NA Digest Sunday, June 2, 1991 Volume 91 : Issue 22

Today's Editor: Cleve Moler

Today's Topics:


From: Michael Berry <>
Date: Mon, 27 May 91 09:06:12 CDT
Subject: Change of Address for Michael Berry

I will be relocating to Knoxville in a few weeks to joint the faculty
at the University of Tennessee. My title will be Assistant Professor
in the Department of Computer Science.


Here's the complete info:

Michael W. Berry
Assistant Professor
Department of Computer Science
University of Tennessee
107 Ayres Hall
Knoxville, TN 37796-1301

Office: 615-974-5607
Fax: 615-974-4404


From: Carlos Perez <>
Date: Tue, 28 May 1991 13:30:24 +0200
Subject: Using Conjugate Gradient for Navier Stokes

I'm investigating the behavior of conjugate-gradient methods for the
solution of the steady incompressible Navier-Stokes equations. I'm trying
to compare the classical line-by-line (LBL) relaxation methods,
solving each line of the domain with the Thomas tridiagonal matrix
algorithm, with the ICCG applied to the pressure equation (the momentum
equations are still solved with the LBL method). I use the SIMPLE and the
SIMPLER methods to treat the pressure-velocity coupling.

The problem is that, although the references I have claim a speed-up
factor of almost 7 when comparing both methods, I don't get any improvement
at all. ICCG reduces the residuals of the pressure equation more than the LBL
at each iteration of the global non linear procedure, but the number of total
iterations (the outer iterations of the non linear iterative scheme, not
the inner iterations of the CG) is almost unaffected. As the time to
perform an ICCG iteration is greater than the time for a LBL iteration, I'm
actually slowing down the code.

- Does anybody have any experience or suggestion about this particular
problem that could be helpful for me?

- I would like to contact via Email with the authors of the reference I
mentioned above. Here they are:

'Three-dimensional unsteady flow simulations: alternative
strategies for a volume-averaged calculation' Chin-Yuan Perng and Robert L.
Street. Environmental Fluid Mechanics Laboratory, Stanford University,
Stanford CA 94305 USA. (Published at the International Journal for
Numerical Methods in Fluids, vol. 9 341-362 (1989)). Does anybody know
their address?

Thanks in advance.

Carlos Perez Caseiras
Fluid Mechanics Dept. Email:
Centro Politecnico Superior
Maria de Luna 3
50015 Zaragoza


From: 0ystein Olsen <>
Date: 28 May 91 09:46:44 GMT
Subject: LSODA (ODE integration) Request

I am looking for LSODA (Petzold and Hindmarsh 1987) for integrating
ODEs. Can anybody tell me where on Internet I can find the package?

Oystein Haug Olsen
University of Glasgow, U.K.



From: Jerzy Wasniewski <>
Date: Wed, 29 May 91 16:41:33 DNT
Subject: Looking for Strassen's Methods

Looking for Strassen's Methods written in Fortran or C

I'm looking for Strassen's methods for the matrix multiplication
and/or matrix inversion written either in Fortran or in C. Any help
will be appreciated.

Thank you very much in advance.

Jerzy Wasniewski
The Danish Computing Center for
Research and Education, UNI-C
Eaddress: jerzy.wasniewski at


From: Steven Abrams <>
Date: Wed, 29 May 1991 12:22:21 EDT
Subject: Constrained Non-linear Optimization


We are trying to find some source code (preferably C, but fortran is
acceptable) for a constrained nonlinear optimization routine. We have been
using the IMSL package's NCONF routine which is a sequential quadratic
programming method. It has yielded acceptable results, but we are looking
for something which is more flexible than IMSL. Methods which require
gradients are acceptable (and in some ways preferable).

Also, if you know of no publicly available source code, any pointers
to other commercial packages which have constrained nonlinear
optimization routines would be appreciated. We're NOT interested in
methods that solve LINEAR or UNconstrained problems, so I believe that
eliminates LINPACK, TOMS, and anything available via netlib, but I'd
love for someone to tell me that I'm wrong.

