NA Digest Sunday, October 15, 1989 Volume 89 : Issue 40
Today's Editor: Cleve Moler
Today's Topics:
-------------------------------------------------------
From: Jack Dongarra and Eric Grosse <dongarra@cs.utk.edu>
Date: Mon, 16 Oct 89 00:23:39 PDT
Subject: netlib@mcs.anl.gov is moving
We are in the process of moving the machine running netlib from Argonne
to Oak Ridge. The machine (a Sequent Balance 8000) was shut down on
Monday, October 9, 1989. We hope to have it back on the air in a
week. During this period, mail to the old Argonne address
netlib@mcs.anl.gov will automatically be forwarded to Bell Labs, so
users should see no disruption in service.
You can get faster service and save a little network traffic by
sending requests directly to
netlib@research.att.com
instead of forwarding through Argonne. The contents of the collection
at Argonne/Oak Ridge and Bell Labs are nearly identical. Interesting
exceptions are the approximation catalog and Fortran-to-C converter at
research.att.com, ANL tools and reports at mcs.anl.gov.
The address for netlib at Oak Ridge National Laboratory will be
netlib@ornl.gov
Mail to the old address netlib@mcs.anl.gov will continue to work
for a month or so by automatically forwarding mail to Oak Ridge.
We will send a message when netlib@ornl.gov is up.
Regards,
Jack and Eric
------------------------------
From: Jack Dongarra <dongarra@cs.utk.edu>
Date: Mon, 16 Oct 89 00:29:49 PDT
Subject: Dongarra's New Address
I am now at Oak Ridge National Laboratory and the University
of Tennessee. I can be reached at the following numbers:
Jack Dongarra
Computer Science Department
University of Tennessee
Knoxville, TN 37996-1301
Phone: 615-974-8295
Internet: dongarra@cs.utk.edu
Fax number 615-974-8296
Regards,
Jack
------------------------------
From: Peter Alfeld <alfeld@science.utah.edu>
Date: Mon 9 Oct 89 10:33:29-MDT
Subject: Positions at University of Utah
The Department of Mathematics of the University of Utah has several
faculty positions open at all ranks. We are particularly interested
in people working in Numerical Analysis or Scientific Computing.
When I mention the above fact to people at conferences I usually
get a blank stare and questions like: Where is Utah? Do they have a
University? We do, and we are on the map!
The Department of Mathematics has about 60 faculty. Our computing
facilities are excellent. Departmental resources include a DEC-20
(which we are in the process of replacing with a Unix based
multiprocessor), a Vax 8600, an Ardent Titan, a network of (public and
private) Sun Workstations, an uncountable number of Macintoshes (I did
try to count them), and an outstanding staff of 5.5 FTE, including the
current president of the national TeX User's Group (Nelson Beebe).
University facilities include an IBM 3090 supercomputer. The
Department ranks in the top group of US Math Departments according to
the classification by the American Mathematical Society.
This is a spectacular place to live, particularly if you like the
outdoors. You can go 500 miles in any direction before reaching a
settlement of comparable size. Some of the world's best skiing
(proposed by the US as the site of the 1998 Winter Olympics) is within
less than an hour driving distance. On the other hand, there are
about a million people within 50 miles North and South of the
University, so you have all the advantages of modern civilization.
Housing is cheap (you can get a 3,000 square foot home close to the
University for under $100,000). A major international airport is
within 20 minutes from the University. The Department is congenial,
and the working atmosphere superb.
Inquire today! I'll be happy to provide any information you may
require. Call or write:
Peter Alfeld
Department of Mathematics
University of Utah
Salt Lake City, Utah 84112
801-581-6842 or 801-581-6851
ALFELD@SCIENCE.UTAH.EDU
------------------------------
From: Alan Edelman <alan@math.mit.edu>
Date: Tue, 10 Oct 89 16:50:14 EDT
Subject: Large Dense Linear System Survey
In reaction to my often being asked whether people really want
to solve large dense systems of equations, I thought I would ask:
Who wants to (or has experience) solving LARGE DENSE linear systems?
Large means n=5000 or greater. Please no mention of SPARSE matrices,
for purposes of this survey.
I wish to put together a survey, hopefully gathering information on
1. In what fields do the solutions of large dense systems of linear
equations arise? Please send a description.
2. For what values of n are solutions currently being obtained, and
what kinds of n would the application really want given faster,
cheaper, computing power?
3. What kind of accuracy is obtained (what precision machine?
conditioning of problem?) What kind of accuracy is desired?
4. References, vague leads to others not directly reachable
via this request, etc., would be appreciated.
