Newsgroups: comp.parallel.mpi
From: Robert van de Geijn <rvdg@cs.utexas.edu>
Subject: Parallel Linear Algebra Package (PLAPACK) Release 1.0 available
Organization: The University of Texas at Austin
Date: 3 Nov 1997 15:27:34 GMT
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <63kqh6$9c4$1@news.cs.utexas.edu>

During the last year and a half, we at UT-Austin have developed an MPI
based Parallel Linear Algebra Package (PLAPACK) designed to provide a
user friendly infrastructure for building parallel dense linear
algebra libraries.  We are happy to announce the first public release
of the package.  The Users' Guide

       "Using PLAPACK: Parallel Linear Algebra Package" 

has been available from The MIT Press since last spring.

WHAT IS DIFFERENT: PLAPACK provides three features not currently found
in other publically available parallel dense linear algebra libraries:
1) A matrix distribution that is a step towards one that is driven 
   by the natural distribution of an application, 
2) An application interface for filling and querying matrices and 
   vectors,
3) A programming interface that allows the code to be written in 
   a way that closely resembles the way algorithms are naturally
   explained, using object based (MPI-like) programming.

WHAT TO EXPECT: We have concentrated on providing an infrastructure
for building parallel dense linear algebra libraries.  Thus, currently
only those who need only basic linear algebra operations (like the
level-1, -2, and -3 BLAS) will be immediately set to link to an
application.  A release of a basic linear system solver package can be
expected shortly.  We encourage people to use this easy-to-use
interface to build higher level algorithms, and contribute to the
project by submitting them to the library.

Thus, we provide all aspects required to build and interface with the
library, including data distribution, accessing data, transparent
communication, and an interface that allows algorithms to be expressed
at a  higher level of abstraction.

Performance is good compared to other such libraries.

SO WHY USE PLAPACK?  To speed development of custom parallel linear
algebra routines.  To avoid dealing directly with indices and data
redistribution (i.e.  inter-processor communication).  Besides, it is
fun (quote from a satisfied customer).

WHAT IS DIFFERENT BETWEEN ScaLAPACK AND PLAPACK?  Comparing ScaLAPACK
and PLAPACK is like comparing PVM and MPI.

WHAT PLATFORMS ARE CURRENTLY SUPPORTED?  As of this moment, we provide
prebuilt versions for IBM, Convex, Cray, SGI, and Sun platforms
Only a C interface is currently supported, but then, most people by
now know how to interface C and Fortran.

SPONSORS: This project is sponsored in part by ARPA, DoE, The Intel
Research Council, and NASA.  Also, the proceeds of the royalties for
the Users' Guide will be donated to UT-Austin to help support the
project.

MORE INFORMATION
   Robert van de Geijn, "USING PLAPACK: Parallel Linear Algebra Package,"
                         The MIT Press, 1997
   http://mitpress.mit.edu
   http://www.cs.utexas.edu/users/plapack
   plapack@cs.utexas.edu

Enjoy,
Robert van de Geijn, for the PLAPACK team

======================================================================
Robert A. van de Geijn              Taylor Hall 4.115C
Associate Professor                 (512) 471-9720 (office)
Department of Computer Sciences     (512) 471-8885 (fax)
The University of Texas             rvdg@cs.utexas.edu
Austin, Texas 78712                 http://www.cs.utexas.edu/users/rvdg