Re: blas, lapack, slatec,scalapack
Forgot to add that Herng-Jeng Jou from SAL describes all the optimized
BLAS options for Linux:
http://www.kachinatech.com/~hjjou/scilib/opt_blas.html
On Mon, 3 Aug 1998, John Lapeyre wrote:
> The free one looks like its going slowly. I downloaded it and
> built it, its three or four routines, blas level 1 . Is this all there is
> to level 1 ?
No, they are not a complete set of blas1 (missing for instance Xasum,
iXamax, Xswap, Xrotm, etc). Would there be a problem in constructing a
complete blas lib from Kessler's routines complemented by all the f77 ones
from netlib?
Too bad that the blas1 are still "too simple" to get a lot of improvement
from coding them in assembler. Everybody says it's in the blas3 where
assembler makes a world of difference. Oh well, if I can get a hold of a
good book on assembler with a good treatment of ix86 floating point (which
seems almost impossible), I'll start learning and helping myself.
> The non-free project prohibits redistribution, so we can't even
> put it in non-free.
Yes, too bad Intel insists on making the assembler source not available.
Many people (including myself) have asked Intel about a Math Kernel
Library for Linux, and they always said it was not planned. Good thing
that the Sandia supercomputer does not run NT (they know better), and that
the Sandia-Intel team are sympathetic to Linux.
I still don't understand why they don't release the assembler source, to
me it seems they don't have anything to lose. Or is it that AMD, Cyrix, et
al. could infer details about the P-II that they don't already know?
> I have looked at slatec . But it seems to requre blas and blacs
> and and || package like pvm . Its a mess . I got blacs to build , but
> only after installing pvm3-dev . slatec will build , but the test
> programs won't build yet-- problem with trailing underscores when mixing C
> and Fortran. I really don't have the patience or desire to finish this
> probably, but if someone wants to do it, you can have my notes and diffs.
Among my colleagues, I have not seen the slatec libraries used that much.
(I can't say anything about other areas of Sci.Comp). So I'd prefer to see
a reliable and stable distribution of the blas and lapack first; then go
on with the blacs, pblas, scalapack and plapack. Just an opinion, though.
Cheers,
Jose
=============================================================================
Jose L. Marín jose@ma.hw.ac.uk
Dept of Maths marin@wanda.unizar.es
Heriot-Watt University
Edinburgh EH14 4AS, U.K.
Phone: +44 131 4513230
Fax: +44 131 451 3249
Former address: Dept. de Física de la Materia Condensada
Facultad de Ciencias, Universidad de Zaragoza
50009 Zaragoza, SPAIN
=============================================================================
--
To UNSUBSCRIBE, email to debian-devel-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: