NA Digest Sunday, March 11, 1990 Volume 90 : Issue 10

Today's Editor: Cleve Moler

Today's Topics:


From: Karl Gustafson <gustafs@boulder.Colorado.EDU>
Date: Tue, 6 Mar 90 13:34:54 MST
Subject: IMACS Boulder Conference Dates

Conference dates for the IMACS conference in Boulder were
inadvertently omitted in last week's NA Net announcement.


University of Colorado,Boulder, Colorado 80309-0426

JUNE 11-15, 1990

For full announcement, email to:


From: Joerg Waldvogel <WALDVOGE%CZHETH5A.BITNET@Forsythe.Stanford.EDU>
Date: Wed, 7 Mar 90 10:14 GMT
Subject: Pronunciation of Cholesky, Lanczos and Euler

The comments on A.L. Cholesky in the NA Digest 90-7 were very interesting
indeed. The only thing missing were hints on the correct pronunciation of
the name Cholesky. Here are some (I am using English phonetics in quotes
to indicate pronunciations):

1) The approximate original Polish or Russian pronunciation is 'Kholesky',
as the name would also be read in German phonetics. Therefore this is
an appropriate pronunciation for Slavic and German, perhaps Spanish
speaking countries.

2) The guttural sound kh does not occur in French or English. There were
(are) two ways to overcome this problem: (a) Best approximation in the
set of native sounds. This results in 'Colesky' for both the French
and the English language. Emigres sometimes adapt the spelling later
on, but apparently the Cholesky family did not do this. (b) Reading
the foreign name in native phonetics, using the original spelling.
This would result in 'Sholesky' for French, the language spoken by
Andre Louis Cholesky. I do not know how Cholesky pronounced his name.

3) In view of this the straight-forward English pronunciation 'Cholesky'=
'T-sholesky' should not be used.

Do you know about any other difficult but important names? Here are two:
The numerical Analyst C. Lanczos is widely mispronounced. The name is
Hungarian and should be read as 'Lantsosh'. On the other hand, it is now-
adays well known that Leonhard Euler is pronounced (but not spelled) as
'Oiler', and not as 'You-ler'.
Joerg Waldvogel, ETH Zurich


From: Are Magnus Bruaset <>
Date: 5 Mar 90 11:22:20 GMT
Subject: Software for Sparse Eigenvalue Problem Wanted

As part of a project on PCG methods used on model problems with discontinous
or highly varying material coefficients I want to compute the eigenvalue
distribution of a sparse matrix. So far I have used MatLab, CLAM and LINPACK
but these packages do not take advantage of the sparsity in my problem, thus
allowing only small systems to be investigated.

I hope someone can guide me to where I can get software that solves
A x = lambda B x, utilizing that both A and B are pentadiagonal matrices
with real entries.

Any ideas will be highly appreciated!

Are Magnus Bruaset (
University of Oslo, Norway


From: David Bernholdt <>
Date: 5 Mar 90 19:34:55 GMT
Subject: Functions to Modify IEEE Floating Point

I use Sun systems, which follows the ANSI/IEEE 754-1985 standard for
floating point arithmetic. Sun provides a number of functions to
manipulate the actions of the system. For example, the Fortran
and the include file <f77/f77_floatingpoint.h>

They also provide things like
which can turn on or off the standard IEEE handling of underflows in
order to save comuptation time, and
which can be called anywhere to provide a summary of the IEEE
exceptions which have occurred so far.

Are these routines available on all machines which follow IEEE
floating point? Do they have the same names or just similar
functionality? Unfortunately, I don't have access to any other IEEE
machines to look at. Any help is appreciated.

David Bernholdt
Quantum Theory Project bernhold@ufpine.bitnet
University of Florida
Gainesville, FL 32611 904/392 6365


From: David Kahaner <>
Date: Mon, 5 Mar 90 09:15:50 +0900
Subject: News Reports from Japan

I have a new e-mail address is a new account and it seems to be reliable.
Please use to contact me instead of any
other you may have.

