Intel compilers and libraries

[Mike Snitzer]

On Thu, Oct 09 2003 at 04:04,
C J Kenneth Tan -- Heuchera Technologies <cjtan at optimanumerics.com> wrote:

> Greg,
> 
> > Is it a 100x100 matrix LU decomposition? Well, no, because Intel's
> > MKL and the free ATLAS library run at a respectable % of peak.
> 
> Our benchmarks concentrate on xGEQRF, xGESVD, xGETRF, xGETRS, xGESV,
> xPOTRF, xPOSV, xPPTRF, xGEEV, extending to xGETRI, and xTRTRI.
> 
> Have you tried DPPSV or DPOSV on Itanium, for example?  I would be
> interested in the percentage of peak that you achieve with MKL and
> ATLAS, for up to 10000x10000 matrices.
> 
> ATLAS does not have full LAPACK implementation.

This gets ATLAS to provide its faster LAPACK routines to a full LAPACK
library: 
http://math-atlas.sourceforge.net/errata.html#completelp

Mike

-- 
Mike Snitzer                           msnitzer at lnxi.com
Linux Networx                          http://www.lnxi.com 

_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf