• About
• Documentation
• Questions
• Speed
• Contact
• Download

Stable Version

• bandicoot-4.0.1.tar.xz

Development Version

• git repo

Related Software

• Armadillo - CPU focused C++ library for linear algebra & scientific computing
• ensmallen - C++ library for efficient numerical optimisation
• mlpack - extensive library of machine learning algorithms

Installation Notes

• See the README file in the .tar.xz package for full installation instructions
  
  
• If you encounter any bugs or regressions, please report them
  
  
• If you use Bandicoot in your research and/or software, please cite the associated papers;
  citations are useful for the continued development and maintenance of the library
  
  

• Linux based operating systems (eg. Fedora, Ubuntu, Red Hat, SUSE, Debian, etc)
  
  
  • Before installing Bandicoot, first install:
    • OpenBLAS along with the corresponding development/header files
    • the CUDA toolkit, if using the CUDA backend
    • OpenCL libraries and runtime, if using the OpenCL backend (consult your system package manager)
    • CLBlast or clBLAS along with the corresponding development/header files, if using the OpenCL backend
    
     
  • Recommended packages to install before installing Bandicoot:
    • Fedora & Red Hat:
      • Install the cmake and openblas-devel packages
      • For the OpenCL backend, install opencl-headers, and one of: intel-opencl, rocm-opencl, or another OpenCL runtime
      • For the OpenCL backend with CLBlast, the clblast-devel package should be installed
      • For the CUDA backend, enable the non-free RPM Fusion repository, and install the xorg-x11-drv-nvidia-cuda, xorg-x11-drv-nvidia-cuda-libs, and xorg-x11-nvidia-cuda-devel packages
    
    
    • Ubuntu & Debian:
      • Install the cmake and libopenblas-dev packages
      • For the OpenCL backend, install opencl-headers, libclblast-dev, and one of: intel-opencl-icd, beignet-opencl-icd, nvidia-opencl-icd, or pocl-opencl-icd
      • For the CUDA backend, install nvidia-cuda-toolkit and nvidia-cuda-dev
    
    

• macOS
  
  
  • The "Accelerate" framework is used for accessing BLAS and LAPACK functions; see the README file in the package for more information
     
  • OpenCL support is installed automatically with Xcode
     
  • CLBlast can be installed via the Homebrew package manager: brew install clblast
     
  • CUDA support can be installed via the NVIDIA-provided installers
     

• Windows
  
  
  • Bandicoot has so far not been thoroughly tested on Windows. Contributions to improve Windows support are appreciated.
    
  
  
  • Caveat: for any high performance scientific/engineering workloads, we strongly recommend using a Linux based operating system