NA Digest Sunday, April 16, 1989 Volume 89 : Issue 15

Today's Editor: Cleve Moler

Today's Topics:


From: Mark Kent <>
Date: Thu, 13 Apr 89 20:14:23 PDT
Subject: Using NA-NET Without "SCORE"

Host, a DEC-20, is disappearing this summer.
The NA-NET started on score early in this decade while
score stopped supporting NA-NET functions in late 1988.

It may be the case that some people are still routing mail through
score. This should not be necessary, but just in case your host
*still* cannot mail directly to the following
redirects should work:

Digest submissions:

Questions, comments, problems:

To get the list of people on the NA-NET (>1400 entries!) send mail to
with the message
send nalist from misc

Note that mail addressed to nanet is read, and acted upon, as soon as
possible by the maintainer of NA-NET. However, due to other time
constraints, there may be a substantial time lag before such mail is
read. The time lag is more suitably measured with a calendar than a



From: George Byrne <GDBYRNE%ERENJ.BITNET@Forsythe.Stanford.EDU>
Date: Mon, 10 Apr 89 10:14:50 EDT
Subject: Eigenvales of Band Matrices

Is there good software around for computing the eigenvalues of
a banded, symmetric matrix, whose bands are too big to fit in
the fast memory of a 4 M-word machine (or any smaller sized machine)?
I am particularly interested in a package that would do well on
a vector machine.

George D. Byrne


From: Gene Golub 415/723-3124 <>
Date: Tue, 11 Apr 1989 17:30:00 PDT
Subject: Meeting in Hungary sponsored by Bolyai Mathematical Society

Budapest, Anker Koz 1-3 I.111 on Numerical Methods
H-1061 August 20-25, 1990
Phone:427-741 Miskolc (Hungary)


Following the Conferences in Tihany(1968), Keszthely(1973, 1977) and
Miskolc (1986), the Fifth Conference on Numerical Methods will be
organized by the Janos Bolyai Mathematical Society in Miskolc between
August 20 - 25, 1990. The conference will take place at the Technical
University for Heavy Industry. Miskolc is located 180 km distance from
Budapest to North-East.


The aim of the meeting is to give a proper survey on the recent
results in Numerical Algebra and in Numerical Solution of Differential

The work of the Confrence will rest on plenary lectures, but the
participants will be offered an opportunity to contribute short
communications and to take part at poster sessions as well.


The Organizing Committee consists of Pal Rozsa (Chairman), Katalin
Balla (Secretary), Zoltan Szarka (Local Representative), Anna Lee,
Istvan Farago, Gisbert Stoyan.

Mailing address: Katalin Balla
Janos Bolyai Mathematical Society
H 1368 Budapest
Pf. 240

Second announcement with Registration Form will be sent at the end of
1989 for those who indicate interest by May 1, 1989.

We hope to meet you at the Conference and would appreciate very much
if you could inform your colleagues about our Conference as well. In
case of any questions, don't hesitate to contact us.

Organizing Committee
Pal Rozsa Katalin Balla
Chairman Secretary.


From: Helen <HDD%CORNELLF.BITNET@Forsythe.Stanford.EDU>
Date: Wed, 12 Apr 89 09:10:43 EDT
Subject: 1989 Summer Institute at Cornell

Cornell National Supercomputer Facility

Advanced Summer Institute in Supercomputing

The Center for Theory and Simulation in Science and Engineering at
Cornell University (Cornell Theory Center) is pleased to announce the
1989 Advanced Summer Institute in Supercomputing, sponsored jointly by
the National Science Foundation and the Theory Center. This Institute
will provide a forum for scientists to meet and learn from leading
researchers about the latest advances in computational methods and
algorithmic developments in a number of important disciplines in
engineering and science. In addition, participants will benefit from
hands-on sessions using the parallel supercomputer resources of the
Center's Cornell National Supercomputer Facility (CNSF) and from
opportunities for informal interaction among themselves and with the
invited speakers.

1989 Summer Institute

The Summer Institute runs for two weeks and is held on the Cornell
University campus. The program consists of four distinct components:
lectures by leading experts in computational science; training in use
of the CNSF facilities, hands-on lab sessions for individual appli-
cations; and informal interaction among participants and speakers.

