ArrayFire is a high performance library for parallel computing with an easy-to-use API. It enables users to write scientific computing code that is portable across CUDA, OpenCL and CPU devices. This project provides Python bindings for the ArrayFire library.

Documentation for this project can be found over here.

# Monte Carlo estimation of pi
def calc_pi_device(samples):
    # Simple, array based API
    # Generate uniformly distributed random numers
    x = af.randu(samples)
    y = af.randu(samples)
    # Supports Just In Time Compilation
    # The following line generates a single kernel
    within_unit_circle = (x * x + y * y) < 1
    # Intuitive function names
    return 4 * af.count(within_unit_circle) / samples

Choosing a particular backend can be done using af.set_backend(name) where name is either "cuda", "opencl", or "cpu". The default device is chosen in the same order of preference.

Currently, this project is tested only on Linux and OSX. You also need to have the ArrayFire C/C++ library installed on your machine. You can get it from the following sources.

• Download and install binaries
• Build and install from source

Please check the following links for dependencies.

• Linux dependencies
• OSX dependencies

Install the last stable version:

pip install arrayfire

Install the development version:

pip install git+git://github.com/arrayfire/arrayfire-python.git@devel

Installing offline

cd path/to/arrayfire-python
python setup.py install

Post Installation

Please follow these instructions to ensure the arrayfire-python can find the arrayfire libraries.

To run arrayfire tests, you can run the following command from command line.

python -m arrayfire.tests

• Slack Chat
• Google Groups

The ArrayFire library is written by developers at ArrayFire LLC with contributions from several individuals.

The developers at ArrayFire LLC have received partial financial support from several grants and institutions. Those that wish to receive public acknowledgement are listed below:

This material is based upon work supported by the DARPA SBIR Program Office under Contract Numbers W31P4Q-14-C-0012 and W31P4Q-15-C-0008. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the DARPA SBIR Program Office.