NA Digest Sunday, June 3, 1990 Volume 90 : Issue 22

Today's Editor: Cleve Moler

Today's Topics:


From: Paul Van Dooren <>
Date: Wed, 30 May 90 15:09:59 +0200
Subject: New Address for Philips Research Laboratory

Philips Research Laboratory Brussels is moving next week to a new address :

Philips Research Laboratory Belgium
Av. Albert Einstein, 4
B-1348 Louvain-la-Neuve

Tel : +32 10 470611
Fax : +32 10 470699

Some Privalte numbers :

Courtois : +32 10 470669
Genin : +32 10 470659
Van Dooren +32 10 470658

E-mail remains the same, but the service will be interupted for about a week
starting from June 11th. We are effectively moving on June 11.

In case of problems, the FAX number should be operating already now.

Paul Van Dooren


From: Karsten Decker <>
Date: 30 May 90 11:05 +0200
Subject: School on Parallel Architectures and Applications


European Physical Society
Computational Physics Group

3rd Graduate Summer Course on Computational Physics
Parallel Architectures and Applications
to be held at
Cret-Berard Puidoux Switzerland
September 3 - 7, 1990

This course is sponsored by the Computational Physics Group of the European
Physical Society, and is the third one of a series which takes place every
second year. The first one in 1986 was on 'Finite Element Methods in Physics',
the second one in 1988 on 'Numerical Methods for Parallel Vector Computers'.


R.W. Hockney, Reading Parallel Architectures
G.C. Fox, CALTECH Hypercube Architectures and Applications
K.C Bowler, Edinburgh Transputer Machines and Applications
J. Richardson, TM Connection Machine and Applications
I. Guyon, AT&T Bell Labs Neural Networks and Applications
R. Gruber, EPFL Software Development Strategies for Parallel
Architectures with Distributed Memory


Parallel architectures

The course will present an overview on parallel architectures in general
with special focus on parallel machines with distributed memory.
Strong and weak points of the different types of architectures are critically
discussed and compared. Software development environments and tools for
parallel architectures are presented and the importance of these tools for
the strongly application oriented user is explained. Models describing
the performance of parallel architectures with shared and distributed
memory architectures are discussed.

Software development strategies for parallel architectures

After presenting an introduction to numerical modelization in general,
the course will discuss the different phases, starting from real experiment,
switching to physical and numerical modeling. Special focus is on the
adaptation of numerical approaches to modern parallel architectures.

Neural networks

The course will describe the current state-of-the-art in neural computing.
It will not assume prior expertise in neural networks. The six lessons
presented will cover machine learning and biological computation, learning
rules and implementations for adaptive artificial neurons, neural network
systems, applications like pattern recognition and signal processing, and
a review of existing software design tools, hardware implementations, neural
network journals, books, and conferences.


A major fraction of the course is dedicated to the presentation of the
successful implementation of a broad spectrum of applications drawn from
physics, engineering, computer science and other areas. The usefulness of
the distinct types of parallel architectures for a collection of very
CPU-intensive applications is compared, and, depending on the type of
architecture, concrete techniques are presented, by means of which these
applications can be mapped to the hardware in an optimal fashion.


During the five days of the course, there will be six hours of lectures
each day, supplemented by informal discussion sessions during the day and
in the evening. A detailed program and abstracts of the lectures will be
issued to those attending. The full collection of the written versions of
the lectures will become available at the time of the summer course. The
proceedings will be published in a special issue of Computer Physics Reports.

Deadline for application: End of July 1990

Address for application: Karsten M. Decker
University of Berne
Laenggassstr. 51
CH-3012 Berne


From: Robert Meyer <>
Date: Wed, 30 May 90 09:52:00 -0500
Subject: Final program for SPO2, Madison, July 23-25, 1990


23 - 25 July 1990

Center for Parallel Optimization
Computer Sciences Department
University of Wisconsin
Madison, Wisconsin 53706

A 3-day symposium of invited presentations on
state-of-the-art algorithms and theory for
the parallel solution of optimization and related problems will be held
at University of Wisconsin at Madison with support from the AFOSR
and in cooperation with SIAM. (The SIAM National Meeting will be
taking place in Chicago the preceding week.)
Emphasis will be on algorithms implementable on parallel and vector
architectures. Refereed proceedings of the Symposium are planned as
a special issue of the new SIAM Journal on Optimization.
The final list of speakers and talks is given below.

