NA Digest Saturday, October 20, 1991 Volume 91 : Issue 42

Today's Editor: Cleve Moler

Today's Topics:

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to


Date: Mon, 14 OCT 91 15:56 N
Subject: ODE Test Problems

During the recent meeting on Parallel ODE Methods (Grado-Italy
Sept. 10-13) it emerged the need of a new set of test initial
value problems which are more suitable than existing ones for
testing parallel algorithms.
This set of test problems should include large scale systems
on which to compare different algorithms, different architectures
and, in general, different parallel computing environments.
On this base, I am inviting the numerical ODE community, as well
as the users of ODE solvers, to send problems revealing large
computational complexity which are expected to be solved on
parallel computers, and problems for which existing codes prove
inadequate anyway.
Test systems should be endowed with analytic solutions, if
available. I will collect, select and distribute them to all
A. Bellen
Dipartimento di Scienze Matematiche
I-34100 TRIESTE, Italy



From: Robert G. Voigt <>
Date: Thu, 17 Oct 91 10:58:26 -0400
Subject: Directorship at ICASE

Effective October 10, 1991, I have resigned as Director of ICASE.
Preliminary steps for initiating a search for a new Director have been taken,
and the details of a transitional plan are being worked out.

For the forseeable future my address remains:

Robert G. Voigt
Mail Stop 132-C
NASA Langley Research Center
Hampton, VA

Phone: (804) 864-2174


Inquiries regarding positions, visits, reports, etc. should be directed to:

Mail Stop 132-C
NASA Langley Research Center
Hampton, VA

Phone: (804) 864-2174


I want to thank all of you who have shown an interest in the ICASE
program over the years and urge you to continue whatever form of
interaction is most appropriate in the future.


From: Raymond Mejia <>
Date: Tue, 15 Oct 91 13:13:26 -0400
Subject: Referees Do Have Some Leverage

Let me add an item to Eduardo Sontag's list of suggestions for improving the
review process:

Be a hard-nosed referee.

I believe that reviewers can spur Editors to be considerate and responsive
in distributing the "burden". I for one will ask to see a paper after
requesting that a major revision be made. If I do not receive it, and
subsequently see the paper in print without addressing the request, well
you can imagine my inclination the next time that a request is made to review
for that journal. Referees have not only an obligation to do a prompt,
thorough review, but the right to expect that an editorial board does it's
job to screen, assign and follow-up. The point is that referees do have some
leverage and a responsibility for what is published.

There certainly appear to be more papers (more poorly written and more poorly
edited) published today in my area of interest than ten years ago. However,
the professions can correct that through the careful work of authors,
editors and referees.

Ray Mejia


From: Joel Storch <>
Date: 15 Oct 91 15:09:24 EDT
Subject: Journal Refereeing

It is obvious from the many errata and 'comments to the author' that appear
in journals, that the original reviewers did not critically read the
manuscript. For those in industry, the paper review process becomes an
additional burden i.e. there employers do not view this task as part of their
work assignment. With limited time available, the review is often hasty and
lacks attention to detail.

As a partial remedy to this problem, R&D organizations should provide
special charge numbers explicitly intended for refereeing of papers. The amount
of time each employee could charge to this number annualy would be set by the
organization. With such a mechanisn in place, the review process is put on
equal footing with any other job assignment. The message conveyed here is that
refereeing of papers is serious business and demands the same degree of
dedication as any other project. At the end of each year, the journal could
print an acknowledgment to organizations participating in this plan along with
the number of hours donated (tax incentive ?)