In doing my job here in Tokyo I write reports on the trips that I make and
the meetings that I attend. Several of these might be of interest to
readers of NA-Digest.
(1) Summary of a trip with Prof Gene Golub (Stanford) to
Institute of Statistical Mathematics (Tokyo)
Research Institute for Mathematical Sciences (Kyoto)
IBM Research (Tokyo)
University of Tsukuba to see PAX project (Tsukuba)
Ryukoku University
(2) Summary of the meeting
Second International Workshop on Software Quality Improvement (Kyoto)
(3) The PAX parallel processing project at University of Tsukuba
(4) Summary of visit to Hiroshima University and Ehime University.

These reports contains descriptive material, references, and my own
reactions. I will be happy to share any or all of these with anyone
who sends me a mail message.

There are also a number of scientific computing related meetings that
will be held in Japan this spring, summer and fall. I can provide some
information on these to anyone who writes to me.

David Kahaner


From: Uri Ascher <>
Date: 5 Mar 90 12:38 -0800
Subject: Report from Conference in Nigeria

The Fourth International Conference on Computational Mathematics
was held at the University of Benin (in cooperation also with
Ondo State University) in Nigeria, February 19-22, 1990.
A workshop on Computers in Education formed part of the program
as well. The first three of these conferences took place in 1983,
1986 and 1988.

The conference is unique in topic and in character for that
part of the world. It also offers visitors from abroad a unique
experience for a scientific meeting. The conference was ably
directed (and at times it seemed single-handedly managed) by
Professor S. O. Fatunla.

About 100 participants attended, mostly from Nigeria. The
list of foreigners included J. Pryce (Swindon, UK), M. van Veldhuisen
(Amsterdam), M. Wulkow (Berlin) and myself (Vancouver). This list
is relatively short as compared to that of 1988, undoubtedly
because of reduced funding for travel outside Nigeria, which
probably reflects the economic slump which this country currently

The impressive opening ceremony was attended by a number of
federal and state officials. Short addresses were delivered by
a representative of the federal Minister of Education, by the
military governor of the state and by the University Vice-Chancellor,
ex-Mathematician Dr. Grace Alele Williams. Professor van Veldhuisen
added his part for the scientists from afar. This ceremony was
followed by a symposium on Computers in Education. Another
symposium the following day was devoted to Computers in Library
Automation. Panels of experts from various Nigerian universities
discussed these important topics in the local context. The
discussions were thorough and the dominant tone was informed
and realistic: the foreign visitors did not have much to add,
just to learn.

In the second half of the conference things got down to what
was to us more familiar grounds. The presentations by the
speakers from abroad, as well as those by a few local scientists,
concentrated on various aspects of the numerical solution of
differential equations.

It has been commented in previous reports on these meetings
(SIAM News March 1986, November 1988) that the experience that
the visiting participant gains transcends well beyond the
usual information gathering in a scientific meeting. The
opportunity of meeting our Nigerian colleagues and their eager
students and of watching them operate on their own turf brings
upon a new perspective in which their efforts and achievements
can be appreciated. One is also led to reflect on all the basic
advantages, taken for granted, which are part of the work
environment in North America and Western Europe.

The trip to Nigeria would not be enjoyed by those who must
have the Hyatt amenities to feel at home. But the Benin University
staff did everything they could to make our stay a comfortable
and enjoyable one, including easing the pains of the Lagos
International Airport. Friendliness, eagerness to please
and to learn and good will by virtually
all colleagues and students whom we met more than compensated
for tentative schedules and bathroom fixtures. Food was OK,
the accomodations spacious and yes, the Nigerian beer is
much better than the ordinary American one and costs almost
nothing in dollars.

Uri Ascher


From: H.J.J. te Riele <>
Date: Wed, 7 Mar 90 16:46:29 +0100
Subject: Symposium Parallel Scientific Computing in Amsterdam



Centre for Mathematics and Computer Science
Kruislaan 413
1098 SJ Amsterdam
The Netherlands

R.J. van der Pas (Convex Computer b.v.) and S.W. Brok (TU Delft),
Generalized Red Black algorithm and its implementation
on the parallel processors DPP84, Alliant FX/8 and Convex C240

S.G. Petiton (Yale Univ., New Haven, USA),
Parallel Subspace Method for non-Hermitian Eigenproblems
on the Connection Machine (CM2)

