[Scipy-tickets] [SciPy] #1503: isolve and arpack test failures on OS X

SciPy Trac scipy-tickets@scipy....
Fri Aug 26 03:19:49 CDT 2011


#1503: isolve and arpack test failures on OS X
---------------------------------+------------------------------------------
 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):

 The QMR failure seems to be a limitation of the mathematical method itself
 -- previously, that particular test case was commented out. Also MATLAB's
 implementation fails for this matrix. Why it succeeds on some compilations
 is probably due to rounding errors working around stagnating convergence.
 So I think this test should be disabled.

 The GMRES issue is more puzzling -- the matrices are so small that GMRES
 should always converge for them (if they are at all invertible). One
 should dump the matrix to a file on arpack.py:943, so that the problematic
 matrix can be analyzed separately:
 {{{
         b, info = self.ifunc(self.M, x, tol=self.tol)
         if info != 0:
             A = self.M.matmat(np.eye(self.M.shape[0]))
             np.savez('dump.npz', A=A, b=x)
 }}}

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1503#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