Joel Storch


From: Gene Golub <golub@sccm.Stanford.EDU>
Date: Thu, 17 Oct 91 0:00:47 PDT
Subject: Proposal Reviewing

We have seen quite a flurry of notes on the problems associated with
journal reviewing. Another activity that we need to consider as
professionals is proposal refereeing for government agencies such as
NSF. Those agencies depend heavily on peer review. The program
managers work very hard in handling proposals and we owe them a debt
of gratitude. But they can only do as well as the reports that they
receive. Therefore, it's important to get reviews to them in a timely

I would like to make another point, however. Frequently, under the
cloak of anonymity reviewers make unprofessional remarks. Perhaps they
feel that they are in competition for funds. I think we might have a
healthier atmosphere if the reviews were signed. This may lead to more
responsible and scholarly reviews.

Gene Golub


From: Richard A. Brualdi <>
Date: Wed, 16 Oct 91 12:20:34 CDT
Subject: LAA Special Issue Honoring Marvin Marcus

MARVIN MARCUS has retired from his position as Professor of Computer
Science at the University of California in Santa Barbara and is now
Emeritus Professor of Computer Science. A special issue of the
for his many important contributions and service to linear and
multilinear algebra. The purpose of this announcement is to solicit
papers for this special issue.

Anyone may submit a paper for this issue. Contributions should be
appropriate for publication in LAA and will be subject to the usual
review process. The deadline for submissions is December 31, 1992.

Papers should be submitted to one of the special editors of the issue:

Bryan E. Cain
Department of Mathematics
Iowa State University
Ames, Iowa 50011

Moshe Goldberg
Department of Mathematics
Technion-Israel Institute of Technology
32000 Haifa, Israel

Robert Grone
Department of Mathematical Sciences
San Diego State University
San Diego, California 92182

Nicholas J. Higham
Department of Mathematics
University of Manchester
Manchester M13 9PL, England.

Papers can also be submitted through the editor-in-chief. Publication
of the special issue is planned for early 1994.
We hope that you will join in this tribute to our esteemed colleague.


From: F. W. Olver <>
Date: Fri, 18 Oct 91 16:36:47 EDT
Subject: IEEE & SLI Arithmetics

In his message to the NA Digest dated September 26, 1991, James Demmel
responded to various criticisms of the IEEE Standard and floating-point
arithmetic that have surfaced recently in the Digest. Inter alia he
suggested, quite rightly, that some of the proposed new arithmetics are
in need of extensive testing before definitive assessments can be made.
It is not our intention to prolong these discussions ad nauseam in the
Digest. The purpose of the present message is simply to clear up some
misconceptions about sli arithmetic that might possibly result from
Demmel's comments. A failure by us to respond to future messages should
not be interpreted as assent on our part.

Two basic problems attend any system of arithmetic used on a computer:
a limitation of precision, and a limitation of range. Our claim is that
sli arithmetic solves the second problem, indeed solves it completely,
but not the first. The precision of any given system can be exhausted
by attempting a sufficiently ill-conditioned problem. In this event, the
only recourse is to employ a higher working precision, that is, to
increase the total word length, or to seek an algebraic or analytical
reformulation of the problem. Floating-point systems, including of
course IEEE arithmetic, solve neither the precision problem nor the
range problem.

Let us examine these two problems in more detail.

PRECISION. Demmel is concerned by the possible severe, or indeed complete,
loss of relative precision associated with the huge numbers that are
representable in the sli system. We first comment that the constant use
of relative precision as error norm may be regarded as a consequence
of the everyday use of floating-point arithmetic that has been forced
upon us by almost all computers. It is not always the right measure.
For example, no physicist would dismiss Dirac's estimate 10**(78+or-1) of
the number of particles in the universe as useless information on the
grounds that this estimate has no relative precision. Only a numerical
analyst steeped in floating-point arithmetic might react in this
manner. For any given number the "correct" measure of its precision
surely depends on subsequent use to be made of this number.

It is true that in many algorithms a combination of relative precision
and absolute precision serves as an adequate error measure for
all numbers generated during the course of the computation.
However, huge sli numbers will not injure these algorithms.
To understand why we make this assertion it suffices to consider
Demmel's example---the formation of a product of a sequence of numbers
by successive multiplications. It doesn't matter a great deal what the
magnitudes actually are. Presumably the Committee on the IEEE Standard
felt that all numbers that appear in everyday computations are bounded
by 2**(2**10), so let us assume that our factors all lie between
1+2**(-52) (the smallest double-precision IEEE number that exceeds unity)
and 2**(2**10). Then it would take between 3.5*(10**12) and
1.2*(10**31) multiplications before an sli number is reached that is
devoid of any relative precision in its double-precision representation.
(See Lozier & Olver SIAM J. Num. Anal.,v.27 pp.1295-1304,1990.)
However, with this number of multiplications there is going to be an
enormous build-up of relative error stemming from the combined effects
of the inherent errors in the original factors, and (especially)
rounding errors made in storing the partial products. So much so that
the problem is too ill-conditioned to be handled in double-precision
sli. And if we elect to overcome this difficulty by shifting to triple
or quadruple precision, that is, to an aggregate word length of 96 or 128
bits, then voila---the sli form is capable of representing the final result
to a relative precision of 2**(-32) or 2**(-64). (Furthermore, even with
this increased precision the usual forms of floating-point arithmetic would
continue to yield no relative precision because of overflow.)

It is also worth reflecting on the size of this pathological problem of
Demmel. Assuming a speed of 10**(-7) seconds, say, for each sli
multiplication it would take between 4 days and 4*(10**16) years of non-
stop operation to complete, without making any allowance for overhead.
(Or maintenance!) And of course if we also introduce factors of
magnitude less than unity, as Demmel's example permits, then these times
will be increased further.

This is not to say that such very large numbers will never be attained.
They can be reached quite rapidly in algorithms that use repeated
squarings, for example. However, an algorithm that uses repeated
squarings is one that is also likely to have a compensating
feature, from the standpoint of precision loss and gain, in the form
of repeated root formation. This means that an intermediate complete
loss of relative precision is of no final consequence. Illustrations
of this phenomenon can be found in Lecture Notes in Mathematics
No.1397 "Numerical Analysis & Parallel Processing", ed. by P.R.Turner,
Springer-Verlag,1989,pp.124-130 and 146-156, and the paper by
Clenshaw and Turner in Computing v.43,pp.171-185,1989.

RANGE. We first comment that the existence of grafted-on devices in
IEEE arithmetic such as NaN's, infinities, gradual underflow and
wrap-around exponents provides an eloquent testimony to its range
failure. Basically these devices do little more than fiddle around at
the edge of the problem.

Nevertheless, Demmel implies that any given problem with floating-point
overflow or underflow can always be overcome by making a sufficient effort
of reformulation and recoding---an observation with which we agree fully.
(For the example discussed above, this might be achieved, for example,
by the nontrivial software device of allocating extra words to the
storage of exponents.) But we also note that a similar argument was once
advanced by defenders of the fixed-point system. The lasting advantage of
floating-point over fixed-point proved to be its greater convenience, that
is, a vast saving of human time and effort. Moreover, this convenience
was realised without the feared effect of an unacceptable loss of
precision (in this case absolute precision). There certainly was not a
"conservation of effort in writing robust code...". Assuming that
satisfactory hardware for sli will become available in due course, then
notwithstanding Demmel's assertion to the contrary, we believe that
the same conclusion will emerge eventually with respect to a changeover
to sli arithmetic.

Daniel Lozier (, Frank Olver (
and Peter Turner (


From: Mike Baines <>
Date: Mon, 14 Oct 91 18:08:02 BST
Subject: ICFD Conference on Fluid Dynamics

The next ICFD Conference on Numerical Methods for Fluid Dynamics
organised by the ICFD at Oxford and Reading, UK, will be held
from April 7th to 10th 1992 at the University of Reading.

This is the 3rd international conference on CFD organised by the
Institute for Computational Fluid Dynamics (ICFD). The aim of the
conference, as in previous years, is to bring together mathematicians
and engineers and other scientists to review recent advances in
mathematical and computational techniques for modelling fluid flows.

The conference will cover all areas of CFD but with special attention
to numerical analysis that needs to be exploited in CFD and numerical
challenges posed by CFD. It is expected to give particular emphasis
to the following topics:

NUMERICAL BOUNDARY CONDITIONS (particularly non-reflective)

The following have accepted invitations to give talks.

D Catherall (Farnborough)
C Farmer (ECL)
B Fornberg (Exxon)
B Gustafsson (Uppsala)
A Hutton (Nuclear Electric)
B Koren (Amsterdam)
R LeVeque (Washington)
K W Morton(Oxford)
B Palmerio (INRIA)
M Pandolfi (Turin)
J Peraire (Imperial College)
P L Roe (Michigan)
E Tadmor (Tel Aviv)
H Yee (NASA Ames)

In addition to the invited lectures the programme will include
contributed talks of 20 mins and poster sessions. Two page abstracts
should be submitted by
3rd December 1991
stating preference
for oral or poster presentation. Notification of acceptance or
rejection will be given by 3rd February 1992. The proceedings
(to contain all oral presentations but not posters) will be
published by the Oxford University Press.

Abstracts and enquiries regarding the conference should be sent
Bette Byrne
ICFD Secretary
Oxford University Computing Laboratory
11 Keble Road
Oxford OX1 3QD
Tel. (0)865 273883 Fax. (0)865 273839


From: Richard A. Brualdi <>
Date: Wed, 16 Oct 91 11:48:02 CDT
Subject: Minnesota IMA Workshop

Minnnesota IMA workshop, Nov. 11-15, 1991

The IMA (Institute for Mathematics and its Applications at the University
IN LINEAR ALGEBRA will take place November 11-15, 1991 in Minneapolis.
The number of talks has been limited in order to leave plenty of free
time for informal discussions. Fourteen of the invited people have been
chosen to speak with a view towards having a broad and diverse program.
The speakers with titles of their talks are:

Avi Berman (Technion): Completely positive graphs;
Michael Boyle (Univ. of Maryland): Symbolic dynamics and inverse problems
for nonnegative matrices;
Fan Chung (BellCore): Laplacians of graphs and hypergraphs;
Miroslav Fiedler (Math. Inst. Acad. Prague): A geometric approach to the
Laplacian matrix of a graph;
Shmuel Friedland (Univ. of Illinois, Chicago): Real eigenvalues of almost
skew symmetric matrices and applications to tournament matrices;
Christopher Godsil (Univ. of Waterloo): Polynomial spaces and the Schur
Clark Jeffries (Clemson Univ.): Some matrix patterns arising in queuing
Charles Johnson (College of William and Mary): Two recent topics of
qualitative matrix theory;
Alex Lubotzky (Hebrew University): Ramanujan Diagrams;
Bojan Mohar (Univ. of Ljubljana, Yugoslavia): Eigenvalues of graphs in
combinatorial optimization;
Kazuo Murota (Univ. of Tokyo): Combinatorial canonical form of layered
mixed matrices;
Peter Rowlinson (Stirling Univ.): Eutactic stars and graph spectra;
Hans Schneider (Univ. of Wisconsin): Ranks of zero patterns;
Walter Wallis (Southern Illinois Univ.): Hadamard matrices.

Questions about the workshop can be directed to Richard Brualdi:
( or 612-624-7073).


From: R. Baker Kearfott <>
Date: Thu, 17 Oct 91 17:54:53 CDT
Subject: 1993 Conference on Interval Methods


A Conference on
Analysis, Applications, and Software
February 25 through March 1, 1993
Lafayette, Louisiana


Interval analysis is applicable in scientific computations in which
reliability, thoroughness, or verification of computational results are

This conference has the following goals.

* To provide an accessible forum for researchers in the field to
exchange the most recent results in interval computations.

