# pyBLAS

**Repository Path**: linius/pyBLAS

## Basic Information

- **Project Name**: pyBLAS
- **Description**: A BLAS wrapper written in Cython
- **Primary Language**: Python
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2014-07-24
- **Last Updated**: 2020-12-18

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

#pyBLAS
============================================
pyBLAS is a partial BLAS wrapper written in Cython.  It allows you
to call various BLAS matrix and vector routines in single and double 
precision without using the numpy/scipy wrapper -- which is in slow 
interpreted Python.  Speed ups range from nothing to over 100 times 
depending on the size of the matrices and vectors and which operations you need to
perform.

Installation
============================================
First of all make sure you have Python and Cython!

Secondly, make sure you have cblas installed.  On a unix system try:

    locate cblas.h

If that finds something for you then you have it and you also now
know where it is!

Now build it in the usual Cython way:

    python setup.py build_ext --inplace

If you get no errors, now run the verification test to check
it's computing the right things:

    python verify.py
    
Next try the single and double precision performance tests:

    python single_speed.py
    python double_speed.py

These will tell you whether it's worth switching your code
to use pyBLAS or not for your machine and matrix sizes.