[Numpy-discussion] numpy.linalg.eig

Nadav Horesh nadavh@visionsense....
Wed Sep 17 08:55:54 CDT 2008


Did you edit site.cfg to use optimised libraries (ATLAS, MKL, whatever)?

  Nadav

On Tue, 2008-09-16 at 17:33 +0100, Tommy Carstensen wrote:

> Dear Robert Kern,
> 
> Thanks for your interest in helping me.
> 
> Previously I used LinearAlgebra.Heigenvectors. I should have switched to
> numpy.linalg.eigh instead of numpy.linalg.eig, since I work with Hermitian
> (non-complex and symmetric) matrices.
> 
> Kind regards,
> Tommy
> 
> -----Original Message-----
> From: numpy-discussion-bounces@scipy.org
> [mailto:numpy-discussion-bounces@scipy.org] On Behalf Of Robert Kern
> Sent: 15 September 2008 17:39
> To: Discussion of Numerical Python
> Subject: Re: [Numpy-discussion] numpy.linalg.eig
> 
> On Mon, Sep 15, 2008 at 09:50, Tommy Carstensen <e-alerts@proteinkemi.dk>
> wrote:
> > To the numpy mailing list,
> >
> > Previously I used python2.3 and the module LinearAlgebra for matrix
> > diagonalization. I recently upgraded to python2.5 and I am now using
> > numpy.linalg.eig for matrix diagonalization. This is approximately 3 times
> > slower than previously. I am using the same work station as previously and
> > it has all lapack libraries installed to my knowledge. Does anyone have an
> > explanation as to why I am experiencing the reduced speed of calculation?
> 
> Although you have the same LAPACK libraries installed, Numeric and
> numpy may have been built differently, and numpy may not be using the
> same libraries. Can you check?
> 
> > This is my first post to this mailing list, so I hope this mail complies
> > with the mailing list standards.
> 
> Yes, certainly.
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://projects.scipy.org/pipermail/numpy-discussion/attachments/20080917/f3f9562e/attachment.html 


More information about the Numpy-discussion mailing list