[SciPy-Dev] Faster expm

[Alex Leach]

On Thursday 15 Nov 2012 17:36:51 Nathaniel Smith wrote:
 > On Thu, Nov 15, 2012 at 12:46 PM, Alex Leach <beamesleach at gmail.com> wrote:
> > Wrapping EXPOKIT for Python has been done before, by a colleague of the
> > original author, so I can't take all the credit. A presentation he gave,
> > with some usage and build instructions can be found here (Pg 16-22):
> > http://sf.anu.edu.au/~mhk900/Python_Workshop/short.pdf
> 
> Did you copy his code, and if so, do we need to worry about its
> license? (I guess to the extent that it's just a literal listing of
> the functions inside the .f file, there's nothing copyrightable,
> but...)

No, not really. He's put some code up from the auto-generated signature file, 
which we'd both edited. I was double-checking his matvec signature against 
what I had generated and edited, and I did copy the argument names he used.

The usage example from Python seems incomplete, and did not inspire my Python 
code at all. The Python code there demonstrates how to call dmexpv, which I 
don't think would be used by SciPy, either.. dmexpv computes Krylov projection 
for Sparse Markov matrices, if that makes sense. I thought we'd just want 
dgexpv, which is for general sparse matrices.

The only bit which I copied was the modification to the matvec calls in 
expokit.f (bottom of Page 20). As it's a modification to Expokit, which we've 
been granted permission to use, modify and distribute, I think we're safe.

Cheers,
Alex
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scipy-dev/attachments/20121115/33f91338/attachment.html>