Lectures: Lectures will be given by industrial researchers with
outstanding reputations and experience on the IBM 3090, Cornell faculty
who are experts in computational science, and current CNSF researchers.
To date, the list of committed speakers from IBM includes Enrico
Clementi, Augustin Dubrulle, Jeno Gazdag, Fred Gustavson, William
Henshaw, Steven Laux and Don Weingarten. Additional speakers have been
invited, and we are awaiting final commitments.
A partial list of topics includes:

Computational Physics:
o Algorithms for Lattice Quantum Chromo Dynamics (QCD) Calculations
o Implementation of QCD Algorithms on the GF11 Parallel Computer
o Monte Carlo Methods for Sub-micron Device Simulation

Computational Fluid Dynamics:
o On the Smallest Scale of Turbulence
o Solution of Navier-Stokes Equations on Composite Overlapping Meshes

Computational Mathematics:
o Concurrent Computing by Sequential Staging of Tasks
o The Key Algorithms in the Parallel ESSL Library
o Future Directions in Sparse Matrix Computations
o Unite-and-Conquer Algorithms for the Solution of Systems
of Equations

Computational Chemistry:
o Molecular Dynamics and Monte Carlo Programs for Proteins and
Nucleic Acids in Solution
o Chemical Application in Fluid Dynamics
o Micro-dynamics

The theme of computational methods will provide the common thread
running through all of these lectures.

Training: Training in the use of the CNSF will cover the VM/CMS
architecture and operating system, concepts and techniques in
parallel processing and vectorization, methods of converting serial
code to parallel, and tools for analyzing code performance.

Lab Sessions: Lab sessions are planned to allow participants to
apply the concepts discussed in lectures to their own projects.
Participants will each be allocated 10 service units (20 CPU hours
on the IBM 3090) to be used during the workshop and later in continuing
work at their home sites. Lecture materials will be given to each
participant, and CNSF staff will be available for consultation
throughout the program.

Informal Interaction: Ample time is planned to allow for informal
interaction among the participants and invited speakers throughout
the program. Speakers will also be invited to attend lab sessions
to learn about participant codes and to provide guidance.

Institute Facilities

The CNSF configuration consists of two IBM 3090-600E supercomputers,
each with six vector facilities, and a peak performance rate of
696 megaflops. (The IBM system has 256 MBytes of memory, 512 MBytes
of expanded storage, and 115 GBytes of disk storage. Each application
may use up to 1 GByte of virtual memory.) The current operating system
is VM/XA SP (CMS) and the primary language is FORTRAN. The CNSF system
provides unique production parallel processing capabilities. An appli-
cation can simultaneously run on six processors using parallelism at
both the loop and task levels while taking full advantage of the large
virtual memory and vector facilities of the 3090.

The Cornell Theory Center

The Center for Theory and Simulation in Science and Engineering -the
Cornell Theory Center - was established in 1984. One of five National
Science Foundation-sponsored National Advanced Scientific Centers, the
Theory Center was created to advance science and technology by making
the most highly advanced computers more accessible to researchers. As
part of this effort, the Theory Center is committed to educating
students, scientists and engineers in the use of supercomputers.

The facility supports about 1,800 users working on more than 460
research projects in the fields of quantum chemistry, theoretical and
applied physics, ecological studies, econometric modeling, biomedicine,
and the social sciences, as well as all areas of engineering.

Application Requirements

Enrollment is limited to 30 participants. Selected applicants will be
graduate students, post-doctoral researchers or university faculty
members who are actively engaged in projects which rely heavily on the
efficiency of computational methods and algorithms as well as vector,
parallel, and memory capabilities.

Participants should bring working versions of their codes which include
only the most critical algorithms in their applications. These working
codes should be written in ANSI FORTRAN 77, and contain minimal machine
dependencies. CNSF staff will provide assistance in migrating code and
data to the IBM 3090 prior to the start of the Institute.

Provisions and Accommodations

Under the NSF grant for the Institute, participants will be reimbursed
for travel expenses; housing and meals will be provided. Details will
be supplied upon acceptance.

Further Information

For further information on the Summer Institute, contact: Robert Feldman
Education and Training, Cornell National Supercomputer Facility, Campus
Road and Central Avenue, Ithaca, NY 14853-8301 607/255-3985.



Supercomputer Summer Institute - July 24-August 4, 1989



Mailing Address:


Electronic Mailing Address:

Institute of Affiliation:

Field of Research:

Principal Investigator:

RETURN THIS FORM along with:

o A letter of interest explaining how the 1989 Institute would
benefit the applicant's research, including a description of
work to be accomplished during the Institute.
o Two letters of recommendation, including one from advisor or
department chair.
o An abstract of current or planned research.
o A brief description of educational background, including
FORTRAN experience.
o Statement of suitability of project for large memory, vector
or parallel processing.