P.J. van der Houwen and B.P. Sommeijer (CWI),
Parallel methods for ordinary differential equations

W.M. Lioen (CWI),
Solving large dense systems of linear equations
using systems with hierarchical memory

Shun Doi (NEC Corporation, Kawasaki, Japan) and A. Lichnewsky (INRIA, Parijs),
On Parallelism and Convergence of Incomplete LU Factorizations

The organizing committee,
H.J.J. te Riele (CWI, 5924106)
Th.J. Dekker (UvA)
H.A. van der Vorst (TUD)


From: H.J.J. te Riele <>
Date: Thu, 8 Mar 90 14:28:00 +0100
Subject: International Conference on Supercomputing

1990 ACM International Conference on Supercomputing
June, 11-15 Amsterdam

CO-CHAIRMEN: Ahmed Sameh (Illinois) & Henk van der Vorst (Delft and CWI)



Burton Smith, "The Tera Computer System"
Tony Chan, "Parallel Multilevel Algorithms for PDE's"
William Wulf, "The Collaboratory: A Larger Context for Support
of Computational Science"
Wolfgang Fichtner, "Iterative Methods and Supercomputers for
VLSI Device Simulation"
Toshitsugu Yuba, "Dataflow Computer Development in Japan"
Ahmed Noor, "Strategies for Large-Scale Structural Problems on
High Performance Computers"
Piet van der Houwen, "Parallel ODE Solvers"
Anthony Hey, To be announced


About 125 papers have been submitted, covering a large number of
subjects. In summary the issues for the ICS'90 Conference are quite
well in agreement with what we believe to be the current trends and
main problems: tools to aid parallel programming as well as parallel
execution, the effects of memory hierarchy, programming techniques,
visualization, effective use of suitable kernels like the BLAS3,
besides the already more classical topics in this rapidly expanding
area: parallel algoritms, large scale applications on vector
multiprocessors and local memory machines, performance
evaluation, and architectures for special applications.
No more tham two parallel sessions will be scheduled at the
same time.


More information, including fees and an electronic registration form,
can be obtained from Frans Snijders:

Mr. Frans Snijders
P.O. Box 4079, Kruislaan 413
1009 AB Amsterdam, The Netherlands
fax.: +31-20-5924199


From: Caroline Foers <>
Date: Tue, 6 Mar 90 16:17 GMT
Subject: NAG Users Association April Workshop


"Scientific and Technical Computing Languages"

Wednesday 4 April 1990, London

In April 1990, NAGUA will be holding its very first one-day workshop.

The title of the workshop is "Scientific and Technical Computing Languages".
It will be given by senior technical NAG staff. Sessions during the day
will focus on key computing languages, to include Fortran, Ada and C.

NAG is involved in library work in all of these languages. In Fortran,
the release of Mark 14 later this year will expand the library to include
around 160 new routines, as well as offering improved documentation and
redeveloped test programs. Plans already exist for Mark 15.

NAG's Ada Library was developed to exploit all the advantages of the language,
and was not simply a translation of the Fortran routines. The recently
launched C Header Files allow the C community to have direct access to NAG's
Fortran Library and Graphics Library.

The provisional programme for the workshop includes

Introduction, Steve Hague, Deputy Director

Overview of Current Practice in Fortran 77, Richard Brankin

Fortran 8x (Fortran 90?), Malcolm Cohen

C, Shah Datardina

Ada, Graham Hodgson

Open Discussion Session

The workshop will be held at Imperial College, University of London, on
Wednesday 4 April 1990. The registration fee will be #50 for NAGUA members
or #80 for non-members, and this will include lunch and tea/coffee.

Booking forms are available from the following address:
NAG Users Association
PO Box 426
0X2 8SD
United Kingdom
Provisional bookings will be accepted by telephone or email
Telephone: 0865 311102
International: +44 865 311102


From: Andrew Sherman <>
Date: Mon, 05 Mar 90 10:41:53 -0500
Subject: Position at Scientific Computing Associates

Scientific Computing Associates, Inc. has immediate openings on its
technical staff for individuals to carry out a variety of tasks related
to scientific software and parallel computing. Initial assignments may
include research and development for the vector and parallel versions
of PCGPAK2, SCIENTIFIC's Fortran subroutine package for iterative
solution of sparse linear equations, as well as some amount of
technical customer support. In the future, there will be opportunities
to move in several directions within the company, depending on interest
and ability. A background in numerical analysis and scientific
computing equivalent to a master's degree is required, and an
additional 3-5 years of work experience with large software packages
and vector or parallel computing is highly desirable.

