llvm icon indicating copy to clipboard operation
llvm copied to clipboard

Compiling with Spack Installed GCC Errors

Open kaanolgu opened this issue 1 year ago • 1 comments

Describe the bug

I am trying to install with spack provided GCC, followed this solution here for external CUDA support but having similar problem for non default located GCC installation.

It's a long line of errors related with GLIBCXX like :

/bin/ld: /workspace/tools/llvm/build/bin/../lib/libsycl.so: undefined reference to std::filesystem::__cxx11::directory_iterator::directory_iterator(std::filesystem::__cxx11::path const&, std::filesystem::directory_options, std::error_code*)@GLIBCXX_3.4.26'

Seems like something is not linking properly with my spack load [email protected] is there any other lines I need to set like the CUDA_LIB_PATH for non default GCC ?

Because if I use the system GCC with module load everything works fine

To reproduce

git clone https://github.com/intel/llvm.git
spack load [email protected]
CUDA_LIB_PATH=$(spack location -i cuda%$COMPILER@${COMPVER})/lib64/stubs CC=gcc CXX=g++ python ${DPCPP_HOME}/llvm/buildbot/configure.py --cuda --cmake-opt="-DCUDA_TOOLKIT_ROOT_DIR=$(spack location -i cuda%$COMPILER@${COMPVER})" -t Release
CUDA_LIB_PATH=$(spack location -i cuda%$COMPILER@${COMPVER})/lib64/stubs CC=gcc CXX=g++ python ${DPCPP_HOME}/llvm/buildbot/compile.py -j8

Environment

  • OS: Linux
  • Target device and vendor: Nvidia GPU with Intel LLVM Clang compiler
  • DPC++ version: fda7dc722161 ( 2024.2.0) also sycl branch gives the same error
  • Dependencies version: cuda 12.6

Additional context

No response

kaanolgu avatar Oct 13 '24 22:10 kaanolgu