NA Digest Sunday, April 29, 1990 Volume 90 : Issue 17

Today's Editor: Cleve Moler

Today's Topics:


From: Paul Saylor <>
Date: 29 Apr 90 18:23 +0200
Subject: Report on Midwest NA Day

Report on Midwest NA Day at the University of Illinois.

Midwest NA Day was held Friday, April 7 1990 on the University of
Illinois, Urbana-Champaign campus. The informal gathering of invited
speakers and conference attendees was organized to pay tribute to
Professor Bill Gear and took special note of his retirement from the
University of Illinois, where he has been head of the Computer Science
Department since 1985.

The conference was originally proposed by Linda Petzold of Lawrence
Livermore National Labs and a former student of Bill Gear's.
The title -- Midwest NA Day -- was suggested by Gene Golub
of Stanford and the conference was organized by Steven Lee with
assistance from Mike Holst and Paul Saylor of the University of
Illinois. Approximately 70 people attended and 10 talks were given.

The speakers were:
Gene Golub (Stanford)
Roland Freund (RIACS)
Germund Dahlquist (Royal Inst. of Technology, Stockholm)
Linda Petzold (LLNL)
Bruce Suter (Air Force Institute of Technology)
Walter Gautschi (Purdue)
Dan Boley (University of Minnesota)
Shikang Li (Purdue)
Greg Ammar (University of Kentucky)
Biswa Datta (Northern Illinois University).

The talks covered topics in orthogonal polynomials, nonsymmetric
Lanczos iterations, ODEs, DAEs, neural networks, Lanczos algorithms
for controllability/observability, discrete least square
approximations, and the numerical solution of control problems.

NA Day was held in the fourth floor Tower Room of the new Beckman
Institute, which gave a panoramic view, such as it is, of the twin
cities of Champaign and Urbana. (The crystal brilliance of a cool,
sunny day partly compensated for the inadequate scenery.) Although the
location was impressive, it was the speakers who made NA Day
exceptional. The purpose of the meeting was a farewell gesture of
friendship for Bill Gear, and in this it succeeded. It may have also
succeeded in promoting closer ties among the NA community in the
midwest (with special thanks to our visitors from outside the midwest.)


From: Andy Froncioni <>
Date: 28 Apr 90 17:27:52 GMT
Subject: Request for Interesting ODE Systems

Does anyone out there have any interesting systems of ODE's
that they'd like solved? If so, please send them to us. We are
developing a tool to solve tough systems and would like to
test the tool.


Andy Froncioni

Andy Froncioni
Rutgers Univ., New Brunswick, N.J.


From: John Holt <jnh@axiom.maths.uq.OZ.AU>
Date: Mon, 23 Apr 90 11:43:43 EST
Subject: Teaching Laboratories


The purpose of this message is to solicit information from colleagues
concerning undergraduate computing laboratories. The University of
Queensland Mathematics Department is planning a significant upgrade to
our laboratory to cater for the large number of students taking courses
with computing content. It is proposed that the laboratory will be used
in courses including n.a., o.r., stats.,calculus, discrete maths, etc.

At present we have 24 graphics terminals to our UNIX system and 16
XT clones which can also emulate terminals to the UNIX system.

It is planned to install around 40 new stations. There is a requirement to
run PC MATLAB for the n.a. people, MINITAB for the statisticians and
various other PC software products. In addition, we have developed
an inhouse computer aided teaching package in C under UNIX which we
wish to run. The hardware solution appears to us to be a network of
PC's (probably AT's) with a 386 file server (or 486), and in addition,
to network the PC's to our PYRAMID 9810 UNIX system. Using terminal
emulation, we can run the UNIX applications as required. This will
also give us access to NAG and other packages on the PYRAMID.

Does anyone have any experience with such dual networking configurations?
Which PC networking software are you happy with?

What sort of security do people use on their computing labs?

We would appreciate any advice that you can offer on the above or any
related issue.