Please respond via e-mail as I am not on this list. I will summarize
responses if they are sufficiently informative.




From: Yves Robert <yrobert%FRENSL61.BITNET@CUNYVM.CUNY.EDU>
Date: Mon, 27 May 91 14:34:27 +0200
Subject: New Book by Yves Robert

Series Algorithms and Architectures for Advanced Scientific Computing


by Yves ROBERT
Professor of Computer Science at the Ecole Normale Superieure de Lyon

Vector and parallel architectures have a tremendous impact on
efficient algorithm design. The book demonstrates this impact and
examines its nature. Using a single algorithm, Gaussian elimination,
which is at the very heart of numerous scientific computing applications,
the key features of each main class of parallel architectures are captured,
and the optimum ways to match algorithm design with architectural
constraints are deduced.

Beginning with background material on Gaussian elimination and
parallel processing terminology, the text introduces pipeline,
vector and parallel architectures. Part 1 is implementation oriented and
describes the re-structuring techniques needed for shared memory vector
multiprocessors, distributed memory systems and systolic arrays.
Part 2 is more theoretical. Task graph scheduling is the subject of
chapter 6. Complexity results and speedup evaluation in a distributed
memory environment are discussed, and automatic synthesis methods
for systolic arrays are introduced.

This book will be both challenging and enjoyable to graduate,
advanced students and researchers in the field of Computer Science,
Scientific Computing, Numerical Analysis and Electrical Engineering.

Manchester University Press (1991)
UK ISBN 0 7190 3365 9
distributed in the USA by Halsted Press, an imprint of John Wiley & Sons
US ISBN 0 470 21703 0

Yves Robert tel + 33 72 72 83 89
Laboratoire LIP/IMAG fax + 33 72 72 80 80
Ecole Normale Superieure de Lyon
46 allee d'Italie, F - 69364 Lyon Cedex 07 yrobert@frensl61.bitnet


From: Daniel B. Szyld <>
Date: Thu, 30 May 91 12:28:13 EST
Subject: Special Session on Numerical Linear Algebra

There will be a Special Session on Numerical Linear Algebra at
the American Mathematical Society (AMS) meeting 868, in Philadelphia,
October 12-13, 1991. Listed below are the speakers and the title
of their presentations. Abstracts of these talks will appear in the
Abstracts of the AMS before the conference. Further information,
including hotel and travel directions, will appear in the
Notices of the AMS later on this summer.
You can also contact the session organizer:
Daniel B Szyld, Department of Mathematics, Temple University 038-16,
Philadelphia, PA 19122-2585, e-mail :

Jesse L. Barlow Error Analysis of Update Methods for the Symmetric
Eigenvalue Problem
Ralph Byers Numerical Computation of an Analytic Singular
Value Decomposition of a Matrix Valued Function
Howard Elman Parallel Implementation of the hp-version of the
Finite Element Method on a Shared-Memory Architecture
Anne Greenbaum Laplace's Equation and the Dirichlet-Neumann Map
in Multiply Connected Domains
Linda Kaufman Modifying the Shougen-Shuqin algorithm for the banded
Symmetric Generalized Matrix Eigenvalue Problem
Ivo Marek Convergent Iterative Methods to Find Stationary
Distribution of Markov Processes
Carl D Meyer A Direct Method For Solving Unstructured Sparse Linear
Michael Neumann Asynchronous parallel iterations for nonlinear systems
Michael Overton On the Convex Hull of Projection Matrices
Ricardo Pantazis Parallel Solution of Generalized Band Eigenvalue
Problems by Sectioning
Mordukh Primak A projection cutting plane algorithm for convex
programming problems
Lothar Reichel A new implementation of the GMRES method
G.W. Stewart Updatable, Rank-Revealing Decompositions
William Stewart Some Convergence Properties of Markov Chains
Gilbert Strang Joint Stability of Two Matrices and their Joint
Spectral Radius
L. N. Trefethen Eigenvalues and pseudo-eigenvalues of nonsymmetric
random matrices
Olof B. Widlund Some Recent Results on Schwarz Type Domain
Decomposition Algorithms


