NA Digest Monday, February 11, 1991 Volume 91 : Issue 6

Today's Editor: Cleve Moler

Today's Topics:


From: Michael Heath <mth@indigo.EPM.ORNL.GOV>
Date: Fri, 8 Feb 91 17:01:37 -0500
Subject: SIAG/LA Prize

There has been some confusion on the deadline for nominations for the
SIAG/LA Prize. My earlier announcement on NAnet stated that the
deadline was January 31, 1991. However, the most recent issue of SIAM
News gives the deadline as "April 1991" (note that no specific day is
given). Thus, in fairness to any potential nominators who may have
seen only the SIAM News announcement, nominations for the Prize will be
accepted through April 30, 1991. For details on the SIAG/LA prize, see
my earlier announcement on NAnet or the write-up on page 8 of the
January 1991 issue of SIAM News.

-Mike Heath,


From: Srinivasan Krishnan <>
Date: 11 Feb 91 04:23:42 GMT
Subject: Programs Which Use LINPACK

I am writing a program that requires for testing -
programs in FORTRAN that use the library LINPACK. If you happen
to have some that you would be willing to mail to me, I would
really appreciate it.

The type of programs that I am most interested
in would use only the LINPACK routines, all other code has been
written by you. ( In other words, if you also used libraries meant
for other purposes, e.g sparse solvers, it would complicate my
testing. )

Thanks very much,

Duke University Computer Science Dept.; Durham, N.C.

ps: please mail to


From: P. J. Lohr <crdgw1!wsqtb8!>
Date: 11 Feb 91 21:10:04 GMT
Subject: Optimization Servers for Workstations

Can anyone help me figure out what my options are in the way of optimization
servers/packages/libraries for UNIX workstations. Both commercial and
public domain packages are fair game.

Thanks in advance,


From: Ramini <>
Date: Mon, 4 Feb 91 00:22:33 EST
Subject: Three Dimensional Grid Generation

Before I describe my problem, let me provide some background information.

Our group is working on 3-dimensional boundary element simulations of
potential flow problems. These problems are governed by Laplace's equation,
and are suitable for boundary element treatment. The problem is dynamic
and involves free/moving boundaries. The surfaces of the boundaries of
the problem are discretized using equal sized triangular elements, and
the problem is discretized assuming a local linear basis on these elements.

And now my problem:
As the problem is dynamic, elements change size, and as the
simulation proceeds the ratio of the sizes of elements can become
large. I would like to regrid (re-triangulate) the surfaces so that at
each step we have equal sized elements. Any regridding must (approximately)

-- conserve volume
-- conserve surface area
-- conserve curvature

In the two-dimensional case we have had success with spline
interpolation. However in the case of surfaces (in 3-space) the problem
seems hard, and I would appreciate any references/guidelines/codes.


From: Frank Luk <luk@jacobi.EE.CORNELL.EDU>
Date: Wed, 6 Feb 91 09:15:04 EST
Subject: SPIE Conference on Signal Processing

SPIE Conference on Signal Processing

An SPIE Conference on Advanced Signal Processing
Algorithms, Architectures and Implementations
will be held in San Diego on July 24-26.
This conference will be dedicated to the memory
of Jeffrey Speiser. The keynote paper will
be presented by Dr Philip Hargrave of STC Technology
of the United Kingdom; his title will be
"Systolic adaptive beamforming - from theory to practice."
There will be five half-day sessions, on Toeplitz matrices,
on Time-frequency Distribution and Nonstationary Signals,
on Bit Level Systolic Arrays and Computer Arithmetic,
on Implementations and Signal Estimation,
and on Array Processing and Beamforming.
Send me e-mail if you are interested in getting
an e-copy of the full program of 43 papers.
A Conference Proceedings will appear in November.


From: Soili Leskinen <LESKINEN@JYLK.JYU.FI>
Date: Wed, 6 Feb 91 09:21 +2
Subject: Summer Semester in Finland

in Physics, Mathematics and Chemistry
August 1 - 31, 1991 Jyv{skyl{, Finland

The Faculty of Mathematics and Natural Sciences of the University of Jyv{skyl{
organizes an International Summer Semester in August 1991 for the students
who are finishing their undergraduate studies or who are beginning their
graduate studies.
The semester offers ten courses in Physics, Mathematics and Chemistry.
The students may choose any courses according to their personal interest.
It is recommended that everybody take at least two courses. Credit will be
given in each course to those who pass the examinations successfully.

