NA Digest Tuesday, September 14, 1993 Volume 93 : Issue 34

Today's Editor:

Cleve Moler
The MathWorks, Inc.

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to


From: Paul Saylor <>
Date: Fri, 10 Sep 93 17:38:33 -0400
Subject: Milton E. Rose

Milton E. Rose died Sunday, August 22 in Lakewood, Colorado at age 68.
Milt retired as Director in 1985 from the Institute for Computer
Applications in Science and Engineering (ICASE). He had held
positions at the National Science Foundation and the Department of
Energy before his appointment at ICASE in 1977. Milt was known for his
work in partial differential equations, and recognized for his
leadership in the mathematical sciences.


From: J.C.T. Pool <>
Date: Wed, 8 Sep 1993 15:37:12 -0500
Subject: Change of Address for Jim Pool

Change of Address for Jim Pool, effective October 1, 1993

James C. T. Pool, Deputy Director
Caltech Concurrent Supercomputing Facility
California Institute of Technology, MS 158-79
Pasadena, CA 91125

Telephone: 818-395-6953
Fax: 818-584-5917


James C. T. Pool, Head Telephone: 215-895-2668
Mathematics & Computer Science Facsimile: 215-895-4999
Drexel University Internet:
Philadelphia, PA 19104-2884


From: Ilse Ipsen <>
Date: Thu, 9 Sep 93 10:14:03 EDT
Subject: Change of Address for Ilse Ipsen

My new address is:

Ilse Ipsen
Department of Mathematics
North Carolina State University
Box 8205
Raleigh, NC 27695-8205, USA

(919) 515-3796


From: Marino Zennaro <>
Date: Thu, 9 Sep 1993 21:42:54 +0200 (WET-DST)
Subject: Change of Address for Marino Zennaro

I inform all colleagues that very soon I am moving from
the University of L'Aquila to the University of Trieste.
Therefore the mailing address to be used to send me
anything from now on is the following:

Marino Zennaro
Dipartimento di Scienze Matematiche
Universita' di Trieste
I-34100 Trieste, Italy

Since the two Univesities are quite far from each
other, any mail which might be sent to my previous
address in L'Aquila is likely to take a very long
time before I actually get it.

Also my e-mail address in L'Aquila
will be soon disabled.

My e-mail addresses are:
which is equivalent to


From: William Mitchell <>
Date: Fri, 10 Sep 93 14:34:37 EDT
Subject: Change of Address for Bill Mitchell

Effective immediately, my address, etc., is:

William F. Mitchell
Bldg 101 Room A238
Gaithersburg, MD 20899-0001

office phone: (301) 975-3808

Friends can find additional address information from na.whois.

Bill Mitchell


From: John D. McCalpin <>
Date: Wed, 8 Sep 1993 14:12:11 -0400 (EDT)
Subject: Well-posedness of "Almost Hyperbolic" Systems

In a variety of important subject areas, one drops time derivative terms
from a hyperbolic system as a result of scaling arguments. The resulting
systems are not hyperbolic and are rather tricky to analyze (e.g. Oliger
and Sundstrum, SIAM J. Appl. Math, 1978).

Does anyone know of any existence and uniqueness proofs for any
configurations of such systems (e.g. the hydrostatic equations)?

A very similar problem arises in the so-called "long-wave" models used in
nearshore ocean wave studies. These equations are essentially a
bi-directional version of the "Regularized Long-wave" equation, which, in
turn, is similar to the well-known KdV equation. The only proof I know of
in these areas is by Bona and Bryant, which shows that the RLW eq forms a
well-posed initial-boundary-value problem with one initial condition and
one boundary value. Unfortunately, since the RLW eq only has propagation
in one direction, the important issue of wave reflection is not addressed
by this proof.

Finally, by some judicious approximating, the long-wave equations can be
put into a form that might be called hyperbolic. That is, the system can
be written such that, in each equation, all derivatives are along a single
characteristic direction, but two of the characteristics are the same
(their eigenvalues are zero). The system is deficient by one eigenvector,
so it cannot be symmetrized. Has anyone out there worked with existence
and uniqueness proofs for a system with this property? I am especially
interested in the number and type of boundary conditions required, and
have already found a regime in which "open" boundary conditions cannot be

John D. McCalpin
Assistant Professor
College of Marine Studies, U. Del.