From: Paul Concus <>
Date: Fri, 31 May 91 16:25:45 PDT
Subject: Postdoctoral Fellowship at LBL


A position is available for a postdoctoral research fellow in
computational and applied mathematics at the University of California
Lawrence Berkeley Laboratory. The fellow would work in a group whose
interests include, but are not limited to, computational fluid
mechanics, turbulence theory, many-body systems, percolation theory,
Monte-Carlo methods, grid generation, combustion, interface problems,
and parallel computing. The position 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
recipient be a U.S. citizen or permanent resident. Interested persons
should send a curriculum vitae and the names of three references before
August 1, 1991 to Prof. Alexandre Chorin, 90-1042 (A6561), Lawrence
Berkeley Laboratory, University of California, Berkeley, CA 94720.
Affirmative Action/Equal Opportunity Employer.

-- Paul Concus


From: John Carroll <>
Date: Mon, 27 May 91 11:42 GMT
Subject: Proceedings of ILIAM VIII


The proceedings of the eight ILIAM seminar which was held in Dublin
City University on Friday, May 17, 1991 under the auspices of the
School of Mathematical Sciences are now available.

The ILIAM (Information Linkage between Industry and Applied
Mathematics) seminar series started in 1984 when a one day meeting,
ILIAM I, was held in the then NIHE, Dublin, as a first attempt to
forge links between the industrial and academic worlds in the
particular context of applied mathematics.

The immediate goals of ILIAM were to direct the attention of applied
mathematicians to some of the real problems of industry and to
acquaint industrialists with recent advances which could enhance their
capability for design, manufacture and distribution.

An intended feature of ILIAM was the presentation of some case studies
which illustrated an engineering or industrial approach to a problem,
indicated some possible difficulties in its implementation and
suggested areas where the applied mathematician might help.

ILIAM VIII Proceedings

"The link with financial services --- an industry view"
Tom Conlon (Management Services, Bank of Ireland)

"Numerical simulation in the semiconductor industry"
Susan Edwards, Analogue Devices

"Modelling environment in a mushroom growing tunnel"
Jim Grant (Teagasc), Denis Williams (Dublin City University)

"Initialization for numerical weather prediction"
Peter Lynch, Meteorological Service

"Interaction of sea waves with large diameter vertical columns"
T. P. Mullarkey, M. D. Gilchrist, J. F. McNamara
(University College Galway and Marine Computation Services)

"A Description of RESOLUTION: A mathematical toolbox for those
working in a quantitative discipline"
Paul Neary (PS Squared)

"DEQSOL/AI-DEQSOL: Towards an integrated problem solving
environment for numerical simulation"
Nobutoshi Sagawa, Donal Finn, Neil Hurley (Hitachi)

"An application of the expert system shell Xi-Plus to fire safety
I. R. Taylor, H. A. Donegan (University of Ulster)

Copies of the proceedings are available at $15 per copy from:

ILIAM VIII Secretariat
Mathematical Sciences
Dublin City University
Dublin 9


and please mark payment to ILIAM VIII.