There is no participation fee for the School. The participating student
will be accommodated free of charge in modern student apartments consisting
of a bedroom and a shared kitchen and bathroom.
The University of Jyv{skyl{ will also grant a scholarship of 350 FIM
(approx 100 US $) to each foreign student. The students have to pay their
travel costs and living expenses.

The University of Jyv{skyl{ is located in Central Finland, which is known
for its beautiful lakes, ridges, fields and forests. The economy of Central
Finland is based on modern pulp and paper industry. The city of Jyv{skyl{,
having 65000 inhabitants, has been the site of the World Cup Rally, the Rally
of the Thousand Lakes, every August since 1950. Jyv{skyl{ and its surroundings
offer numerous free-time activities including tennis, golf, track and field,
hiking, fishing, canoeing and lake cruises.

To register for the International Summer Semester ask the secretary to send the
registration form to you and mail it to the organizers by the end of March 1991.

Deadline for the registration is March 31, 1991.

Summer Semester Address: Telephone: 358-41-602203
M.Sc. Soili Leskinen (Secretary) Fax: 358-41-602201
Faculty of Mathematics and Bitnet: MLTK @ FINJYU
Natural Sciences
University of Jyv{skyl{
P-O. Box 35
SF-40351 Jyv{skyl{



MA 1. Functional analysis and partial differential equations
MA 2. Sobolev spaces
MA 3. Numerical analysis
MA 4. Numerical methods for differential and partial differential equations
MA 5. Nonlinear potential theory


PH 1. Introduction to solid state physics
PH 2. Accelerator physics
PH 3. Surface Physics
PH 4. Magnetism


CH 1. Atomic and Molecular Spectroscopy
CH 2. Molecular Dynamics Simulations


From: George Corliss <>
Date: Thu, 7 Feb 91 16:06:03 CST
Subject: Workshop on Automatic Differentiation

January 6 - 8, 1991, Breckenridge, Colorado
Andreas Griewank (
George Corliss (

Automatic differentiation is a technique for the fast,
efficient, accurate generation of ordinary and partial derivatives.
Prof. Louis Rall characterised the technique as a computer
arithmetic in which operations are performed on object which are
gradients, Hessians, Taylor series, or other similar objects.
As explained by Prof. Masao Iri and Prof. George Cybenko, the
top down or reverse mode of automatic differentiation was already
utilized in the early seventies by Seppo Linnainmaa (1970) for
the estimation of rounding errors, and by Paul Werbos for training
neural networks by backpropagation. At the same time, the forward
mode had already been implemented, e.g., in the computer language
PROSE, co-authored by Joseph Thames, who presented his integrated
applications package at this workshop. Several author speakers
presented automatic differentiation implementations, including
precompilers for Fortran programs, integrated symbolic/numerical
environments for PC's, and implementations by overloading in C++,
Ada, and other advanced languages. There was a very strong
interest in practical implementations and applications.
The presentations of Prof. Bruce Char and Dr. Victor Goldman
were particularly valuable in discussing the relationship between
fully symbolic computer algebra and the more numerical technique,
which formed the primary focus of this workshop. Both acknowledged
that automatic differentiation is significantly more efficient in
certain situations and advocated a merging of the two approaches in
order to obtain the best of both worlds in terms of flexibility,
convenience, and efficiency. There were many excellent talks on
large scale applications, including weather modeling, petrolium
reservoir modeling, mechanical systems systems simulation,
oceanography, beam tracing in optics, and orbit analysis. Some of
the speakers delineated very clearly the remaining deficiencies of
currently available automatic differentiation techniques in
comparison to handcoded derivative evaluation programs. On the
basis of the ensuing discussions, it can be expected that the
software developers present will accept the challenge of closing
this gap in efficiency without sacrificing user convenience.
Several speakers pointed beyond the traditional core concern of
evaluating point derivatives of first and second order from
user-defined programs. They proposed the exploitation of automatic
differentiation techniques for the solution of various computational
task, e.g., concurrent scheduling on advanced architectures, the
numerical solution of differential equations with guaranteed bounds,
the efficient calculation of Newton steps, and the parallel solution
of unconstrained optimization problems. Prof. Y.F. Chang pointed
to the importance of extracting physically significant analytic
information such as the locations and orders of singularities from
Taylor series. It became clear, that there remain many theoretical
and practical challenges, for example regarding the differentiation
of multivariate implicit functions arising in differential algebraic
equations or efficient utilization of parallelism.
Proceedings of the workshop are being published by SIAM.


From: Martin Berzins <>
Date: Wed, 6 Feb 91 08:12:43 GMT
Subject: Position at University of Leeds, U.K.

