DPC++ toolchain

cristianvasileachim — Thu, 04 Jan 2024 11:19:28 GMT

Hello,

I am trying to set up my one toolchain in my own folder on a cluster. I have no root access.

I was able to compile the intel llvm wiht nvidia gpu support on top on cuda toolkit 11.5.0. I am able to run sycl codes using the gpu selector , but when I use th ecou selector it does not find any device.

How do I compile hte intel llvm to have support for both nvidia gpus and cpus without root access?

Best regards,

Cristian

Edit:

Trying to compile with cpu fsycl-targets I get the error below:

$DPCPP_HOME/llvm/build/bin/clang++ -std=c++17 -O3 -fsycl -fsycl-targets=nvptx64-nvidia-cuda,cpu --cuda-path=$CUDA_HOME -L$GCC_INSTALL_RO OT/lib64 enumerate_gpu.cpp clang++: error: SYCL target is invalid: 'cpu' clang++: /scratch/project_2008874/sycl_workspace/llvm/clang/lib/Driver/Driver.cpp:6350: virtual bool {anonymous}::OffloadingActionBuilder::SYCLActionBuilder::initialize(): Assertion `TCIt != ToolChains.end() && "Toolchain was not created for this platform"' failed. PLEASE submit a bug report to https://github.com/intel/llvm/issues and include the crash backtrace, preprocessed source, and associated run script. Stack dump: 0. Program arguments: /scratch/project_2008874/sycl_workspace/llvm/build/bin/clang++ -std=c++17 -O3 -fsycl -fsycl-targets=nvptx64-nvidia-cuda,cpu --cuda-path=/appl/spack/v017/install-tree/gcc-11.2.0/cuda-11.5.0-mg4ztb -L/appl/spack/v017/install-tree/gcc-8.5.0/gcc-11.2.0-zshp2k/lib64 enumerate_gpu.cpp 1. Compilation construction 2. Building compilation actions Stack dump without symbol names (ensure you have llvm-symbolizer in your PATH or set the environment var `LLVM_SYMBOLIZER_PATH` to point to it): 0 clang++ 0x0000000001f513bb llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) + 219 1 clang++ 0x0000000001f4e914 2 libpthread.so.0 0x00007f3dc0a33ce0 3 libc.so.6 0x00007f3dbf61ba9f gsignal + 271 4 libc.so.6 0x00007f3dbf5eee05 abort + 295 5 libc.so.6 0x00007f3dbf5eecd9 6 libc.so.6 0x00007f3dbf6143f6 7 clang++ 0x0000000002804eba 8 clang++ 0x00000000027cd515 9 clang++ 0x0000000002812a95 clang::driver::Driver::BuildActions(clang::driver::Compilation&, llvm::opt::DerivedArgList&, llvm::SmallVector<std::pair<clang::driver::types::ID, llvm::opt::Arg const*>, 16u> const&, llvm::SmallVector<clang::driver::Action*, 3u>&) const + 11205 10 clang++ 0x000000000281786f clang::driver::Driver::BuildCompilation(llvm::ArrayRef<char const*>) + 5135 11 clang++ 0x0000000000b3bdb5 clang_main(int, char**, llvm::ToolContext const&) + 4293 12 clang++ 0x0000000000a85201 main + 33 13 libc.so.6 0x00007f3dbf607cf3 __libc_start_main + 243 14 clang++ 0x0000000000b3698e _start + 46 Aborted

Later edit:

Running the sycl-ls I get this:

LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$DPCPP_HOME/llvm/build/lib srun --time=00:15:00 --partition=gputest --account=project_2001498 --nodes=1 --ntasks-per-node=1 --cpus-per-task=1 --gres=gpu:a100:4 $DPCPP_HOME/llvm/build/bin/sycl-ls [ext_oneapi_cuda:gpu:0] NVIDIA CUDA BACKEND, NVIDIA A100-SXM4-40GB 8.0 [CUDA 12.0] [ext_oneapi_cuda:gpu:1] NVIDIA CUDA BACKEND, NVIDIA A100-SXM4-40GB 8.0 [CUDA 12.0] [ext_oneapi_cuda:gpu:2] NVIDIA CUDA BACKEND, NVIDIA A100-SXM4-40GB 8.0 [CUDA 12.0] [ext_oneapi_cuda:gpu:3] NVIDIA CUDA BACKEND, NVIDIA A100-SXM4-40GB 8.0 [CUDA 12.0]

No cpu are found.

topic DPC++ toolchain in Intel® oneAPI DPC++/C++ Compiler

DPC++ toolchain