Thank you very much
Alan Edelman
alan@math.mit.edu
na.edelman
------------------------------
From: Olof Widlund <widlund@csd23.nyu.edu>
Date: Tue, 10 Oct 89 10:31:57 EDT
Subject: Department Chair Position at NYU
NEW YORK UNIVERSITY
Computer Science Department
Chair Search
A senior computer scientist with outstanding credentials as a researcher as
well as administrative experience is being sought to serve as Chair of the
Computer Science Department at New York University.
The resources of the department have grown substantially in recent years and it
now has thirty-two full-time faculty members and a substantial research staff.
The department is part of New York University's Faculty of Arts and Science
and the Courant Institute of Mathematical Sciences, a major research institute
which has a large computing center and several large research groups in computer
science and applied mathematics. Major funding in parallel computing,
robotics, computer vision, high level programming languages, the Ada language,
software prototyping, natural language processing and numerical analysis is
provided by a number of federal agencies. A laboratory for robotics and
computer vision, funded in part by a CER grant from NSF, opened in 1983.
The university, one of the largest private institutions in the country, is
located in one of the most attractive residential areas of Manhattan.
Resumes should be sent to: Professor Martin Davis, Chair, Department of
Computer Science, New York University, Courant Institute of Mathematical
Sciences, 251 Mercer Street, New York, NY 10012. Internet: davism@cs.nyu.edu.
New York University is an Equal Opportunity/Affirmative Action Employer.
------------------------------
From: Phil Rasch <pjr@cgdra.ucar.edu>
Date: 15 Oct 89 20:39:21 GMT
Subject: Solving PDEs on the Sphere
I am interested in learning of innovative and/or powerful methods for
solving partial differential equations on the sphere that are
relatively unknown in my particular discipline (meteorology).
More precisely, I am interested in solving primarily hyperbolic and
elliptic equations associated with fluid flow and transport in this
geometry, and I want to avoid the following problems.
P1.) Numerical problems associated with the singularities in the
natural coordinate system (spherical polar coordinates) at the pole.
P2.) CFL problems associated with the convergence in the vicinity
of the pole of natural (e.g. latitude, longitude) grids.
P3.) The overshoot/undershoot phenomena (and computational expense at
high resolution) associated with spectral transform methods.
There is quite a large body of work in meteorology on finite
difference (on approximately rectilinear grids) and spectral transform
methods in meteorology, but in my opinion, no really satisfactory
accurate and conservative methods exists which also:
A.) resolve in space in an optimal fashion. This could be construed to
be in an approximatly uniform fashion (both for esthetic reasons, and
to reduce CFL type problems), or in a fashion to resolve the small
scale features.
B.) can be extended to a monotonic or non-oscillatory form.
Rare, or entirely missing in this body of work are methods which use:
1.) triangular meshes: these might include finite element, or control volume
approaches.
2.) numerically generated (and/or) adaptive grids: These might include
also moving to other (possibly) patched curvilinear coordinate systems with
singularities off the sphere.
3.) monotone, total variation diminishing, or essentially non-oscillatory
techniques
4.) Adaptive Lagrangian Eulerian codes.
I would like to hear about particularly successful techniques used
elsewhere which could contribute to solving P1-P3. Solutions which use
spherical geometry are obviously the easiest to map to my applications,
but I would also be interested in untested candidates with potential. I
intend to work on these problems myself, but want to avoid repeating
previous work. It seems likely to me that others in fields like
physics, engineering, aerodynamics, geophysics, and computational
mathematics must have considered these problems, but a comprehensive
search seems difficult.
Thanks in advance for your help.
Phil Rasch
Climate and Global Dynamics Division
National Center for Atmospheric Research
Boulder CO, 80303
(303) 497-1368
Internet: pjr@cgdra.ucar.edu
------------------------------
From: Peter Arbenz <arbenz@inf.ethz.ch>
Date: 11 Oct 89 15:03:20+0100
Subject: CONPAR 90 in Zurich
C O N P A R 9 0
Swiss Federal Institute of Technology ETH
Zuerich, Switzerland
September 10-13, 1990
Sponsored by
Schweizer Informatiker Gesellschaft SI
IEEE Switzerland Section
in cooperation with
Gesellschaft fuer Informatik GI-PARS
British Computer Society BCS-PPSG
Computer Society of the IEEE
Swiss Chapter of the ACM
The past decade has seen the emergence of the two highly successful series
of CONPAR and of VAPP conferences on the subject of parallel processing.