NAG Software Engineer for
Development of Mathematical Software

The post of NAG Software Engineer in the
School of Computer Studies at the University of Leeds, U.K. is available
for a fixed period of three years.
The post which is jointly funded by NAG Ltd. and the School
is to assist in the writing of mathematical software for time-dependent
partial differential equations on high performance computers, that can be
exploited through the NAG Library.
Applicants should have a good honours degree in a scientific or
engineering discipline, have a practical interest in mathematical software
and have experience of C and/or Fortran programming.
Salary up to UK #18,165 according to age, qualifications and experience.
Informal enquiries may be made to Dr. Martin Berzins (email na.berzins or

Thank-you, Martin Berzins.


From: Arden Ruttan <ruttan@sun ... >
Date: Sun, 10 Feb 91 15:10:57 EST
Subject: Position at Kent State University

Department of Mathematical Sciences
Senior Position in Applied Mathematics/Scientific Computation

Applications are invited for a faculty position at the associate or full
professor level beginning Fall Semester 1991 (or earlier). The ideal
candidate would have a strong training in classical/modern applied mathematics
and some experience with large-scale scientific computation. He or she would
be expected to have a solid record of research, publication, and external
funding, as well as a commitment to quality teaching. The appointed faculty
member would be expected to enhance the Department's outreach and
interdisciplinary research efforts, supervise graduate students, and
contribute to curricular planning and development. A competitive salary is

The Department of Mathematical Sciences at Kent State University comprises
pure and applied mathematics, statistics, computer science, and the Institute
for Computational Mathematics. This new position is intended to complement
existing strengths in applied analysis (especially numerical analysis and
approximation theory) and computer science (especially symbolic computation,
expert systems, and parallel computing).

The infrastructure of the Department is very good: the equipment inventory
includes a significant workstation network plus Wavetracer, Encore, Sequent,
and Warp parallel-processing computers and a variety of peripherals. The
University also maintains an IBM 3090 mainframe and a high-performance
(interactive) link to the Cray Y-MP/864 at the Ohio Supercomputer Center in
Columbus, on which computing time is readily available.

Application deadline is April 26, 1991. Applicants should submit a resume and
arrange to have three letters of recommendation sent to

Arden Ruttan
Chair, Applied Mathematics Search Committee
Department of Mathematical Sciences
Kent State University
Kent, OH 44242

Kent State University is an Affirmative Action/Equal Opportunity Employer.


From: Marc Jacobs <MQJACOBS@VTVM1.CC.VT.EDU>
Date: Wed, 06 Feb 91 08:41:18 EST
Subject: Position in Computational Mathematics at AFOSR


The Air Force Office of Scientific Research (AFOSR) is seeking a
Mathematical Scientist to direct the Computational Mathematics grants
and contracts program (approximately $3.5M annually) within the
Mathematical and Information Sciences Directorate.
AFOSR's research managers are highly qualified individuals responsible
for the formulation, organization, execution, and transition of the
entire Air Force basic research program performed at universities,
industries, and Air Force and other government laboratories. If
successful, you will identify research opportunities, communicate Air
Force interests to the scientific community, evaluate proposals, review
ongoing projects, manage fiscal resources, and represent the program and
assist in the transition of research results within the Air Force and DOD.

The Computational Mathematics program has national scope. Previous
research managers have been responsible for the funding and direction of
research leading to significant advances such as multigrid, p and h-p
finite element techniques, shock capturing techniques, and fast parallel
matrix algorithms. You will be the catalyst for multidisciplinary AFOSR
research in high performance computing to meet Air Force needs in
scientific and engineering design and in operational uses. Interaction
with leading scientists and research managers in other disciplines such
as computational fluid dynamics, structures, propulsion, combustion
phenomena, control, electromagnetics, and weather modelling is encouraged.

A Ph.D. in Mathematics or Computer Sciences is highly desired. You
should have first-hand research experience and broad knowledge in the
latest developments in the field as demonstrated by a record of
publications in refereed scientific journals.
This is a career civil service position.

