Re: [Beowulf] Re: Redmond is at it, again
Joshua Baker-LePain <email@example.com> writes:
> On Thu, 10 Jun 2004 at 11:36am, Robert G. Brown wrote
> > can improve by 10-20% from that. Finally, in C one can implement linear
> > algebra code in particular using ATLAS, which can all by itself speed
> > operations up by as much as a factor of three.
> Actually, Matlab's lapack *is* ATLAS. They provide several copies of it,
> compiled for different processors, that get called based on a runtime
> architecture check. My R13 install has atlas_Athlon.so, atlas_P4.so,
> atlas_PII.so, atlas_PIII.so, and atlas_PPro.so. In addition, there are
> directions floating about for compiling your own optimized ATLAS on your
> particular box and substituting that for the ones that come with Matlab.
This is much like the setup with atlas on Debian, where quite a few
packages now are transparently accelerated via the selection of an
atlas .so appropriate for the running machine. Custom builds/timings
are supported too with one command.
On i386, we have PII, sse1, sse2, and 3dnow builds
On ppc, generic and altivec
On sparc, generic and v9
On alpha, generic and ev5
Others just have generic builds at present.
Here are a list of packages making use of this facility. Packages
using the libs mentioned below also thereby use the atlas
> Joshua Baker-LePain
> Department of Biomedical Engineering
> Duke University
> Beowulf mailing list, Beowulf@beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
Camm Maguire firstname.lastname@example.org
"The earth is but one country, and mankind its citizens." -- Baha'u'llah