[Scipy-tickets] [SciPy] #1515: Arpack complex_nonsymmetric_modes failure on 64-bit Linux

SciPy Trac scipy-tickets@scipy....
Mon Sep 26 10:57:03 CDT 2011


#1515: Arpack complex_nonsymmetric_modes failure on 64-bit Linux
---------------------------------+------------------------------------------
 Reporter:  rgommers             |       Owner:  wnbell
     Type:  defect               |      Status:  new   
 Priority:  normal               |   Milestone:  0.10.0
Component:  scipy.sparse.linalg  |     Version:  0.9.0 
 Keywords:                       |  
---------------------------------+------------------------------------------

Comment(by pv):

 Ok, it seems that this could be worse than I thought. However, it's still
 an issue with tolerances, so non-fatal.

 This is true also for the failing GMRES case -- it does not IMHO represent
 a realistic use case: ARPACK + iterative inversion (!) is most useful only
 for large matrices, and then the user typically has to set a tolerance
 larger than the machine precision so that the computation does not take
 ages.

 The best idea is probably to specify a larger tolerance for *only* the
 single-precision `aslinearoperator`, and leave the other tests alone (so
 we won't weaken them).

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1515#comment:3>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list