Difference between revisions of "User:Richli/Scientific Python"

From ArchWiki
Jump to navigation Jump to search
(Created page with "==Introduction== Several packages exist for scientific/numerical computing using Python: typical ones include numpy, scipy, and matplotlib. These are commonly used ...")
 
Line 16: Line 16:
  
 
Show code examples to use python-fftw.
 
Show code examples to use python-fftw.
 +
 +
===Math Libraries===
 +
TODO: Add info about IMKL/ACML?

Revision as of 06:18, 21 March 2012

Introduction

Several packages exist for scientific/numerical computing using Python: typical ones include numpy, scipy, and matplotlib. These are commonly used to provide a free alternative to matlab. (TODO: mention other packages, like chaco, mayavi, scikits, etc?)

Installation

numpy, scipy, and matplotlib are all available in extra/community (TODO: add links).

Improving Performance

ATLAS

numpy and scipy use basic blas/lapack routines, but using atlas (AUR) will fine-tune these basic routines to the processor architecture (TODO: talk more about atlas).

Unfortunately, numpy does not seem to recognize atlas unless it's rebuilt with atlas already present on the system. So we use the ABS to rebuild numpy after installing atlas.

FFTW

(TODO: Add background to FFTW) Recent versions of numpy/scipy no longer use FFTW. python-fftw3 from AUR is a python wrapper to extra/fftw.

Show code examples to use python-fftw.

Math Libraries

TODO: Add info about IMKL/ACML?