem” “/usr/include” “-internal-isystem” “/usr/local/cuda/include” “-include” “__clang_cuda_runtime_wrapper.h” “-fdeprecated-macro” “-fdebug-compilation-dir” “//test” “-ferror-limit” “19” “-fmessage-length” “192” “-fobjc-runtime=gcc” “-fcxx-exceptions” “-fexceptions” “-fdiagnostics-show-option” “-fcolor-diagnostics” “-x” “cuda” “//test/kernel.cu”; ’

Error while processing //test/kernel.cu.

I searched thoroughly online before posting here but could not find relevant answers. Any help would be appreciated.

Thanks!

-Izzat

2

+cfe-dev

3

When the clang driver is handed a .cu file, it compiles that file more
than once, to build the host code and the device code. clang-check
can't handle this, thus the error "expected exactly one compiler job".
I expect it's a similar situation to if you asked clang-check to check
two .cpp files at once.

To make this work you'd need to teach clang-check to handle this case
properly, or you'd need to convince the clang driver to invoke the
compiler only once (for just the host, or just the device). The
relevant flags to clang proper are --cuda-device-only or
--cuda-host-only; I don't know offhand how to make those interact with
clang-check.

Good luck,
-Justin

4

Thanks Justin! This is very helpful!

-Izzat