From: Richard Brualdi <>
Date: Tue, 28 May 91 07:35:57 cdt
Subject: Table of Contents: Linear Algebra and Its Applications

Table of Contents

Volume 153 July 15, 1991 PP. 1-228

S. M. Rump (Hamburg, Germany)
Estimation of the Sensitivity of Linear and Nonlinear Algebraic Problems

Jinchuan Hou (Shanxi, China)
On Operator Inequalities and Linear Combinations of Operators

Juan-Enrique Martinez-Legaz (Barcelona, Spain) and
Ivan Singer (Bucharest, Romania)
Compatible Preorders and Linear Operators on Rn

Matjaz Omladic and Peter Semrl (Ljubljana, Yugoslavia)
Spectrum-Preserving Additive Maps

H. Osaka (Tokyo, Japan)
Indecomposable Positive Maps in Low Dimensional Matrix Algebras

Ji-guang Sun (Beijing, China)
Perturbation Expansions for Invariant Subspaces

Carsten Scherer (Wurzburg, Germany)
The Solution Set of the Algebraic Riccati Equation and the Algebraic
Riccati Inequality

K. T. Arasu (Dayton, Ohio), Willem H. Haemers (Tilburg, The Netherlands),
Dieter Jungnickle, and Alexander Pott (Giessen, Germany)
Matrix Constructions of Divisible Designs

S. J. Lee (Tampa, Florida) and M. Z. Nashed (Newark, Delaware)
Normed Linear Relations: Domain Decomposability, Adjoint Subspaces,
and Selections

Peter Nylen and Tin-Yau Tam (Auburn, Alabama)
Numerical Range of a Doubly Stochastic Matrix

Mohamed Barraa (Marrakech, Maroc) and Bernard Charles (Montpellier, France)
Sous-espaces Invariants d'un Operateur Nilpotent sur un Espace de Banach

H. Valiaho (Helsinki, Finland)
A Polynomial-Time Test for M-matrices

Artibano Micali (Montpellier, France) and Moussa Ouattara (Burkina Faso)
Dupliquee d'une Algebre et le Theoreme d'Etherington

Jennifer Ryan (Denver, Colorado)
Decomposing Finitely Generated Integral Monoids by Elimination

Book Reviews
John H. Randall (Stellenbosch, South Africa)
Review of Linear Least Squares Computations by R. W. Farebrother

Pamela G. Coxson (Berkeley, California)
Review of Nonnegative Matrices in Dynamic Systems
by A. Berman, M. Neumann, and R. J. Stern


Date: Wed, 29 May 91 13:45 EDT
Subject: Table of Contents: SIAM Scientific and Statistical Computing

Scientific and Statistical Computing

September 1991 Volume 12, Number 5

Runge--Kutta Defect Control Using Hermite--Birkhoff Interpolation
Desmond J. Higham

Iterated Runge--Kutta Methods on Parallel Computers
P. J. van der Houwen and B. P. Sommeijer

A Control Volume Finite Element Approach to NAPL Groundwater Contamination
Peter A. Forsyth

Stability Analysis and Improvement of the Block Gram--Schmidt Algorithm
W. Jalby and B. Philippe

Cell Discretization of Nonselfadjoint Linear Elliptic Partial Differential
John Greenstadt

An Improved Newton Iteration for the Generalized Inverse of a Matrix,
with Applications
Victor Pan and Robert Schreiber

The Fast Gauss Transform with Variable Scales
John Strain

Comparing Averaged-Out Utilities of Probability Trees Having Random Parameters
R. D. Eisenhut, R. S. Dittus, S. D. Roberts, and J. R. Wilson

A Stabilized Treatment of the Biharmonic Operator with Spectral Methods
Wilhelm Heinrichs

A Scheme for Handling Rank-Deficiency in the Solution of Sparse Linear
Least Squares Problems
Esmond Ng

Limiting Communication in Parallel Sparse Cholesky Factorization
Laurie Hulbert and Earl Zmijewski

An Adaptive Finite Element Method for Two-Phase Stefan Problems in Two
Space Dimensions. Part II: Implementation and Numerical Experiments
R. H. Nochetto, M. Paolini, and C. Verdi

The Equations of Motion of Particles in Smoothed Particle Hydrodynamics
G. V. Bicknell

Timely Communication

Block Gauss Reduction to Hessenberg Form
Augustin A. Dubrulle


End of NA Digest