RETURN TO: Donna Smith, Conference Coordinator
Cornell National Supercomputer Facility
Summer Institute
Campus Road and Central Avenue
Ithaca, NY 14853-8301



From: Hans Stetter <E115N06%AWITUW01.BITNET@Forsythe.Stanford.EDU>
Date: 14 APR 89 10:49:18
Subject: Conference Fee at London ODE Conference

Ref: Conference Fee at London ODE Meeting

A week ago, I had voiced my anger about the exorbitant conference
fee of 185 British pounds = 320 US $ = 600 DM in the NA-Digest (issue 14).
This brought me a pile of responses: All (except one) were fully in agree-
ment with my position; many indicated that they may or will not attend
under these circumstances. There were some very well-known numerical
analysts amongst those responding.

Also, two members of the Organisation Committee (= program committee)
told me that they have been fighting with IMA about these excessive fees
for some time but without success, and that they would appreciate support.

May I therefore suggest the following: Everybody who has considered
or planned to go to London and who agrees that these fees are completely
out of proposition for a scientific conference held at a university,
should write a letter to the Institute of Mathematics and its Applications.
Just to stay away or to complain to colleagues will have no effect; IMA
must be made aware of the situation. You may also quote registration fees
of comparable conferences to point out the discrepancy. E.g., the last
Computational ODE Conference at Toronto (1988) which was fully comparable
in size and scope with the London Meeting, had a registration fee below
50 US $, viz. 50 CAN $.

(Besides, the information received with the application form is
quite lacunary. E.g. what quality of accomodation the residential fee will
buy is not even indicated. Thus, the organisation of the conference is not
even done in a professional manner.)

Your action should come very soon in order to be effective, it will
be a service to the whole Numerical Analysis community.

Hans J. Stetter


From: Gene Golub <>
Date: Fri, 14 Apr 1989 15:56:37 PDT

The SIAM Journal on Matrix Analysis and Applications (SIMAX) is approaching
the middle of its second year. I'd like to update you on our progress.

SIMAX is still a very new journal and the style of the journal is still
developing. We continue to focus on the matrix research that has to do with
applications to areas such as signal processing, systems and control theory,
statistics, Markov chains, mathematical programming, economic and biological
modeling. We encourage you to submit your articles to us, especially in those
areas of unusual application. In general, our editorial board members have
been successful in receiving prompt reviews of journal submissions. We are
taking steps to remedy the delicate problem of delayed reviews. Our waiting
time from acceptance to publication is six to nine months and it is our
intention to maintain this processing time.

We want you to know that every issue will contain articles of interest to each
of our subscribers. Thus, as a matter of policy, we do not plan to devote
more than a partial issue to any special topic.

You may now submit your accepted manuscripts in Tex, LATex, and AMS-Tex. SIAM
journal macros are available through Net-lib and directly from SIAM via e-mail
and floppy disk.

Although authors may need to spend some time making editorial changes, there
are advantages in submitting your manuscripts to us in Tex. First, the author
retains greater control over the final appearance of his manuscript. Second,
the introduction of errors from a second keyboarding is avoided. Third, one
round of proofreading is eliminated. Fourth, authors who typeset their own
articles receive 100 gratis reprints.

Circulation for SIMAX has increased nicely during the past year--we now have
over 1100 subscriptions and we expect that number to continue to increase at a
reasonable rate for the next several years.

SIMAX is an affordable journal. The current subscription rate for SIAM
individual members is $36 ($39 for individual members outside the United
States, Canada, and Mexico). The institutional rate is $95 ($118 for
institutions outside the United States, Canada, and Mexico). You can have the
issues of your personal subscription to SIMAX delivered to your home or
office. Institutions may order several subscriptions delivered to addresses
they designate. Our academic members receive a subscription to SIMAX as a
part of their membership. Of course, we encourage you to request your local
library to subscribe to SIMAX.

Remember, for those of you who are SIAM members, SIMAX is your journal. For
those of you who are not members, please consider joining SIAM and subscribe
to SIMAX at the member rate.

We will always welcome your help, advice, and submissions.


Gene Golub
Managing Editor
SIAM Journal on Matrix Analysis and Applications


Please submit your manuscripts to: SIAM Publications,
P.O. Box 7541, Philadelphia, PA 19101.

Requests for journal macros should be sent by e-mail to Corey Gray
at SIAM@WHARTON.UPENN.EDU or by mail to Corey Gray at SIAM, 117
South 17th Street, Philadelphia, PA 19103-5052.

Requests for membership information should be sent to the
e-mail or 17th St. address, attention: Customer Services.


End of NA Digest