From: Roland Freund <>
Date: Tue, 7 Sep 93 22:17 EDT
Subject: QMRPACK is Available from NETLIB

QMRPACK, a software package with FORTRAN-77 implementations of the QMR
method and variants thereof, is now available from NETLIB. The package
is distributed in the form of the compressed tar file "qmrpack.tar.Z" in
the "linalg" section of NETLIB. You can obtain QMRPACK by using xnetlib
or by anonymous ftp. In the latter case, ftp to "" and
then get the file "qmrpack.tar.Z" from the directory "netlib/linalg".

QMR is a Lanczos-based Krylov subspace iteration for solving nonsingular
nonsymmetric systems of linear equations; it also works for certain
singular square systems. The two main features of QMR are (i) its
smooth convergence behavior due to the global quasi-minimal residual
property of its iterates, and (ii) the use of look-ahead techniques
to avoid breakdowns and near-breakdowns in the underlying Lanczos

The package provides two different implementations of QMR, one based
on three-term recurrences, and one based on coupled two-term recurrences.
Since several people find it desirable to have iterative methods
that require only a few lines of code, we have also included versions
of QMR without look-ahead. In particular, there is code for "QMR from BCG",
which generates QMR by simply adding one extra SAXPY to each BCG iteration.
However, in view of their enhanced stability, we recommend the use of
the "true" QMR methods with look-ahead. The package also contains a
no-look-ahead version of TFQMR, which is a transpose-free variant of QMR.
The package comes with two preconditioners: SSOR and a variant of
Youcef Saad's ILUT preconditioner. Finally, the package also includes
code for computing eigenvalues of nonsymmetric matrices, using the
look-ahead Lanczos algorithm.

At the moment, QMRPACK only contains single and double precision code
for real systems. In the near future, we will add codes for complex systems,
special variants of QMR for complex symmetric systems and symmetric indefinite
systems with symmetric indefinite preconditioners, and an implementation
of TFQMR with look-ahead. These additions to QMRPACK will be announced
on na-digest.

Roland Freund and Noel Nachtigal


From: Joe Grcar <>
Date: Thu, 09 Sep 93 19:16:46 -0700
Subject: Difficulties with IEEE Arithmetic, part 2

Here's another reality check for IEEE arithmetic fans. The
point of relating this story is: in my limited circle of R&D
professionals I've encountered two unsolicited complaints
about IEEE arithmetic in as many years. I suspect there are
many scientists and engineers who've said "gee, that's a
really stupid thing to do," but who haven't been heard by
anyone who understands the ramifications of the complaint.

A colleague with a PhD in mechanical engineering just walked
into may office and said:

"I like my new SGI Indigo, but I really don't like those NaN
things it has. I'm rewriting a piece of software and when I
looked at the output it was all full of NaN's. I found out
this means I'm dividing by zero somewhere, but where? I
liked the Vax a lot better because when it divided by zero it
would just stop. That made it easy to find mistakes. I've
spent all day in the debugger looking at NaN's trying to
figure out where they're coming from.

"Have you ever heard of this? Apparently the SGI is built so
it can add these NaN's together, and multiply and divide them
too, just like numbers. But that's a really stupid thing to
do, because they aren't numbers. Is there any way to add
something to not-a-number and get a-number?

(I said I didn't think so.)

"So what good are they? Why would anyone go through all this
trouble just to get output that nobody wants?

(So I said there is some way to turn off the NaN's. By
chance I once asked SGI technical support where to find
documentation on IEEE arithmetic. So I knew "man math"
mentions NaN's, but in fact, it has only happy talk and a
long list of "see also's". One of those, "man sigfpe,"
begins by discussing a struct called sigfpe_template. After
several pages of that, there is something about environmental
variables, but I doubt non-hackers get that far.)

Regards, Joe Grcar


From: L. M. Delves <>
Date: Fri, 10 Sep 93 15:45:33 +0100 (BST)
Subject: F90 Test codes

Institute of Advanced Scientific Computation,
University of Liverpool

IASC is in the final stages of a collaborative project
(in fact, two separate projects) to develop:
1) A Fortran90 compiler, with parallel extensions for MIMD message-passing
2) An HPF Fortran source-source translator: HPF => F90+PVM

These have been under development within two European-funded ESPRIT
projects: Supernode II, and PPPE (Portable Parallel Programming

We are now seeking material to test these on; and expressions of interest
in the results of the projects.

Offers of test codes: email me, or send the codes via email to:

Expressions of interest: to the same address;

or write me at

Mike Delves
University of Liverpool
Liverpool, Meresyside, UK

or voice +44 51 794 4752;
or Fax +44 51 794 4754

All offers will be greatly appreciated.
We solicit also codes accompanied by performance data on other
F90 OR F77 compilers: please supply details of host configuration.

Our compiler is currently hosted on 386/486 under OS/2,
but implementations on Transputer T800/T9000, and on Sparc/Unix, are
following close behind.

Mike Delves


From: Arieh Iserles <>
Date: Wed, 8 Sep 93 11:50:06 +0100 (BST)
Subject: Re: SIAM Meetings

In the last issue of na-net Rob Schreiber reacted briefly to Gene
Golub's recent complaint about the dedication of a session at a SIAM
meeting to a commercial product. The points raised by Gene are of
fundamental importance and their impact goes well beyond the
organization and scheduling of scientific meetings. Hence, I believe
that it is vital to discuss them and, specifically and without
rancour, to disagree with Rob's main premise.

It is true that scientists use commercial products in their research
and that the discussion of such tools is legitimate and, indeed,
occasionally helpful. This, however, is not the end of the story. Many
of us must have come across blatant commercialization of the (excuse
the high-sounding phrase) scientific enterprise:

* Thus, presentation and hard-sell of commercial products under the
guise of a bona fide scientific session. It is fair to say that
MathWorks Inc. (the Matlab people) have behaved, in my experience,
with an enviable integrity. Other vendors have less scruples and I have
witnessed my share of presentations that would not have `disgraced' a
television commercial (or television evangelism, for that matter).

* Thus, talks in scientific conferences that consist, to larger or
lesser degree, in overtly promoting a commercial product in which the
speaker has a pecuniary interest. How am I to believe the speaker that
the new algorithm is, indeed, superior to everything in sight, if the
boundary between the scientific and the commercial is blurred?
The statement that `FFT is faster than matrix multiplication' is
scientific and you can argue its merits on scholarly grounds. The
jingle `Donkin Donuts -- the best coffee in the world', however, is a
purely commercial hyperbole.

* Thus, the relatively new phenomenon of individuals (usually with
former university base) becoming software entrepreneurs. In principle,
this is a welcome development and the field of scientific software will
benefit from the influx of experienced, often brilliant, researchers.
The situation, however, is fraught with danger. Thus, assume that
Professor Bloggs, of great and well-deserved reputation, abandons a
university career to manufacture and market software based on her/his
latest algorithm to digitize differential widgets. Few years later, as
is in the nature of things, Dr Cloggs, a young postdoc at the other
end of the world, comes up with arguably superior (and completely
different) algorithm. However, Bloggs the entrepreneur is also Bloggs
the reputable scientist -- a referee, an editor, member of numerous
committees that decide on careers and budgets. How will Bloggs react
to a development that is likely to wipe out her/his new livelihood?
How will Cloggs trust in Bloggs's honesty?

Commercialization is with us and it cannot be wished away. We should
realize this and, as a community, take steps that allow us to
reconcile integrity with the new state of affairs. After all,
scientists, as human beings, are divided on a multitude of grounds --
national, political, religious... We belong to different scholarly
cultures and different affinity networks. So far, the `system' managed
to get it mostly right. How well, however, will it cope with
commercial pressures? To which extent will we be able to trust
speakers, papers, referee reports, grant-awarding agencies, if we know
the overwhelming commercial interest in the `outcome'? It is not just
a matter of individuals behaving with honesty and integrity -- and I
trust that most of them will -- but also of the recognition at the
receiving end that it is so.

Inasmuch as there are no simple answers, we should discuss this state
of affairs, with a view toward agreeing on a code of conduct that will
allow the numerical community to rip the benefits of commercial
developments without falling pray to their perils. This should not
interfere with the overwhelming majority of individuals trying to make
their livelihood from software or endeavouring to use it in their
research. To the contrary -- maintaining clear and acknowledged
standards should help many to reconcile commercial pressures with
scientific integrity and assist the numerical community as a whole in
trading commercial hype off for valuable information.

In the last decade or so universities in different parts of the world
have been coming under an increasing pressure from their paymasters to
behave as commercial organizations. This is a sad development with
immediate and severe dangers to the fabric of research and teaching.
The more we distinguish clearly and overtly between the commercial and
the scholarly, the better we should be able, as a community, to
recognize (and, hopefully, to resist) unwelcome pressures from
politicians who, in a memorable phrase, know the price of everything
but the value of nothing.

To return to the starting point and Rob's reaction to Gene's complaint
-- there is nothing wrong in discussing a software system at a
scientific meeting as long as there is a firm distinction between the
scholarly and the commercial part and as long as everybody knows
exactly what is what.

-- Arieh Iserles


From: Julio G. Dix <>
Date: Wed, 8 Sep 1993 11:01:34 -0500
Subject: Electronic Journal of Differential Equations (EJDE)


Mathematicians at Southwest Texas State University and
at the University of North Texas have collaborated to
establish a new journal, the ELECTRONIC JOURNAL OF
electronic journal dealing with all aspects of differential
equations. Articles will be submitted as TeX files, sent to
referees electronically, and then disseminated electronically,
free of charge.

Although the time between submission and dissemination
will be greatly reduced, only original research of high
quality will be accepted. Each article will be subject to as
rigid a peer review process as is applied by the finest of
today's printed journals. The EJDE is calling for papers
now. There are no page charges.

The EJDE can be accessed via ftp (login: ftp), gopher, and
telnet (login: ejde) to "" or to
"". Examples illustrating these options are:

1. "telnet", login: "ejde" . (It may be
necessary to set your terminal to emulate a VT100.)
2. "telnet", login: "e-math", password: "e-math",
select "Mathematical Publications", then "Other Mathematical
Publications", and then "Electronic Journal of Differential
3. "ftp", login: "ftp", and "cd pub".
4. Provided that the gopher-client software is loaded on the
reader's computer."gopher".

Readers can transfer the TeX and Postscript files to
their own computers and then read them or print hard copies.

A free subscription to the abstracts of new articles in
the EJDE is available by sending an e-mail message to
"". Suggestions and comments should be sent
to "" or to "".

Identical copies of the EJDE will be originated and
maintained at Southwest Texas State University and at the
University of North Texas. For posterity and for interlibrary
loan, a hard copy exists in the libraries at both institutions.

The Managing Editors of EJDE are Alfonso Castro, Julio
Dix, Gregory Passty, and Ricardo Torrejon. The Editorial
Board consists of

P. Bates (Brigham Young University)
A. Bloch (Ohio State University)
J. Bona (Pennsylvania State University)
K. J. Brown (Heriot-Watt University)
L. Caffarelli (Institute for Advanced Study)
C. Castillo-Chavez (Cornell)
C. Chui (Texas A & M University)
M. Crandall (University of California at Santa Barbara)
E. Di Benedetto (Northwestern University)
G. B. Ermentrout (University of Pittsburgh)
J. Escobar (Indiana University)
L. C. Evans (University of California at Berkeley)
J. Goldstein (Louisiana State University)
C. Groetsch (University of Cincinnati)
I. Herbst (University of Virginia)
C. Kenig (University of Chicago)
R. Kohn (Courant Institute)
A. Lazer (Miami University)
J. Neuberger (University of North Texas)
P. H. Rabinowitz (University of Wisconsin)
R. Shivaji (Mississippi State University)
R. Showalter (University of Texas)
H. Smith (Arizona State University)
P. Souganidis (University of Wisconsin)
N. Walkington (Carnegie-Mellon University)


From: Lothar Reichel <>
Date: Wed, 8 Sep 93 19:56:55 EDT
Subject: New Book, Numerical Linear Algebra


Title: Numerical Linear Algebra
Editors: L. Reichel, A. Ruttan and R.S. Varga
Publisher: W. de Gruyter, Berlin, 1993.
ISBN: 3-11-013784-4

About the Book:

On March 13-14, 1992, a meeting of the Western Pennsylvania and Eastern
Ohio Section of SIAM, with the title Numerical Linear Algebra and Scientific
Computing was held at Kent State University, Kent, Ohio.

In the short period of two days, the roughly 60 participants heard new
research results in Numerical Linear Algebra from acknowledged leaders in the
field. The talks displayed the present activity in this area of Numerical
Analysis and illustrated the wide diversity of the ongoing research.

Some of the topics covered in papers of the proceedings:

* Iterative methods for large linear systems of equations. Papers
present new variants of the QMR, GMRES and SOR methods.

* Preconditioning of linear systems of equations. Papers discuss block
circulant preconditioners with application to image restoration and the
use of the block-ADI iterative method for preconditioning.

* Generalized M-matrices and ultrametric matrices.

* Algorithms for eigenvalue problems. A new more stable variant of
Rutishauser's qd algorithm is presnted, as well as a divide-and-conquer
algorithm for the generalized symmetric tridiagonal eigenvalue problem.

* An algorithm of interest for the computation of the singular value
decomposition of products of matrices.

* Ill-posed problems in image resoration.

* Implementation of iterative methods on sequential and parallel computers.

Contributing authors:

A. Bojanczyk, C.F. Borges, D. Calvetti, M. Eiermann, K.V. Fernando,
R.W. Freund, W.B. Gragg, M. Hanke, S. Ma, R. Nabben, N.M. Nachtigal,
J. Nagy, B.N. Parlett, J. Petersen, R. Plemmons, L. Reichel, Y. Saad,
P. Van Dooren, R.S. Varga

Information on ordering:

For the USA, Canada and Mexico:

Walter de Gruyter,Inc.
200 Saw Mill Road
Hawthorne, NY 10532, U.S.A.
Phone (914) 747-0110
Telex 646677
Fax (914) 747-1326
Price: US $79.95

For other countries:

Walter de Gruyter & Co.
Postfach 303421
10728 Berlin
Phone +49-30-26005-161
Telex 184027
Fax +49-30-26005-222
Price: DM 168,-


From: Gil Strang <gs@BOURBAKI.MIT.EDU>
Date: Thu, 9 Sep 93 11:53:21 EDT
Subject: New Book, Introduction to Linear Algebra

Gil Strang's linear algebra text and MATLAB diskette

The new textbook Introduction to Linear Algebra is published by
Wellesley-Cambridge Press. The good way to obtain a copy is by email to Chapters 1-6 are for the basic course in linear algebra.
The book ends with the SVD and FFT and matrix applications to networks
and computer graphics.

The accompanying MATLAB diskette (PC or Mac) will soon be available from It contains 25 Teaching Codes, each with an M-file of
examples and exercises written by Cleve Moler and Gil Strang. The Toolbox
also comes by anonymous ftp to in file pub/toolkits/Strang
or by email request to

The text is widely adopted already this fall. It starts more gradually
than the earlier book Linear Algebra and Its Applications - and computing
gives students a way to work with matrices.

-- Gil Strang


From: Larry Norris <>
Date: Fri, 10 Sep 93 17:01:56 EDT
Subject: Cornelius Lanczos International Centenary Conference

December 12-17,1993
North Carolina State University
Raleigh, NC, USA

To celebrate the 100th anniversary of the birth of
Cornelius Lanczos (1893-1974) , North Carolina State University
together with other sponsors will hold an international conference
from Sunday 12 December through Friday 17 December 1993. The
conference will reflect the wide interests of Cornelius Lanczos
in computational mathematics, theoretical physics, and astrophysics.


The program of the conference includes approximately twenty-six
invited plenary speakers and twenty-five minisymposia which will
deal with the research topics listed below. The invited plenary
papers and the minisymposium papers will be published in the
proceedings of the conference. Also, the conference will include
sessions of contributed papers in the form of twelve minute talks
and poster sessions. The contributed talks and posters can be on
any topic in the areas of computational mathematics, theoretical
physics, and astrophysics.


The closing date for receiving abstracts for oral contributed and
poster papers for publication in the final program has been extended
to 1 October 1993. There is also a late paper deadline of 1 December,
1993, for both oral and poster papers. We cannot guarantee that
abstracts received after 1 October will be published in the final
program. However, we will publish abstracts of late papers in an
addendum to the final program which will be distributed with the
final program and other registration materials at the conference.


