目录
develop_deprecated

rocSPARSE

[!CAUTION] The rocSPARSE repository is retired, please use the ROCm/rocm-libraries repository

rocSPARSE exposes a common interface that provides Basic Linear Algebra Subroutines (BLAS) for sparse computation. It’s implemented on top of AMD ROCm runtime and toolchains. rocSPARSE is created using the HIP programming language and optimized for AMD’s latest discrete GPUs.

Documentation

[!NOTE] The published rocSPARSE documentation is available at rocSPARSE in an organized, easy-to-read format, with search and a table of contents. The documentation source files reside in the rocSPARSE/docs folder of this repository. As with all ROCm projects, the documentation is open source. For more information, see Contribute to ROCm documentation.

To build our documentation locally, run the following code:

cd docs
pip3 install -r sphinx/requirements.txt
python3 -m sphinx -T -E -b html -d _build/doctrees -D language=en . _build/html

Alternatively, build with CMake:

cmake -DBUILD_DOCS=ON ...

Requirements

  • Git
  • CMake (3.5 or later)
  • AMD [ROCm] 3.5 platform or later

Optional:

  • GoogleTest (required only for tests)
    • Use GTEST_ROOT to specify a location
    • If you don’t have GoogleTest installed, CMake automatically downloads and builds it

Build and install

  1. Build rocSPARSE using the install.sh script.

    # Clone rocSPARSE using git
    git clone https://github.com/ROCm/rocSPARSE.git
    
    # Go to rocSPARSE directory
    cd rocSPARSE
    
    # Run install.sh script
    # Command line options:
    #   -h|--help         - prints help message
    #   -i|--install      - install after build
    #   -d|--dependencies - install build dependencies
    #   -c|--clients      - build library clients too (combines with -i & -d)
    #   -g|--debug        - build with debug flag
    ./install.sh -dci
  2. Compile rocSPARSE (all compiler specifications are automatically determined).

    # Clone rocSPARSE using git
    git clone https://github.com/ROCm/rocSPARSE.git
    
    # Go to rocSPARSE directory, create and go to the build directory
    cd rocSPARSE; mkdir -p build/release; cd build/release
    
    # Configure rocSPARSE
    # Build options:
    #   BUILD_CLIENTS_TESTS      - build tests (OFF)
    #   BUILD_CLIENTS_BENCHMARKS - build benchmarks (OFF)
    #   BUILD_CLIENTS_SAMPLES    - build examples (ON)
    #   BUILD_VERBOSE            - verbose output (OFF)
    #   BUILD_SHARED_LIBS        - build rocSPARSE as a shared library (ON)
    CXX=/opt/rocm/bin/amdclang++ cmake -DBUILD_CLIENTS_TESTS=ON ../..
    
    # Build
    make
    
    # Install
    [sudo] make install

Unit tests and benchmarks

To run unit tests, you must build rocSPARSE with -DBUILD_CLIENTS_TESTS=ON.

# Go to rocSPARSE build directory
cd rocSPARSE; cd build/release

# Run all tests
./clients/staging/rocsparse-test

To run benchmarks, you must build rocSPARSE with -DBUILD_CLIENTS_BENCHMARKS=ON.

# Go to rocSPARSE build directory
cd rocSPARSE/build/release

# Run benchmark, e.g.
./clients/staging/rocsparse-bench -f hybmv --laplacian-dim 2000 -i 200

Issues

To submit an issue, a bug, or a feature request, use the GitHub issue tracker.

License

Our license file is located in the main repository.

关于

AMD GPU上稀疏矩阵计算的高性能BLAS库。

23.2 MB
邀请码
    Gitlink(确实开源)
  • 加入我们
  • 官网邮箱:gitlink@ccf.org.cn
  • QQ群
  • QQ群
  • 公众号
  • 公众号

版权所有:中国计算机学会技术支持:开源发展技术委员会
京ICP备13000930号-9 京公网安备 11010802032778号