[Numpy-tickets] [NumPy] #627: linalg.svd takes 100% CPU and never ends

NumPy numpy-tickets@scipy....
Sun Dec 9 09:11:54 CST 2007


#627: linalg.svd takes 100% CPU and never ends
--------------------------+-------------------------------------------------
 Reporter:  Czarny        |        Owner:  somebody
     Type:  defect        |       Status:  new     
 Priority:  high          |    Milestone:  1.0.5   
Component:  numpy.linalg  |      Version:  none    
 Severity:  normal        |   Resolution:          
 Keywords:  svd           |  
--------------------------+-------------------------------------------------
Comment (by Czarny):

 Replying to [comment:8 cdavid]:
 >
 > > PLD Linux, a post RedHat derivate: http://pld-linux.org
 >
 > Mm, ok, I don't know this one :). One thing you could try is the
 following: compile numpy, and disable LAPACK (I strongly suspect that
 lapack is wrongly packaged)
 >
 > {{{
 > LAPACK=None python setup.py install
 > }}}
 >

 Ok -- I've tried just that a minute ago. It actually works now!

 The show_config after the recompiling and reinstalling:
 {{{
 [czarny@kacper /tmp]$ python -c "import numpy; numpy.show_config();"
 blas_info:
     libraries = ['blas']
     library_dirs = ['/usr/lib']
     language = f77

 lapack_info:
     libraries = ['lapack']
     library_dirs = ['/usr/lib']
     language = f77

 atlas_threads_info:
   NOT AVAILABLE

 blas_opt_info:
     libraries = ['blas']
     library_dirs = ['/usr/lib']
     language = f77
     define_macros = [('NO_ATLAS_INFO', 1)]

 atlas_blas_threads_info:
   NOT AVAILABLE

 lapack_opt_info:
     libraries = ['lapack', 'blas']
     library_dirs = ['/usr/lib']
     language = f77
     define_macros = [('NO_ATLAS_INFO', 1)]

 atlas_info:
   NOT AVAILABLE

 lapack_mkl_info:
   NOT AVAILABLE

 blas_mkl_info:
   NOT AVAILABLE

 atlas_blas_info:
   NOT AVAILABLE

 mkl_info:
   NOT AVAILABLE
 }}}

 So the question is how to correctly build and pack the LAPACK stuff to
 make them work. (what exactly is that stuff anyway ;p)

 > The report the problem. Another thing which would be useful to know is
 whether you used g77 or gfortran (which one the PLD gcc-fortran package
 gives you ?)
 >

 The buildlog I've posted link to shows clearly that gfortran is used:
 [czarny@kacper /tmp]$ rpm -ql gcc-fortran | grep bin
 /usr/bin/g95
 /usr/bin/gfortran
 /usr/bin/i686-pld-linux-gfortran

 So we've got no g77, only g95 but then again the log clearly shows gfotran
 being called.

 Thx man -- that'll make my master thesis a lot easier.
 But still as a PLD developer I'd like to fix the issue a little more
 kosher then LAPACK=None ;)

-- 
Ticket URL: <http://scipy.org/scipy/numpy/ticket/627#comment:9>
NumPy <http://projects.scipy.org/scipy/numpy>
The fundamental package needed for scientific computing with Python.


More information about the Numpy-tickets mailing list