Pyculib

High Performance Computing

Pyculib is a package that provides access to several numerical libraries that are optimized for performance on NVidia GPUs.

Pyculib was originally part of Accelerate, developed by Anaconda, Inc.

The current version, 1.0.1, was released on July 27, 2017.

Note

Pyculib is currently available for archival purposes and is not receiving updates. We think that CuPy provides a much more complete interface to standard GPU algorithms, and CuPy arrays work with Numba-compiled GPU kernels now.

Features

• Bindings to the following CUDA libraries:

  • cuBLAS
  • cuFFT
  • cuSPARSE
  • cuRAND
  • CUDA Sorting algorithms from the CUB and Modern GPU libraries

Installation

This section contains information related to:

• Requirements
• Installing Pyculib
• Updating Pyculib

User guide

This section contains information related to:

• CUDA libraries
• Environment variables

Release notes

• Release notes