Although the symposium will be comprised of invited talks as
indicated above, registration (early registration by May 30: $50; late
registration $65) is open to all persons wishing to attend.
A registration form and information on lodging is deposited in
netlib and may obtained via email to netlib (mail
with the request:
send SPO from meetings

For information beyond that in netlib, contact the SPO2 Secretary, Laura Cuccia,
or one of the organizers, O. L. Mangasarian, R. R. Meyer at the above address.
Secretary: (608)262-0017, email:, FAX (608)262-9777.

SPO2 Speakers and Talks:

Richard S. Barr & William Stripling Parallel approaches to fixed charge
Southern Methodist University & network problems
American Airlines

Dimitri P. Bertsekas Parallel forward path search: A new
MIT dual coordinate ascent algorithm for
shortest paths

Donald E. Brown & C.L. Huntley Parallel genetic algorithms as control
University of Virginia structures for local optimizers

Renato De Leone & Robert R. Meyer Serial and parallel solution of
University of Wisconsin large-scale multicommodity networks

John E. Dennis Jr. & Virginia Torczon Direct search methods for parallel
Rice University machines

L. C. W. Dixon On parallel truncated Newton and
Hatfield Polytechnic (England) and variable metric algorithms

Michael C. Ferris & Olvi L. Mangasarian Parallel distribution of convex
University of Wisconsin programming constraints

John J. Grefenstette Parallel search with genetic algorithms
Naval Research Laboratory

Heinz Muhlenbein Parallel genetic algorithms and
Carnegie-Mellon Univesity combinatorial optimization

Stephen G. Nash & Ariela Sofer A practical truncated-Newton method
George Mason University for parallel optimization

James M. Ortega Preconditioned conjugate gradient
University of Virginia methods for elliptic boundary value

Yu. E. Nestorov & A. S. Nemirovsky Parallel computations and interior-point
USSR Academy of Sciences (Moscow) methods for structured nonlinear convex

J. Ben Rosen, R.S. Maier & J. Glick Parallel solution of large-scale,
University of Minnesota block-angular concave programs

Russell A. Rushmeier & A Cooperating multiple search approach
George L. Nemhauser for parallel integer programming
Rice University & Georgia Tech

Ahmed Sameh & Randall Bramley Projection methods for solving sparse
University of Illinois nonsymmetric linear systems on

Paul Tseng On the rate of convergence of partially
MIT asynchronous gradient algorithms

Dirk Van Gucht Parallel genetic algorithms applied to
Indiana University the traveling salesman problem

Layne Watson & Alexander P. Morgan Serial and parallel global
Virginia Polytechnic Institute optimization of polynomial programs
via homotopy algorithms

Stephen J. Wright Partitioned dynamic programming
Argonne National Laboratory

Stavros A. Zenios Massively parallel network
University of Pennsylvania optimization with a financial


From: Israel Nelken <>
Date: Thu, 31 May 1990 15:20:45 -0400
Subject: The Eighth Parallel Circus


Eighth Parallel Circus
Location: Toronto, Canada
Dates: October 26-27, 1990

Continuing the tradition that began at Yale in 1986, the Department of
Computer Science at the University of Toronto will be hosting the Eighth
Parallel Circus in Toronto on Friday and Saturday, October 26-27, 1990.

The Parallel Circus is an informal conference which emphasizes parallel
algorithms for scientific computing.

This is the first time the Circus will be held outside of the USA, and we
hope to have many attendees from Canada, the USA and other countries.
The circus is unique in that it has one and only one focus, scientific
computing on parallel machines. Another feature is that there has been
lots of informal discussion and a very healthy mix of industrial and
academic participants.


*** Air Transportation ***

We have arranged special conference air fares with Air Canada. To get
the special rate, call them toll free (800) 361-7585 from either Canada
or the USA. Mention the code number 90-972 and they will give you:

1) The best available fare.
2) a. 15 % off economy fares for flights within Canada.
b. 25 % off coach for flights from the USA.
c. 35 % off coach for USA flights if available.