TheVector and Parallel Processors in Computational Science meetings were
held in Chester (VAPP I, 1981), Oxford (VAPP II, 1984) and Liverpool (VAPP
III, 1987). The International conferences on Parallel Processing took place
in Erlangen (CONPAR 81), Aachen (CONPAR 86) and Manchester (CONPAR 88). The
format of the joint meeting will follow the pattern set by its predecessors.
It is intended to review hardware and architecture developments together
with languages and software tools for supporting parallel processing.
Another objective of the conference will be to highlight advances in
algorithms and applications software on vector and parallel architectures.
It is expected that the programme will cover languages/software tools,
hardware/architecture, algorithms/software and applications.
Also special sessions will be devoted to the field of application and/or
programming language specific architectures; i.e. machines, where
performance has been gained through limiting the field of applications, or
systems designed according to a joint optimization of programming language
and architecture.
Other topics of special interest are
* performance analysis for real-life applications
* testing and debugging of parallel systems
* portability of parallel programs
* paradigms for concurrency and their implementation
The conference should appeal to anyone with an interest in the design and
use of vector and parallel machines.
Call for Papers
Original papers are invited for the conference. Five copies of the full
paper (maximum of 10 pages) are to be submitted no later than Feb 1, 1990.
The proceedings of the joint VAPP/CONPAR conference will be published as a
volume in the Springer Lecture Notes in Computer Science series.
Further Information
Prof. Dr. Helmar Burkhart
Institut fuer Informatik
Universitaet Basel
Mittlere Strasse 142
CH-4056 Basel
Switzerland
phone: +41 61 449967
e-mail: burkhart@urz.unibas.ch
------------------------------
From: Dan Sorensen <sorensen@rice.edu>
Date: Fri, 13 Oct 89 11:58:00 CDT
Subject: Panel on Scientific Computing Curriculum
SC89 Panel Discussion on Scientific Computing Curriculum
The preliminary program of the SC89 meeting did not contain a description
of the panel discussion on a University Curriculum in Scientific Computing
that is to be held during session 11A at 4:30 pm on Thursday 16 Nov.
Here is a description of that panel discussion.
PANEL: University Education in Scientific Computing
The purpose of this panel will be to discuss the nature of a University
curriculum in Scientific Computing. The goal will be to outline the
structure of such a program. The panel will attempt an overview of
some programs that are already in place and then discuss additional
concepts that might be incorporated to improve the existing programs.
One goal of the panel discussion is to produce an outline of a "model"
curriculum that might be used to initiate new Scientific Computing
programs at those institutions that do not already have one in place.
For example, a program in scientific computing might contain the core
material for a Masters Degree. Recipients of this degree would be
prepared to bring the full potential of modern computing capabilities
to the industrial research environment. In addition to this master's
degree track, there could be courses and graduate seminars designed to
prepare Ph.D. students for beginning thesis work. A major component of
the program would be a central one semester course designed to appeal
to all the sciences which now rely on computing. The intent would be
to offer a unified approach to computing which would prepare a student
to take a selection of special topic courses. The primary course would
be a prerequisite for the subsequent courses, but they could be
somewhat independent of each other and be drawn in many cases from
existing courses with minor alterations to content. This would give
the program more flexibility to incorporate students from science &
engineering. The topics courses would foster sound computational
science techniques, introduce state of the art numerical techniques,
cover concepts in high-performance computing and promote the use of
visualization and graphics. Hands-on experience with state of the art
super computers and parallel computers, perhaps culminating with a
significant project, would be a major component of this program.
FORMAT:
Each panelist will give brief description of their program in
Scientific Computing and will express their view of what such a program
should consist of. Then there will be a discussion among the panel
members about the various proposals. Finally, the discussion will be
opened to audience participation.
The entire session will be recorded and a written report will be
produced based upon the content of the discussion. The report will be
distributed with the intent of fostering further interaction and
eventually some recommendations from the scientific community.
PANELISTS:
Prof. Geoffrey Fox
Physics Department
California Institute of Technology
Pasadena, California 91125
gcf@procyon.caltech.edu
Prof. Gene Golub
Computer Science Dept.