John Holt


From: David Levine <>
Date: Tue, 24 Apr 90 16:20:11 CDT
Subject: Argonne Parallel Programming Class

The Advanced Computing Research Facility (ACRF) at Argonne National Laboratory
is offering introductory courses on parallel computing to familiarize
potential users with the ACRF multiprocessors and parallel programming in
general. The courses are limited to 22 people on a first come first serve

Topics to be covered include:

1) Parallelizing compilers.
2) The Monitor package for portable parallel programming.
3) Programming the Butterfly 2.
4) Programming the AMT DAP.
5) Programming the Connection Machine-2.
6) Introduction to the LAPACK project.

The format of the course is alternating lectures and hands-on work
with the parallel computers in the ACRF. Fortran will be emphasized as the
primary programming language. Knowledge of Fortran and Unix will be assumed.
A portion of the third day is available for each attendee to work on their
own particular project.

The schedule of classes for the remainder of 1990 is:

June 13-15 1990
August 22-24 1990
October 17-19 1990
December 5-7 1990

Parallel computers currently in the ACRF are:

4-processor Ardent Titan
8-processor Alliant FX/8
16-processor Intel iPSC-VX hypercube
20-processor Encore Multimax
24-processor Sequent Balance 21000
32-processor Intel iPSC-1 hypercube
45-processor Butterfly TC2000
1024-processor Active Memory Technology DAP
16384-processor Thinking Machines CM-2

Those interested in the class should contact:

Teri Huml
Mathematics and Computer Science Division
Argonne National Laboratory
Argonne, IL 60439-4844
(312) 972-7163

There will be a $25.00 registration fee per person for universities,
federal laboratories and government organizations and $100.00 for
commercial organizations.


From: Paul Concus <>
Date: Thu, 26 Apr 90 13:42:14 PDT
Subject: Postdoctoral Fellowship at LBL

Applications Invited for
Hans Lewy Postdoctoral Fellowship in Mathematics
at Lawrence Berkeley Laboratory

Applications are invited for the Hans Lewy postdoctoral
fellowship in the Mathematics Department of the Lawrence
Berkeley Laboratory. The fellowship, which was initiated
last year, is intended to enhance scientific research in the
U.S.A. by fostering a young Ph.D. to work in areas related
to the numerical solution of partial differential equations.
Of particular emphasis are those areas in which the LBL
Mathematics Department is active, such as fluid mechanics,
polymer physics, interface methods, iterative methods, and
parallel processing. Concomitant interests in the use of
advanced-architecture computers are encouraged.

Hans Lewy, who died in 1988, spent more than a half-
century as part of the Berkeley mathematics community. Much
of his work was in the area of partial differential equa-
tions, and a portion of it forms the foundation of the
theory of modern difference schemes for solving evolutionary
partial differential equations numerically.

Favorable funding may permit a new fellow to be
appointed as early as this fall. Interested applicants
should send a curriculum vitae and names of three references
to Ms. V. Heatlie, Lawrence Berkeley Laboratory, University
of California, Berkeley, CA 94720. Applications for this
fall should arrive before May 23, 1990. If the anticipated
early funding does not materialize, then applications sub-
mitted before May 23 will be held active, for applicants who
so desire, until selection early next year of a fellow whose
term will begin in the fall of 1992. The fellowship is for
one year, with possibility of renewal for a second year.
Support is provided primarily by special funds from the
Department of Energy Applied Mathematical Sciences Research
Subprogram, for which there is a stipulation that the reci-
pient be a U.S. citizen or permanent resident.


From: Howard Wilson <HWILSON%UA1VM.BITNET@Forsythe.Stanford.EDU>
Date: Thu, 26 Apr 90 18:26:47 CDT
Subject: Offer to Share MATLAB Programs

This note is addressed to people interested in using high
level programming languages in physical applications.