For (2a) and (2c) there is a requirement of 7 days advance booking. A
minimal stay of 2 nights but less than 15 days is required.

*** Accommodations ***

We have made arrangements for a block of discount rooms at the Delta-
Chelsea Inn, Toronto. It is just a short walk from the University Campus.
The price is $86 CAD per night for a single or $96 CAD double
room. To get the discount rate you should use the code letters "GIMAL"
and please reserve your room by Sept. 26th., 1990.

Delta-Chelsea Inn, Toronto
73 Gerrard St. West

Phone numbers: (800) 268-9070
(416) 595-1975

Toronto also has several less expensive but comfortable Bed & Breakfast
outfits. Their prices are $45-55 CAD for a single and $65-75 CAD for
a double. Call Susan Oppenheim at (416) 598-4562 and she'll be glad to
book you a room. Mention that you need a place close by to the University
of Toronto campus.

Note: All prices are quoted in Canadian Dollars (CAD). Currently $1 CAD
is about $0.85 US.

*** Chinese Banquet ***

Continuing the tradition, we will have a Chinese Banquet on Friday
evening. Please indicate if you would like to come. We estimate the
price to be about $25 CAD.

*** Presentations ***

Presentations will be about 30 minutes long -- 25 minutes plus
5 minutes for questions and discussion. Their actual lengths
will depend on the number of participants.

The circus will begin on Friday morning. Although there is no
prescribed program, we will end by early Saturday afternoon.
Participants who give a talk and leave are generally regarded
as anti-social so you should plan to attend all of the talks.

*** Visas to Canada ***

Visas to Canada may be required for non-US citizens. Please check with a
Canadian Consulate.

Organizers: Gene Golub, Izzy Nelken, Ken Jackson and Christina Christara.

For further information, including an electronic registration form,
please contact:

Izzy Nelken
Department of Computer Science
University of Toronto
Toronto, Canada M5S 1A4
Phone: (416) 978-5899
Fax: (416) 978-4765


From: Joel Saltz <>
Date: Thu, 31 May 90 18:21:41 EDT
Subject: Unstructured Computation on Multiprocessors

Unstructured Scientific Computation on
Scalable Multiprocessors

October 28-30, 1990
Ramada Inn, Nags Head, NC
Sponsored by ICASE

Unstructured and dynamically varying algorithms play an
increasingly critical role in scientific computation. Thus,
mapping such algorithms onto scalable multiprocessors is a
pressing practical concern. This workshop will address issues
raised when implementing such algorithms on architectures, such
as the present generation of hypercube machines, which can be
scaled to the teraflop performance range. Problems of greatest
interest include:

PDE solvers using unstructured and/or adaptive meshes

Algorithms for problems in sparse linear algebra

Adaptive fast multipole methods

Particle-in-cell and vortex-chasing methods

The focus here is on problems in which the information
necessary for effective mapping and load balancing is not
available until run time. Invited speakers will address aspects
of these problems, including pacing issues in algorithm design,
problem partitioning and load balancing, software tools, and
algorithm performance characteristics:

Geoffrey Fox -- Irregular Problems on Concurrent

Manuel Salas -- Unstructured Adaptive Methods for CFD

Harold Trease -- Free Lagrangian Calculations on SIMD

David Nicol -- Static and Dynamic Load Balancing

Roy Williams -- Irregular Meshes and Distributed

David Keyes -- Adaptive Domain Decomposition

Contributed papers are also solicited for this workshop.
Papers, selected on the basis of originality and relevance, will
be allocated 30 minute sessions; to be considered, submit a 1000
word extended abstract to the address below, by September 1,
1990. Authors will be notified by September 15, 1990, and full
papers for inclusion in the proceedings must be in our hands by
January 1, 1991.

Ms. Emily Todd, ICASE, MS 132C, NASA Langley Research
Center, Hampton, VA 23665.

Phone: (804) 864-2174

Organizing Committee: P. Mehrotra, J. Saltz, J. Scroggs, J. Van
Rosendale, R. Voigt


End of NA Digest