ArrayFire Python Wrapper


ArrayFire is a high performance scientific computing library with an easy to use API.

>>> # 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

Programs written using ArrayFire are portable across CUDA, OpenCL and CPU devices.

The default backend is chosen in the following order of preference based on the available libraries:

  1. CUDA
  2. OpenCL
  3. CPU

The backend can be chosen at the beginning of the program by using the following function

>>> af.backend.set(name)

where name is one of ‘cuda’, ‘opencl’ or ‘cpu’.

The functionality provided by ArrayFire spans the following domains:

  1. Vector Algorithms
  2. Image Processing
  3. Signal Processing
  4. Computer Vision
  5. Linear Algebra
  6. Statistics


arrayfire.algorithm Vector algorithms (sum, min, sort, etc).
arrayfire.arith Math functions (sin, sqrt, exp, etc).
arrayfire.array Array class and helper functions.
arrayfire.base Implementation of BaseArray class.
arrayfire.bcast Function to perform broadcasting operations.
arrayfire.blas BLAS functions (matmul, dot, etc)
arrayfire.cuda Functions specific to CUDA backend. Functions to create and manipulate arrays.
arrayfire.device Functions to handle the available devices in the backend.
arrayfire.features Features class used for Computer Vision algorithms. Graphics functions (plot, image, etc).
arrayfire.image Image processing functions.
arrayfire.index Index and Seq classes used in indexing operations.
arrayfire.interop Interop with other python packages.
arrayfire.lapack Dense Linear Algebra functions (solve, inverse, etc).
arrayfire.library Module containing enums and other constants.
arrayfire.opencl Functions specific to OpenCL backend.
arrayfire.signal Signal processing functions (fft, convolve, etc).
arrayfire.statistics Statistical algorithms (mean, var, stdev, etc).
arrayfire.timer Functions to time arrayfire.
arrayfire.util Utility functions to help with Array metadata. Computer vision functions (FAST, ORB, etc)