Many people interested in applied mathematics already know
that MATLAB has been gaining wide use as a wonderful tool
for performing matrix computations and complex analysis.
However, MATLAB has probably been less widely recognized as
a powerful programming medium providing a very attractive
alternative to FORTRAN in many engineering and scientific
applications. After having used FORTRAN as my primary
computing tool for many years, it is a real joy to work with
a program language like MATLAB which greatly reduces many of
the pains which are so frustrating in the debugging of
FORTRAN code. I am thinking of the tedium of correcting
mistakes in long argument lists, and making sure the typing
and sizes of data arrays are correct. Since MATLAB creates
intermediate data variables dynamically and treats variables
as complex numbers, whenever appropriate, many of the
traditional 'stupid errors' which waste a lot of time when
we have to work with older generation program languages
simply go away. Consequently, we can get a program working
much sooner and can proceed to the primary task of getting
numerical results from a problem of interest.

Much of my professional effort as a teacher of mechanics is
devoted to instructing engineering students on how to solve
physical problems by use of computers. Any new tool is
welcomed which can accelerate the facility with which
numerical results are obtained. The appeal of not having to
tolerate debugging frustrations which go with along with
programming in widely used languages like Fortran, Basic,
Pascal or C is a welcome change. This is especially true to
someone whose primary priority is to solve physical problems
as opposed to embracing computer programming as a way of
life. Since MATLAB works so well for me and many of my
colleagues, I am eager to see its use grow and to share
instructional applications with other people of similar
interest. For this reason, Chris Bischof and I (with the
kind help of Jack Dongarra) have been able to establish a
MATLAB program collection under the NETLIB software
collection at ORNL. Any people who are not already familiar
with the MATLAB collection might like to drop a note to
Chris at Argonne ( and have your name added
to the MATLAB user group list. Furthermore I would be happy
to send any interested person an Email copy of the following
collection of MATLAB programs consisting of about 3000
source lines . Some of the programs are already included in
NETLIB. Others are simple instructional examples which may
not be appropriate for inclusion in a library. The
collection is certainly not a research contribution, but it
does show the versatility of MATLAB as an efficient
alternative to FORTRAN.

*** HWILSON Library Illustrating MATLAB Applications ***

BARIMAPC Longitudinal vibrations of a bar
BEMIMPAC Transverse vibrations of a beam
BESTERP Bessel interpolation and differentiation
BRACHIST Brachistochrone problem by nonlinear programming
CABLEDYN Nonlinear dynamic response of a cable
CHAOS Routines to illustrate chaos
DESTAB Stability regions for Runge-Kutta methods
DVDTP Interpolation by divided differences
ELIPS Rational function approximation for ellipse mapping
FHRSDE Forced harmonic response in structural dynamics
FOUPLA Fourier series expansion for general functions
GAUSSQ Routines for Gauss integrations
GJEROU Gauss-Jordan reduction with rounding
NELVIB Nelder-Mead function fit to vibration data
PENFORC ODE78 integrator for differential equations
POLHEDR Inertial properties of polyhedra
RECTOR Ractangle torsional analysis by point matching
SLAPERR Solution of Laplace's equation in a rectangle
SPACCURV Interpolation of data points on a space curve
SQUARMAP Conformal mapping of a circle onto a square

Any opportunity to confer with other kindred spirits who like
MATLAB would be welcomed by me and other users at the University
of Alabama.

Howard Wilson
Engineering Mechanics Department
University of Alabama
Box 870278
Tuscaloosa, Alabama 35487-0278
Telephone (205 348-1617)


From: Bengt Aspvall <>
Date: Fri, 27 Apr 90 16:18:28 +0200
Subject: Scandinavian Workshop on Algorithm Theory


The Second Scandinavian Workshop on Algorithm Theory (SWAT 90)
Bergen, Norway, July 11-14, 1990

The program committee has selected 34 papers for presentation. In addition,
there will be invited lectures by Juris Hartmanis, Robert E. Tarjan, and
David S. Johnson.

Springer-Verlag will publish the proceedings.
BIT will devote a special issue to papers from SWAT 90.

To obtain the complete program and registration forms, please contact

Bengt Aspvall,
Department of Informatics, University of Bergen,
Thormohlensgate 55, N-5008 Bergen, Norway
Phone: 47-5-544156
Fax: 47-5-544199


End of NA Digest