Dear R-help,
Here's a posting to the most recent NA-digest:
From: Robert van de Geijn <rvdg at cs.utexas.edu>
Date: Fri, 13 Dec 2002 11:15:23 -0600
Subject: Fast BLAS Libraries for Current Architectures
Recent research by Kazushige Goto, Visiting Scientist at UT-Austin,
has resulted in high-performance BLAS libraries for the Intel (R)
Pentium (R) III and 4 processors, the HP/Compaq/DEC alpha processor,
and the IBM Power 3 and 4 architectures. Performance improvements
appear to be substantial. For example, by linking this library
instead of other commonly used BLAS libraries, the performance of the
600 processor (Pentium 4 processor based) cluster at the University at
Buffalo-SUNY was increased from roughly 1.5 TeraFLOPS to 2.0 TeraFLOPS
(HPL LINPACK benchmark used for the TOP500 list. See
http://www.ccr.buffalo.edu/newsReleases/newsRelease.htm).
To help us evaluate this library, kindly visit
http://www.cs.utexas.edu/users/flame/goto/
For information on the techniques underlying the implementation, see
Kazushige Goto and Robert van de Geijn. On Reducing TLB Misses in
Matrix Multiplication. FLAME Working Note #9, The University of Texas
at Austin, Department of Computer Sciences. Technical Report
TR-2002-55. Nov. 2002.
available from
http://www.cs.utexas.edu/users/flame/pubs.html
Regards
Robert van de Geijn
=========================================
Anyone interested in trying to link R against it and see how much of a
difference it makes? (I would, except I won't have high speed net access
until I get back to office next year, and d/l large files are too
painful...)
Cheers,
Andy
------------------------------------------------------------------------------