* To further delineate the role of interval computations in
practical (applied and industrial) problems, and to identify tasks
which must be completed to facilitate its optimal use in such

* To highlight the role of interval mathematics in more purely
academic pursuits, such as automatic theorem proving.

Theory, software, computational results, etc. will be presented.

Topics covered include, but are not limited to


We plan to publish a refereed proceedings.


* G. Alefeld (University of Karlsruhe)
* G. Corliss (Marquette University)
* B. Kearfott (University of Southwestern Louisiana)
* U. Kulisch (University of Karlsruhe)
* H. Stetter (Technical University of Vienna)


To obtain further information, contact

Interval Methods Conference
C/O R. Baker Kearfott
Department of Mathematics
University of Southwestern Louisiana
U.S.L. Box 4-1010
Lafayette, LA 70504-1010

Office phone: (318) 231-5270
Home phone: (318) 981-9744
email: (Internet)


From: Alvaro R.De Pierro <>
Date: Fri, 18 Oct 1991 19:21 GMT-0300
Subject: IMPA Summer Workshop in Rio de Janeiro


The Workshop will take place at the Institute for Pure and
Applied Mathematics(IMPA),Rio de Janeiro,in January 8-10,1992.
It will consist of one-hour lectures by invited speakers and
sessions for short communications. Additional lectures by
invited speakers will be scheduled in January 6-7.

Abstracts of short communications consisting of no more than 4
double spaced pages,must be submitted before November 1st,1991.
Oral presentation of short communications should last no more than
20 minutes and must be offered either in english or portuguese.

Some of the topics that will be discussed at the workshop are:
-Lanczos method for solving ill-posed problems
-duality in regularized problems
-stopping rules for iterative methods for ill-posed problems
-iterative methods for radiation therapy planning
-the EM algorithm and its application in emission tomography
-implicit iterative methods for ill-posed problems .

Invited speakers:

More information on the Workshop may be obtained from:
Dr.Alvaro R.De Pierro:
Dr.Alfredo N. Iusem:iusp@lncc.bitnet


From: Jack Dongarra <>
Date: Sat, 19 Oct 91 09:10:30 -0400
Subject: Symposium on Massively Parallel Computation

FRONTIERS '92: The 4th Symposium on the
Frontiers of Massively Parallel Computation

October 19-21, 1992
McLean Hilton McLean, Virginia

Sponsored by IEEE Computer Society
NASA Goddard Space Flight Center

This symposium is the fourth in a series of biannual meetings on massively
parallel computation, focusing on research related to, or adaptable for,
systems with 1,000 or more processors. Submissions of original research
papers about any aspects of the design, analysis, development, and/or use
of massively parallel computers are solicited. Papers relating to high
performance computing and communications are of particular interest.


Professor H. J. Siegel
School of Electrical Engineering
1285 Electrical Engineering Building
Purdue University
West Lafayette, IN 47907-1285, USA

Submission related questions should be sent to the internet address:

Selected papers will be eligible for publication in a dedicated issue of the
Journal of Parallel and Distributed Computing, to be guest-edited by Joseph
JaJa, University of Maryland, and David Schaefer, George Mason University.

Professor Isaac D. Scherson
Department of Information & Computer Sciences
University of California- Irvine
Irvine, CA 92717, USA
fax: (714) 856-4056


Ms. Judy Devaney
Building 225, Room B-146
Gaithersburg MD 20899-0001, USA
phone: (301) 975-2882


Professor Abdou Youssef
Department of EE and CS
School of Engineering and Applied Science
George Washington University
Washington, DC 20052, USA
phone: (202) 994-5513

GENERAL CHAIR - Pearl Wang, George Mason University

PROGRAM CHAIR - H.J. Siegel, Purdue University

Algorithms - Leah Jamieson, Purdue University
Architectures - Ken Batcher, Kent State University
Applications - Jack Dongarra, Univ Tenn/ORNL
Software - Andre van Tilborg, ONR