SCIENTIFIC is a small company (around 15 employees) that is committed to the
development of leading-edge solutions to computational problems arising
from scientific and engineering applications. A major technical focus
for SCIENTIFIC has been the effective use of vector and parallel
computers. The Company has been the recipient of a large number of
Small Business Innovation Research (SBIR) grants awarded by government
agencies to foster the commercial development and application of new
research ideas. In addition to PCGPAK2, SCIENTIFIC presently markets
three other software products: SMPAK, a set of Fortran subroutines for
sparse Gaussian elimination; CLAM, an interactive environment for
numerical computation and graphics; and LINDA, a set of extensions to
various languages that facilitate portable and efficient programming in
diverse parallel computing environments.

Most computing at SCIENTIFIC is done on a network of Sun SPARCstation 1
workstations. Also on site are a Decstation 3100 workstation and
multi-processor Apollo DN 10000 and Silicon Graphics Personal IRIS
computers. Access to numerous other vector and parallel computers is
available as needed for research and development work.

Salary for these positions will be competitive and will vary depending
on the qualifications of the successful applicants. Scientific
Computing Associates, Inc. is an equal opportunity/affirmative action employer.

Inquiries regarding these positions should be directed to:

Andy Sherman
Manager of Mathematical Software
Scientific Computing Associate, Inc.
246 Church Street, Suite 307
New Haven, CT 06510

Tel: (203) 777-7442
Fax: (203) 776-4074
email: or yale!sca-sun!sherman.


From: William Alexander <>
Date: Tue, 6 Mar 90 13:30:24 EST
Subject: An Interpreted Matrix Language

MATRIX: An Interpreted Matrix Language

MATRIX is a user oriented matrix language interpreter for the entire line of
IBM PC's and true compatibles. MATRIX exists as an environment in which
commands are processed with the basic unit of data being a matrix. There
are over one hundred commands in the command set. These commands may either
be given to MATRIX singly in an interactive mode or a set of commands forming
a program (macro) may be run from a file.

Many statistical, mathematical and engineering techniques can be conveniently
and quickly implemented in MATRIX. It is a tool both for teaching and
research. Many ``what if'' questions can be investigated without the need
to resort to traditional programming languages. MATRIX has the standard
matrix operations such as matrix multiplication, element-wise operations,
inversion, determinant, Cholesky factorization, (symmetric) eigenvalue,
singular value and QR decompositions, generalized inverse, sweep, Kronecker
product, trace, and block. It also features a number of statistical commands
such as random number generation, pdf, cdf, and quantiles from 23 continuous
distributions, kernel regression and density smoothers and fast Fourier
transform. MATRIX also has a great many non-mathematical functions. There are
convenient ways to create and save overlay graphs. These graphs may be
printed on printers using either the Epson graphics or HP Laserjet
command set. Data files may be read and written in both ASCII and binary
formats. MATRIX supports conditional execution of commands by IF/ELSE and
WHILE constructs. There is an extensive on-line help facility and a built-in
screen-oriented editor. Macros can be run directly from the editor, reducing
development time.

Users with an understanding of Microsoft Pascal or C can add their
own commands to the MATRIX command set via the EXTERN command.
Such commands have access to any defined matrix. Hence, MATRIX can be
customized to almost any need.

MATRIX is primarily written in Microsoft Pascal. However, many of the
routines are written in assembly language to maximize their speed.
MATRIX requires at least 384k of RAM to load. More memory will allow the user
to define more matrices. A color graphics adapter (CGA) is required.
MATRIX will run in the CGA mode on machines equipped with an EGA or VGA.
A math coprocessor is recommended, but not required.

For more information, including prices, contact:

Dr. William P. Alexander
Department of Statistics
859 Patterson Office Tower
University of Kentucky
Lexington, KY 40506-0027


End of NA Digest