********** Computational Mathematics **********

James W. Cooley Lanczos and the FFT:
A discovery before its time

Jane K. Cullum Lanczos algorithms for large scale nonsymmetric
and symmetric matrix eigenvalue computations

Roland W. Freund The Look-Ahead Lanczos Process for
Nonsymmetric Matrices and Its Applications

Gene H. Golub Uses & Abuses of the Lanczos Algorithm

Anne Greenbaum The Lanczos and Conjugate Gradient
Algorithms in Finite Precision Arithmetic

Martin H. Gutknecht The Lanczos Process and Pade Approximation

Eduardo L. Ortiz The Tau Method of Lanczos and Related
Numerical Methods for Differential Equations

Chris Paige Solving equations and computing eigencomponents
via the Lanczos process

Beresford Parlett Do we fully understand the symmetric
Lanczos algorithm yet?

Robert J. Plemmons Some Matrix Computations in Adaptive Optics

Pal Rozsa On Generalized Band Matrices and Their Inverses

Youcef Saad Theoretical Error Bounds and General Analysis of
a Few Lanczos Algorithms

G. W. Stewart Lanczos and Linear Systems

John Todd Reminiscences of Cornelius Lanczos

********** Theoretical Physics & Astrophysics **********

V. I. Arnold (to be confirmed)

James B. Hartle The Classical Domain in a Quantum Universe

Christopher Isham Quantum Gravity: What Are We Doing?

Jerrold Marsden Some New Perspectives on Variational
Principles in Mechanics

Roger Penrose Relativity, Quantum Theory, and Computation

Tsvi Piran Neutron Star Mergers and Gamma-Ray Bursts,
Possibly the Strongest Explosions in the Universe

John J. Stachel Lanczos's Contributions to General Relativity

Yasushi Takahashi Four Dimensional Vector and the Gauge Transformation

Claudio Teitelboim Black Hole Entropy and Dimensional Continuation

Kip S. Thorne Gravitational Waves: Challenges, Plans and Prospects

Michael S. Turner The Remarkable Success Story of the Big Bang Cosmology

John A. Wheeler

James W. York Black Holes In Thermal Equilibrium


Computational Mathematics:

M. Berry Software for Lanczos-based Algorithms
A. Bjorck Least Squares
Dan Boley Control Applications
D. Calvetti Moments in Numerical Analysis
M.T. Chu & John Lewis Eigenvalue Computations
C. K. Chui Wavelets
J. Cooley Development of the FFT
E. Ortiz Tau Method and Chebyshev Polynomials
R.J. Plemmons The FFT in Signal Processing
H. Van der Vorst Iterative Methods for Linear Systems

Theoretical Physics & Astrophysics:

Paul Anderson Black Hole Evaporation and Thermodynamics
J. Blondin & J. Stone Computational Magnetohydrodynamics in Astrophysics
A. Kashlinsky Galaxy Formation and Large-Scale Structure of the Universe
M. Scholer & D. Winske Numerical Simulations of Collisionless Space Plasmas
C.M. Will Detection of Gravitational Radiation from Astrophysical
J. Isenberg Cauchy Problem of General Relativity
D. Garfinkle Cosmic Censorship
P. Dolan & A.H. Taub Lanczos H-tensor
M. J. Gotay & P. Olver Symplectic Methods in Physics
K.V. Kuchar The Problem of Time in Quantum Gravity
J.J. Halliwell Decoherence and the Foundations of Quantum Mechanics
T. Ratiu & T. Bloch Geometric Mechanics
L. Smolin New Variables and Loop Quantization
P.H. Frampton,
T.W. Kephart & M. Sher Supercollider Physics
C. Carlson
& A. Szczepaniak Open Questions in Particle Theory


A first circular was sent out in late Fall, 1992, through the SIAM,
APS and GRG mailing lists. A second circular giving details about the
program and related information was distributed in early June, 1993.
If you wish to receive a copy of the second circular, or wish to receive
the preliminary program (scheduled for mailing in early October) please
write to:

Cornelius Lanczos International Centenary
Conference, Attn: Sheehan/Heggie,
NCSU/OCE&PD, Box 7401, Raleigh, NC, 27695-7401, USA

or contact through e-mail:


From: Jack Dongarra <>
Date: Sat, 11 Sep 93 09:53:34 -0400
Subject: Scalable High Performance Computing Conference


Knoxville, Tennessee
May 23 - 25, 1994

University of Tennessee
Oak Ridge National Laboratory

Sponsored by
IEEE Computer Society

Jack Dongarra David W. Walker

David Bailey Paul Pierce
William Gropp Sanjay Ranka
Rolf Hempel Gary Sabot
Anthony Hey Robert Schreiber
Charles Koelbel Bernard Tourancheau
Steve Otto Robert van de Geijn
Cherri Pancake Katherine Yelick