AFOSR, a part of Air Force Systems Command, is located in an attractive
setting on Bolling AFB, DC. Information regarding AFOSR and the
technical scope of the program can be obtained from the Director of the
Mathematical and Information Sciences Directorate: Dr Charles J. Holland
To apply you should send Standard Form 171, a resume, and a list of
publications to arrive by 1 March 1991 at the address
Civilian Personnel Office
1776 ABW/MSCA-90-650
ATTN: Deborah Williams
Andrews AFB, MD 20331-5964


Date: Thu, 7 Feb 91 10:16 EDT
Subject: Contents, SIAM Journal on Numerical Analysis

SIAM Journal on Numerical Analysis
JUNE 1991 Volume 28, Number 3


Stability of Higher-Order Hood-Taylor Methods
Franco Brezzi and Richard S. Falk

A Posteriori Error Estimates for the Stokes Problem
Randolph E. Bank and Bruno D. Welfert

Efficient Preconditioning for the P-Version Finite Element Method
in Two Dimensions
I. Babuska, A. Craig, J. Mandel, and J. Pitkaranta

Error Bounds for Numerical Solutions to Hydrodynamical Problems
Involving Shocks
D. L. Hicks and K. L. Kuttler

Upstream Differencing for Multiphase Flow in Reservoir Simulation
Yann Brenier and Jerome Jaffre

Convergence of the Galerkin Method for Two-Dimensional Electromagnetic
H. P. Urbach

Alternating Direction Collocation for Separable Elliptic Partial
Differential Equations
K. D. Cooper and P. M. Prenter

An Adapted Boundary Element Method for the Dirichlet Problem in
Polygonal Domains
M. Boulard, S. Nicaise, and L. Paquet

An Explicit Finite Element Method for Convection-Dominated Steady State
Convection-Diffusion Equations
Gerard R. Richter

The Sinc-Galerkin Method for the Fourth-Order Differential Equations
Ralph C. Smith, Gary A. Bogar, Kenneth L. Bowers, and John Lund

Numerical Computation and Continuation of Invariant Manifolds Connecting
Fixed Points
Mark J. Friedman and Eusebius J. Doedel

The Numerical Solution of Nonlinear Equations Having Several
Parameters, Part III: Equations with Z2-Symmetry
A. D. Jepson, A. Spence, and K. A. Cliffe

The Analysis of Generalized Backward Difference Methods
Applied to Hessenberg Form Differential Algebraic Equations
J. B. Keiper and C. W. Gear

Alternating Direction Implicit Iteration for Systems with Complex Spectra
Nancy S. Ellner and Eugene L. Wachspress

The Spectra of Super-Optimal Circulant Preconditioned Toeplitz Systems
Raymond H. Chan, Xiao-Qing Jin, and Man-Chung Yeung

Error Estimates in Gaussian Quadrature for Functions of Bounded Variation
Klaus-Jurgen Forster, and Knut Petras

For information regarding the SIAM Journal on Numerical Analysis,
please contact Vickie Kearn, Publisher, SIAM, 3600 University
City Science Center, Philadelphia, PA 19104-2688; telephone
(215) 382-9800; FAX: (215) 386-7999; e-mail:


arom: SIAM Publications Department <SIAMPUBS@WILMA.WHARTON.UPENN.EDU>
Date: Fri, 8 Feb 91 14:38 EDT
Subject: Contents, SIAM Journal on Computing

June 1991 Volume 20, Number 3


Fast Matching Algorithms for Points on a POlygon
Odile Marcotte and Subhash Suri

The Communication Complexity of Atomic Commitment and of Gossiping
Ouri Wolfson and Adrian Segall

Some Lower and Upper Complexity Bounds for Generalized Fourier
Transforms and Their Inverses
Ulrich Baum and Michael Clausen

On Vertical Visibility in Arrangements of Segments and the Queue
Size in the Bentley-Ottmann Line Sweeping Algorithm
Janos Pach and Micha Sharir

On Polynomial-Time Bounded Truth-Table Reducibility of NP Sets to Sparse Sets
Mitsunori Ogiwara and Osamu Watanabe

Nondeterministic Computations in Sublogarithmic Space and Space
Viliam Geffert

A 4n Lower Bound on the Combinational Complexity of Certain Symmetric
Boolean Functions over the Basis of Unate Dyadic Boolean Functions
Uri Zwick

Near-Testable Sets
Judy Goldsmith, Lane A. Hemachandra, Deborah Joseph, and Paul Young

Compression and Ranking
Andrew V. Goldberg and Michael Sipser

Algorithms for Scheduling Imprecise Computations with Timing Constraints
Wei-Kuan Shih, Jane W.S. Liu, and Jen-Yao Chung

Boolean Functions, Invariance Groups, and Parallel Complexity
Peter Clote and Evangelos Kranakis

Tests for Permutation Polynomials
Joachim von zur Gathen

For additional information, please contact Vickie Kearn,
Publisher, 3600 University City Science Center, Philadelphia,
PA 19104-2688; telephone: 215-382-9800; fax: 215-386-7999;


End of NA Digest