[Scipy-tickets] [SciPy] #231: A wrapper for ARPACK

SciPy scipy-tickets at scipy.net
Tue Oct 10 10:58:40 CDT 2006


#231: A wrapper for ARPACK
--------------------------+-------------------------------------------------
 Reporter:  nils          |        Owner:  somebody
     Type:  enhancement   |       Status:  new     
 Priority:  normal        |    Milestone:          
Component:  scipy.sparse  |      Version:          
 Severity:  normal        |   Resolution:          
 Keywords:                |  
--------------------------+-------------------------------------------------
Comment (by NeilenMarais):

 Hi.

 I've also tried wrapping ARPACK with similar results. I actually only
 tried wrapping the general (i.e. non-symmetric) matrix routines. I
 narrowed the problem down the call to the post-proccessing <s,d>neupd
 routine. If eigenvectors aren't requested, it does calculate the correct
 eigenvalues reliably. If eigenvectors are requested, repeated runs will
 eventually turn up the correct results.

 I have tried all sorts of things (super-sizing some of the workspaces) to
 no avail. I wonder if the problems are caused by a bug in F2PY, or I guess
 some subtle ARPACK bug that isn't aparent if there is no python+scipy also
 in the process. I'm interested to hear that the symmetric and complex
 routines work properly.

 I have also gotten the shift-invert mode working, again without
 calculating eigenvectors, since I'm solving the generalised eigensystem
 produced by my FEM code. By combining it with linsolve.splu() I can solve
 fair sized problems fairly quickly. Using instead the unpreconditioned
 iterative solvers seems to run a very very long time... But at least I can
 solve large problems without killing my system :) Actually I don't know if
 it works properly yet, since I'm still waiting.

-- 
Ticket URL: <http://projects.scipy.org/scipy/scipy/ticket/231#comment:2>
SciPy <http://www.scipy.org/>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list