Fran Berman, UC - San Diego Tom Blank, Maspar Computer
Jim Browne, Univ Texas-Austin Tom Casavant, Univ Iowa
Janice Cuny, Univ Massachusetts Larry Davis, Univ Maryland
Doug DeGroot, Texas Instruments Hank Dietz, Purdue Univ
John Dorband, NASA GSFC John Feo, Lawrence Livermore Nat Lab
Jeanne Ferrante, IBM TJ Watson Raphael Finkel, Univ Kentucky
Geoffrey Fox, Syracuse University Richard Freund, NOSC
John Gustafson,Ames Lab/Iowa St. Susanne Hambrusch, Purdue Univ
Mary Jane Irwin, Penn State Univ Anita Jones, Univ Virginia
Russ Miller, SUNY - Buffalo Dennis Parkinson, AMT
Donna Quammen, George Mason U John Riganati, SRC
Sartaj Sahni, Univ Florida Thomas Schwederski,
Inst. Microelectronics Stuttgart
Marc Snir, IBM TJ Watson Ted Tabloski, Thinking Machines
Stephen Taylor, Caltech Patricia Teller, NM State Univ
Elizabeth Williams, SRC Michael Wolfe, Oregon Grad Inst.
Pen-chung Yew, Univ Illinois Abdou Youssef, Geo Washington U

Ken Batcher, Kent State Univ Jerry Brackbill, LANL
Harold Breaux, BRL Mel Ciment, NSF
Hank Dardy, NRL Larry Davis, Univ Maryland
Marvin Denicoff, TMC John Dorband, NASA GSFC
Milt Halem, NASA GSFC R. Michael Hord, GE
Gregory McRae, CMU Paul Messina, Caltech
Tor Opsahl, CIT David Schaefer, George Mason U
Bill Scherlis, DARPA Paul Schneck, SRC
Francis Sullivan, NIST Charles Taylor, UCLA

If you would like an advance program and registration information
for FRONTIERS '92, please contact

James Fischer
Frontiers '92
Code 932.1
NASA Goddard Space Flight Center
Greenbelt, MD 20771 USA


Date: Wed, 16 Oct 91 09:04 EDT
Subject: Table of Contents, SIAM Applied Mathematics

Table of Contents
SIAM Journal on Applied Mathematics
Vol. 52, No. 1, February 1992

The Oscillations of a Bubble Moving in an Inviscid Fluid
James Q. Feng

Flooding and Flow Reversal in Annular Two-Phase Flow
A. C. Fowler and P. E. Lisseter

Homogenization of Linear Transport Equations with Oscillatory Vector Fields
Thomas Y. Hou and Xue Xin

Homogenization Approach to Light Scattering from Polymer-Dispersed Liquid
Crystal Films
Avner Friedman and Bei Hu

Multicomponent Chromotography in a Two Phase Environment
Olav Dahl, Thormod Johansen, Aslak Tveito, and Rangar Winther

On Shock Wave Solutions in Extended Discrete Kinetic Theory
G. Spiga and S. Oggioni

Singular Perturbations and a Free Boundary Problem in the Modeling of Field
Effect Transistors
Michael J. Ward

Physical Parameters Reconstruction of a Free-Free Mass-Spring System from
Its Spectra
Yitshak M. Ram and James Caldwell

Diffusion Controlled Smoulder Propagation Parallel to a Plane Surface
J. Adler and D. M. Herbert

Diffusion and Reaction Caused by Point Catalysts
Donald A Dawson and Klaus Fleischmann

Rotating Chemical Waves in the Gray--Scott Model
W. W. Farr and M. Golubitsky

Global Dynamics of a Mathematical Model of Competition in the Chemostat:
General Response
Functions and Differential Death Rates
Gail S. K. Wolkowicz and Zhiqi Lu

Asymptotic Analysis of an Integrated Digital Network
Margo L. Mankus and Charles Tier

Intrinsic Random Functions and the Paradox of l/f Noise
Victor Solo

On the Feasibility of Cross-Validation in Image Analysis
Peter Hall and Inge Koch

For more information contact Vickie Kearn, Publisher, SIAM, 3600 University
City Science Center, Philadelphia, PA 19104-2688. Phone: (215) 382-9800.


