[SciPy-dev] ARPACK (large scale eigen-solver) wrappers

Neilen Marais nmarais at sun.ac.za
Mon Sep 4 09:18:44 CDT 2006


As promised a while ago I've started messing around with ARPACK wrappers
for SciPy. The typical form of an ARPACK solution is that you iteratively
call an ARPACK routine which then asks you to do certain matrix-vector
operations on the matrix that you need the eigenvalues of. I've only
played around with the simplest mode of the dnaupd routine where it will
only ask you for matrix-vector products.

I seem to have a reasonable f2py wrapper for that routine. ARPACK seems to
think that it has converged after a resonable seeming number of
iterations. But getting at the actual eigenvalues requires postprocessing
using the dneupd routine. At the moment I get segfaults if I run it, or
otherwise just after. 

I'm not quite sure where the problem is. I'll spend a bit more time on it,
but if anyone is interested I could put up a bzr archive of what I've done
so far. It's not a whole lot yet, though :)


you know its kind of tragic 
we live in the new world
but we've lost the magic
-- Battery 9 (www.battery9.co.za)

More information about the Scipy-dev mailing list