Reikna, a pure Python GPGPU library

Reikna is a library containing various GPU algorithms built on top of PyCUDA and PyOpenCL. The main design goals are:

• separation of computation cores (matrix multiplication, random numbers generation etc) from simple transformations on their input and output values (scaling, typecast etc);
• separation of the preparation and execution stage, maximizing the performance of the execution stage at the expense of the preparation stage (in other words, aiming at large simulations)
• partial abstraction from CUDA/OpenCL

The installation is as simple as

$ pip install reikna

Community resources

• Source repository on GitHub;
• Issue tracker, ibid.;
• Discussion forum on Google Groups.

Contents

• Introduction
  • CLUDA
  • Computations
  • Transformations
• Tutorial: modules and snippets
  • Snippets
  • Modules
  • Shortcuts
  • Module and snippet discovery
  • Nontrivial example
• Tutorial: basics
  • Usage of computations
  • Computations and transformations
  • Transformation tree
  • Transformation restrictions
• Tutorial: advanced topics
  • Mako basics
  • Writing a transformation
  • Writing a computation
• API reference
  • Version queries
  • Helpers
  • CLUDA layer
  • Core functionality
  • Computations
  • Transformations
• Release history
  • 0.6.8 (18 Dec 2016)
  • 0.6.7 (11 Mar 2016)
  • 0.6.6 (11 May 2015)
  • 0.6.5 (31 Mar 2015)
  • 0.6.4 (29 Sep 2014)
  • 0.6.3 (18 Jun 2014)
  • 0.6.2 (20 Feb 2014)
  • 0.6.1 (4 Feb 2014)
  • 0.6.0 (27 Dec 2013)
  • 0.5.2 (17 Dec 2013)
  • 0.5.1 (30 Nov 2013)
  • 0.5.0 (25 Nov 2013)
  • 0.4.0 (10 Nov 2013)
  • 0.3.6 (9 Aug 2013)
  • 0.3.5 (6 Aug 2013)
  • 0.3.4 (3 Aug 2013)
  • 0.3.3 (29 Jul 2013)
  • 0.3.2 (29 Jul 2013)
  • 0.3.1 (25 Jul 2013)
  • 0.3.0 (23 Jul 2013)
  • 0.2.4 (11 May 2013)
  • 0.2.3 (25 Apr 2013)
  • 0.2.2 (20 Apr 2013)
  • 0.2.1 (8 Mar 2013)
  • 0.2.0 (3 Mar 2013)
  • 0.1.0 (12 Sep 2012)
  • 0.0.1 (22 Jul 2012)

Indices and tables

• Index
• Module Index
• Search Page