Tutorials Published Proceedings

Guy Blelloch, Carnegie Mellon University
Phil Colella, University of California, Berkeley
David Culler, University of California, Berkeley
Monica Lam, Stanford University
Marc Snir, IBM T.J. Watson Research Center

Architectures Load Balancing
Artificial Intelligence Linear Algebra
Compilers Neural Networks
Concurrent Languages Non-numerical Algorithms
Fault Tolerance Operating Systems
Image Processing Programming Environments
Large-scale Applications Scalable Libraries

Extended Abstracts November 1, 1993
Abstracts of posters November 1, 1993
Notified of acceptance January 14, 1994
Camera-ready copy of paper February 14, 1994
Lodging Reservations March 1, 1994
Early Registration March 1, 1994

More detailed questions can be sent to:


From: Karsten M. Decker <>
Date: Sun, 12 Sep 93 15:45:04 +0200
Subject: IFIP Conference on Programming Environments


April 25 - 30, 1994
Monte Verita, Ascona, Switzerland

Massively parallel systems with distributed resources will play a very
important role for the future of high performance computing. One of
the current obstacles of these systems is their difficult programming.
The proposed conference will bring together active researchers who are
working on ways how to help programmers to exploit the performance
potential of massively parallel systems. The working conference will
consist of sessions for full and short papers, interleaved with poster
and demonstration sessions.

The Conference will be held April 25 - 30, 1994 at the Centro Stefano
Franscini, located in the hills above Ascona at Lago Maggiore, in the
southern part of Switzerland. It is organized by the Swiss Scientific
Computing Center CSCS ETH Zurich. The conference is the forthcoming
event of the working group WG 10.3 of the International Federation for
Information Processing (IFIP) on Programming Environments for Parallel
Computing. The conference succeeds the 1992 Edinburgh conference on
Programming Environments for Parallel Computing.


Submission of papers is invited in the following areas:

-- Programming models for parallel distributed computing
-- Computational models for parallel distributed computing
-- Program transformation tools
-- Concepts and tools for the design of parallel distributed algorithms
-- Reusability in parallel distributed programming
-- Concepts and tools for debugging massively parallel
systems (100+ processing nodes)
-- Concepts and tools for performance monitoring of massively
parallel systems (100+ processing nodes)
-- Tools for application development on massively parallel systems
-- Support for computational scientists: what do they really need ?
-- Application libraries (e.g., BLAS, etc.) for parallel distributed
systems: what do they really offer ?
-- Problem solving environments for parallel distributed programming

Authors are invited to submit complete, original, papers reflecting
their current research results. All submitted papers will be refereed
for quality and originality. The program committee reserves the right
to accept a submission as a long, short, or poster presentation paper.

Manuscripts should be double spaced, should include an abstract, and
should be limited to 5000 words (20 double spaced pages); The contact
authors are requested to list e-mail addresses if available. Fax or
electronic submissions will not be considered.

Please submit 5 copies of the complete paper to the following address:

PD Dr. Karsten M. Decker
Via Cantonale
CH-6928 Manno


Deadline for submission: December 1, 1993
Notification of acceptance: February 1, 1994
Final versions: March 1, 1994


Karsten M. Decker
Via Cantonale
CH-6928 Manno
phone: +41 (91) 50 8233
fax: +41 (91) 50 6711

For more information, send email to

== Karsten M. Decker


From: Ray Zahar <>
Date: Mon, 13 Sep 93 03:44:18 EDT
Subject: Symposium in Honor of Walter Gautschi

In recognition of Walter Gautschi's contributions to mathematics
and computer science, and to celebrate his sixty-fifth birthday, a
conference is to be held in his honor:

International Symposium
Special Functions, Approximation,
Numerical Quadrature and Orthogonal Polynomials

Celebrating the Sixty-Fifth Birthday
Walter Gautschi

SPONSOR: Purdue University

DATES: The conference begins at 2:00pm on December 2, and
continues until the evening of December 4, 1993

LOCATION: Purdue Memorial Union
Purdue University
West Lafayette, Indiana
U.S.A. 47907

SPEAKERS: The program is comprised of invited talks on the four main
themes of the conference:

ONE-HOUR SPEAKERS: R.A. Askey, E.W. Cheney, G. Dahlquist, G.H. Golub
W.B. Gragg, J. Korevaar, J.N. Lyness, F.W.J. Olver, R.S. Varga

HALF-HOUR SPEAKERS: A. Bellen, J.C. Butcher, J.R. Cash, P.J. Davis,
C. de Boor, B.D. Flury, D. Gottlieb, M.H. Gutknecht, A. Iserles,
W.B. Jones, H.J. Landau, S. Li, G. Mastroianni, C.A. Micchelli,
G.V. Milovanovic, G. Monegato, M.E. Muldoon, S.E. Notaris,
L. Reichel, W.C. Rheinboldt, T.J. Rivlin, S. Ruscheweyh,
E. Saff, F. Stenger, N.M. Temme, J. Wimp, R. Wong, M. Zhang

PUBLICATION: Proceedings of the conference will be published in
book form. Participants (other than the invited speakers)
are invited to contribute articles to the refereed proceedings.
The final date for submission of papers is December 5, 1993.

ABSTRACTS: All abstracts should be sent by October 15, 1993 to
the e-mail address:
or to:
Professor R.V.M. Zahar
Departement d'IRO
Universite de Montreal
C.P. 6128, Succ. "A"
Montreal, Quebec
Tel: (514) 343-7481

FURTHER INFORMATION: Details on registration, accommodation and travel
as well as updated information on the talks, can be obtained
via anonymous ftp at:
directory: pub/numerik/Gautschi_Symposium
filename: Gautschi_Symposium


From: D. Sloan <>
Date: Thu, 9 Sep 93 17:41:15 +0100 (BST)
Subject: Positions at University of Srathclyde


The University of Srathclyde, Glasgow, invites applications for two new
appointments in the Department of Mathematics--

1. Lectureship/Senior Lectureship in Continuum Mechanics (Ref 84/93)

Candidates should have research experience in continuum mechanics and
preferably in anisotropic fluids. The successful candidate will strengthen
research and teaching in continuum mechanics. An exceptionally well-qualified
candidate may be offered appointment at Senior Lectureship level.
Salary:- up to 29,788 pounds per annum.

2. Lectureship in Nonlinear Analysis (Ref 82/93)

Candidates should have research experience in nonlinear analysis and
preferably in nonlinear dynamical systems. The successful candidate is
expected to straddle current departmental research interests in applied
analysis and numerical analysis. Salary:- 13,601 - 25,107 pounds per annum.

For application form and further particulars (please quote reference number)
contact the Personnel Office, University of Strathclyde, Glasgow, G1 1XQ.
Applications closing date:- 13th October 1993.


From: SIAM <>
Date: Thu, 09 Sep 93 14:39:29 EST
Subject: SIAM Catalogs Available


The SIAM 93-94 Publications Catalog is available upon request.
It includes descriptions of all SIAM books in print.

The SIAM 1994 Periodicals Catalog is also available and includes
SIAM journal descriptions and editorial boards.

Catalog requests should be sent to:

Please specify which catalog(s) you are interested in recieving.

SIAM* 3600 Science Center* Philadelphia, PA* 19104* USA
215-382-9800* FAX 215-386-7999


From: Jack Dongarra <>
Date: Fri, 10 Sep 93 08:21:27 -0400
Subject: Contents: International Journal of Supercomputer Applications

International Journal of Supercomputer Applications
Volume 7.3 Fall 1993 Anna Nagurney, Guest Editor
Special Issue on Computational Ecomonics

Simulating Normal Rectangle Probabilities and Their Derivatives:
The Effects of Vectorization
Vassilis Hajivassiliou

A Globally Convex Agricultural Production System: Parameter Estimation
Agapi Somwaru and Kenneth Hanson

Econometric Model Simulation on Parallel Computers
Manfred Gilli and Giorgio Pauletto

A Numerical Solution Algorithm for Solving Models with Incomplete Markets
ayse Imrohoroglu, Selahattin Imrohoroglu, and Douglas Joines

Forward Looking Behavior and Learning in Stochastic Control
Hans Amman and David Kendrick

Computational Issues in the Statistical Design and Analysis of Experimental
Mahmoud El-Gamal, Richard McKelvey, and Thomas Palfrey


End of NA Digest