Date: Wed, 16 Oct 91 11:29 EDT
Subject: Table of Contents, SIAM Scientifc and Statistical Computing

SIAM Journal on Scientific and Statistical Computing
January 1992 Volume 13, Number 1

Special Issue Devoted to the Copper Mountain Conference on Iterative Methods,
April 1--6, 1990

A Comparison of Adaptive Chebyshev and Least Squares Polynomial Preconditioning
for Hermitian Positive Definite Linear Systems
Steven F. Ashby, Thomas A. Manteuffel, and James S. Otto

Preconditioned Iterative Methods for Homotopy Curve Tracking
Colin Desa, Kashmira M. Irani, Calvin J. Ribbens, Layne T. Watson, and Homer F.

A Block Projection Method for Sparse Matrices
Mario Arioli, Iain S. Duff, Joseph Noailles, and Daniel Ruiz

Sparse Approximation for Solving Integral Equations with Oscillatory Kernels
Francis X. Canning

A Highly Parallel Multigrid-Like Method for the Solution of the Euler Equations
Ray S. Tuminaro

Fast Parallel Iterative Solution of Poisson's and the Biharmonic Equations on
Irregular Regions
A. Mayo and A. Greenbaum

Compact Multigrid
Victor Pan and John Reif

Parallel Performance of Domain-Decomposed Preconditioned Krylov Methods for PDEs
with Locally Uniform Refinement
William D. Gropp and David E. Keyes

A Large, Sparse, and Indefinite Generalized Eigenvalue Problem from Fluid
Hans D. Mittlemann, Cindy C. Law, Daniel F. Jankowski, and G. Paul Neitzel

Conjugate Gradient-Type Methods for Linear Systems with Complex Symmetric
Coefficient Matrices
Roland W. Freund

Row Projection Methods for Large Nonsymmetric Linear Systems
R. Bramley and A. Sameh

A set of New Mapping and Coloring Heuristics for Distributed-Memory Parallel
Claude Pommerell, Marco Annaratone and Wolfgang Fichtner

Multilevel Filtering Preconditioners: Extensions to More General Elliptic
Charles H. Tong, Tony F. Chan, and C.C. Jay Kuo

Domain Decomposition Algorithms for Indefinite Elliptic Problems
Xiao-Chuan Cai and Olof B. Widlund

Preconditioning Second-Order Elliptic Operators: Experiment and Theory
Wayne Joubert, Thomas Manteuffel, Seymour Parter, and Sze-Ping Wong

Fast Iterative Solution of Carrier Continuity Equations for Three-Dimensional
Device Simulation
O. Heinreichsberger, S. Selberherr, M. Stiftinger, and K.P. Traar

The Hierarchical Basis Extrapolation Method
U. Rude

Fourier Analysis of Incomplete Factorization Preconditioners for
Three-Dimensional Anisotropic Problems
June M. Donato and Tony F. Chan

Line Iterative Methods for Cyclically Reduced Discrete Convection-Diffusion
Howard C. Elman and Gene H. Golub

An Optimal Domain Decomposition Preconditioner for the Finite Element Solution
of Linear Elasticity Problems
Barry F. Smith

An Unconventional Domain Decomposition Method for an Efficient Parallel
Solution of Large-scale Finite Element Systems
Charbel Farhat and Francois-Xavier Roux

Domain Decomposition Methods for Problems with Partial Refinement
James H. Bramble, Richard E. Ewing, Rossen R. Parashkevov, and Joseph E. Pasciak


End of NA Digest