Release Notes

v3.0.2

Bug Fixes

Documentation

  • Added missing documentation for constant()
  • Added missing documentation for array::scalar()
  • Added supported input types for functions in arith.h

v3.0.1

Bug Fixes

  • Fixed header to work in Visual Studio 2015
  • Fixed a bug in batched mode for FFT based convolutions
  • Fixed graphics issues on OSX
  • Fixed various bugs in visualization functions

Other improvements

  • Improved fractal example
  • New OSX installer
  • Improved Windows installer
    • Default install path has been changed
  • Fixed bug in machine learning examples


v3.0.0

Major Updates

  • ArrayFire is now open source
  • Major changes to the visualization library
  • Introducing handle based C API
  • New backend: CPU fallback available for systems without GPUs
  • Dense linear algebra functions available for all backends
  • Support for 64 bit integers

Function Additions

  • Data generation functions
  • Computer Vision Algorithms
    • features()
      • A data structure to hold features
    • fast()
      • FAST feature detector
    • orb()
      • ORB A feature descriptor extractor
  • Image Processing
  • Linear Algebra
  • Other functions
    • lookup() - lookup indices from a table
      • batchFunc() - helper function to perform batch operations
  • Visualization functions
    • Support for multiple windows
    • window.hist()
      • Visualize the output of the histogram
  • C API
    • Removed old pointer based C API
    • Introducing handle base C API
    • Just In Time compilation available in C API
    • C API has feature parity with C++ API
    • bessel functions removed
    • cross product functions removed
    • Kronecker product functions removed

Performance Improvements

  • Improvements across the board for OpenCL backend

API Changes

  • print is now af_print()
  • seq(): The step parameter is now the third input
    • seq(start, step, end) changed to seq(start, end, step)
  • gfor(): The iterator now needs to be seq()

Deprecated Function APIs

Deprecated APIs are in af/compatible.h

Known Issues

  • OpenCL backend issues on OSX
    • AMD GPUs not supported because of driver issues
    • Intel CPUs not supported
    • Linear algebra functions do not work on Intel GPUs.
  • Stability and correctness issues with open source OpenCL implementations such as Beignet, GalliumCompute.