We plan to provide ArrayFire packages for as many Linux distributions as possible. Currently, only Ubuntu packages are available.

Given below is the outline of this document for quick navigation.

• ArrayFire Packages
• How To Add Apt Repository In Ubuntu
• Docker Specific Instructions
  • Ubuntu

apt-key adv --fetch-key https://repo.arrayfire.com/GPG-PUB-KEY-ARRAYFIRE-2020.PUB
echo "deb [arch=amd64] https://repo.arrayfire.com/debian all main" | tee /etc/apt/sources.list.d/arrayfire.list

To install ArrayFire's CUDA backend packages or an all-backends package such as arrayfire or arrayfire-dev, please add NVIDIA provided CUDA Ubuntu repository if you don't have it already.

• Development Packages

  • arrayfire - Development files for all backends of ArrayFire
  • arrayfire-cpu3-dev - ArrayFire CPU backend development files
  • arrayfire-cuda3-dev - ArrayFire CUDA backend development files
  • arrayfire-opencl3-dev - ArrayFire OpenCL backend development files
  • arrayfire-unified3-dev - ArrayFire Unified backend development files
  • arrayfire-headers - ArrayFire C/C++ Header files
  • arrayfire-cmake - Configuration files for using ArrayFire via CMake
  • arrayfire-doc - ArrayFire Doxygen Documentation
  • arrayfire-examples - ArrayFire Samples C++ source files

• Runtime Packages

  • arrayfire-cpu3-mkl - ArrayFire CPU backend shared libraries using Intel MKL
  • arrayfire-cpu3-openblas - ArrayFire CPU backend shared libraries using OpenBLAS
  • arrayfire-cuda3-cuda-10-2 - ArrayFire CUDA backend shared libraries built using CUDA 10.2 (supports older GPUs)
  • arrayfire-cuda3-cuda-11-6 - ArrayFire CUDA backend shared libraries built using CUDA 11.6 (Adds support or new GPUs)
  • arrayfire-opencl3-mkl - ArrayFire OpenCL backend shared libraries using Intel MKL for CPU fallback routines
  • arrayfire-opencl3-openblas - ArrayFire OpenCL backend shared libraries using OpenBLAS for CPU fallback routines
  • arrayfire-unified3 - ArrayFire Unified backend shared libraries. Requires other runtime backends to function.

ℹ️ All installation and associated commands need sudo privileges if they modify system files.

The commands for Ubuntu docker image are same as earlier except that you might need some additional packages installed on the image to successfully the run the commands mentioned in ubuntu section.

apt install -y gnupg2 ca-certificates apt-utils software-properties-common

Similar to a regular system, if CUDA backend is required please add NVIDIA provided CUDA Ubuntu repository.