[Scipy-tickets] [SciPy] #1689: Bug in ARPACK interface with MKL

SciPy Trac scipy-tickets@scipy....
Mon Jul 9 13:29:27 CDT 2012


#1689: Bug in ARPACK interface with MKL
---------------------------------+------------------------------------------
 Reporter:  jmhite               |       Owner:  somebody   
     Type:  defect               |      Status:  new        
 Priority:  normal               |   Milestone:  Unscheduled
Component:  scipy.sparse.linalg  |     Version:  0.10.0     
 Keywords:                       |  
---------------------------------+------------------------------------------
 When calling the sparse eigenvalue routine eigs in sparse.linalg, the
 following traceback is generated:

 {{{
 /abdelkhalik_data/home/jmhite/.python/lib/python2.7/site-
 packages/scipy/sparse/linalg/eigen/arpack/arpack.pyc in eigs(A, k, M,
 sigma, which, v0, ncv, maxiter, tol, return_eigenvectors, Minv, OPinv,
 OPpart)
    1274
    1275     while not params.converged:
 -> 1276         params.iterate()
    1277
    1278     return params.extract(return_eigenvectors)

 /abdelkhalik_data/home/jmhite/.python/lib/python2.7/site-
 packages/scipy/sparse/linalg/eigen/arpack/arpack.pyc in iterate(self)
     728                                     self.tol, self.resid, self.v,
 self.iparam,
     729                                     self.ipntr,  self.workd,
 self.workl,
 --> 730                                     self.info)
     731         else:
     732             self.ido, self.tol, self.resid, self.v, self.iparam,
 self.ipntr, self.info =\

 ValueError: need more than 6 values to unpack
 }}}

 This happens when solving both the standard eigenvalue problem and the
 generalized form, for every combination of arguments that I tried (varying
 which, k tol etc). Importantly, this only occurs on a system where I have
 scipy linked against the most recent version of MKL, testing with a
 version linked against ATLAS works fine.

 The following code will generate the error, when using the MKL-linked
 version:

 {{{
 import numpy
 from scipy.sparse.linalg import eigs

 x = numpy.random.randn(50, 50)
 l, v = eigs(x)
 }}}

 Here's show_config():

 {{{
 umfpack_info:
   NOT AVAILABLE
 lapack_opt_info:
     libraries = ['mkl_rt', 'pthread']
     library_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/lib/intel64']
     define_macros = [('SCIPY_MKL_H', None)]
     include_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/include']
 blas_opt_info:
     libraries = ['mkl_rt', 'pthread']
     library_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/lib/intel64']
     define_macros = [('SCIPY_MKL_H', None)]
     include_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/include']
 lapack_mkl_info:
     libraries = ['mkl_rt', 'pthread']
     library_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/lib/intel64']
     define_macros = [('SCIPY_MKL_H', None)]
     include_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/include']
 blas_mkl_info:
     libraries = ['mkl_rt', 'pthread']
     library_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/lib/intel64']
     define_macros = [('SCIPY_MKL_H', None)]
     include_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/include']
 mkl_info:
     libraries = ['mkl_rt', 'pthread']
     library_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/lib/intel64']
     define_macros = [('SCIPY_MKL_H', None)]
     include_dirs =
 ['/abdelkhalik_data/home/jmhite/.prefix/opt/intel/composer_xe_2011_sp1/mkl/include']
 }}}

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1689>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list