Stanford University
Stanford, CA 94305
golub@patience.stanford.edu
Prof. R. J. Plemmons
Mathematics Department
North Carolina State University
Raleigh, North Carolina 27695-8205
plemmons%matple@ncsuvx.ncsu.edu
Prof. Garry Rodrigue
Dept. of Applied Science
Univ. of California, Davis
PO Box 808
Livermore, California
rodrigue@lll-crg.llnl.gov
Prof. Ahmed Sameh, Assoc. Director
Center for Supercomputer Research and Development
University of Illinois
305 Talbot Lab, 104 S. Wright St.
Urbana, Illinois 61801
sameh@s12.csrd.uiuc.edu
Prof. D.C. Sorensen (Chair)
Department of Mathematical Sciences
P.O. Box 1892
Rice University
Houston, Texas 77251
sorensen@rice.edu
ABSTRACTS:
PHYSICAL COMPUTATION AND COMPLEX SYSTEMS (PCCS)
Prof. Geoffrey Fox
California Institute of Technology
This is a new Ph.D. program within the Physics option whose objective
is a unified approach to abstraction, modeling, and computation applied
to the natural world. The program involves fundamental education in
mathematical physics, simple classical and quantum physical systems,
fundamental properties of complex systems, physical optimization
methods, and the appropriate computational techniques needed for
solving large scale scientific problems. Physical Computation and
Complex Systems gives a thorough training in basic physics, but unlike
traditional physics programs, these physical techniques are applied to
topics in computation and complex systems. PCCS has been motivated by
the success and growing importance of two multidisciplinary research
thrusts, namely, computational science and complex systems. These two
themes are synergistically linked by their reliance on high performance
computers.
SCIENTIFIC COMPUTING AND COMPUTATIONAL MATHEMATICS AT STANFORD
Prof. Gene Golub
Director, Scientific Computing/ Computational Math, Stanford University
For the last three years we have had a new program in place for
granting degrees in Scientific Computing and Computational Mathematics
at Stanford. The Program is an interdisciplinary program and can admit
students and grant degrees. The Faculty is made up of faculty from
other departments; we currently reside in the School of Engineering. I
shall describe the requirements of our program and its scope.
SCIENTIFIC COMPUTING AT NORTH CAROLINA STATE UNIVERSITY
Prof. Robert J. Plemmons
North Carolina State University
There is a selection of programs in Scientific Computing at North
Carolina State University. The Computer Science and Mathematics
Departments both offer joint courses and degree programs with
scientific computing components. In addition, the Colleges of
Engineering and Physical & Mathematical Sciences have a broadly based
interdisciplinary program. The student receives a degree in a specific
department but is required to select certain courses in scientific
computing from the Mathematics and Computer Science Departments. Course
topics here include parallel processing and supercomputing, with
hands-on experience. The courses formerly used the ACRF computing
facilities at the Argonne National Lab, but we now have our own
equipment, including Alliant and Sequent shared memory computers and an
Intel hypercube distributed memory system. A Cray Y-MP is now available
in our Research Triangle Park so we no longer use the NSF
Supercomputing sites. Interest in advanced scientific computing at
North Carolina State University is fostered by our Center for Research
in Scientific Computation, which was established by the Board of
Governors in 1986.
THE NEW ACADEMIC FIELD OF COMPUTATIONAL SCIENCES.
Prof. Garry Rodrigue
Univ. of California, Davis
The University of California at Davis has started a new graduate
educational program in the area of Computational Sciences. The program
is a joint program involving the fields of Applied Science, Chemistry,
and Computer Science. In contrast to Computer Science, the field of
Computational Sciences is mathematically based with Computational
Mathematics, Applied Mathematics, and a new course entitled
Computational Science as its core. The curriculum for students will
include classes in Mathematical Physics, Computer Engineering, Computer
Science, and Computational Mathematics. The main strength of the
program is in its affiliation with the Lawrence Livermore National
Laboratory (LLNL). Students can participate in the ongoing
Computational Science research at LLNL through laboratory sponsored
research assistantships and at the same time pursue their graduate
degree as a full-time student at a U.C. Davis satellite campus located
next to LLNL.
COMPUTATIONAL SCIENCE AND ENGINEERING
Prof. Ahmed Sameh
University of Illinois
The availability of powerful new computers has made it possible to use
computational methods in larger and broader areas of science and
engineering. Yet, making effective use of such computers is difficult,
partly due to the complexity of their architecture and system
software. The extensive knowledge required to use effectively this new
technology can best be provided through the formal education and
training of students in the field of Computational Science and
Engineering. Unlike what is now regarded as traditional computer
science, a CSE program would attempt to focus on the whole computational
process. Such a program has been established at the University of
Illinois within the departments of Computer Science and Electrical and
Computer Engineering. It covers the following topics:
(I) Computer Architecture and Design
(II) System Software and Compiler Technology
(III) Applications Software and Algorithm Development
(IV) Performance Evaluation
------------------------------
End of NA Digest
**************************
-------