[SciPy-user] SciPy-user Digest, Vol 31, Issue 35

Curtis Cooper curtis at lpl.arizona.edu
Thu Mar 23 12:00:44 EST 2006


> Message: 5
> Date: Thu, 23 Mar 2006 16:10:14 +0000
> From: Peter Bowyer <peter at mapledesign.co.uk>
> Subject: [SciPy-user] The "Performance Python with Weave" article
> To: "SciPy Users List" <scipy-user at scipy.net>
> Message-ID: <7.0.1.0.0.20060323155841.023b7ec0 at mapledesign.co.uk>
> Content-Type: text/plain; charset="us-ascii"; format=flowed
>
> Hi,
>
> Referring to
> http://old.scipy.org/documentation/weave/weaveperformance.html (I
> can't find it on the new site - weave appears to have vanished), I
> note that there is no date on it and neither are there any details of
> the versions of the libraries uses.  Could anyone enlighten me as to
> how accurate it is, as I was planning to reference it in my
> dissertation writeup as an example of how Python code can be speeded
> up if required.
>

I use weave in my own projects.  For computationally intensive loops that
cannot be written with NumPy's vector operations (use that if you can),
switching to C++ often gives you ~100 times performance improvement above
the same algorithm written in Python.  Don't do it if you don't need to,
but for performance critical code that must be executed repeatedly in your
calculations, weave is a great solution.

Cheers,
Curtis




More information about the SciPy-User mailing list