NA Digest Sunday, July 8, 1990 Volume 90 : Issue 24

Today's Editor: Cleve Moler

Today's Topics:

-------------------------------------------------------


From: G. W. Stewart <stewart@cs.UMD.EDU>
Date: Fri, 6 Jul 90 08:19:23 -0400
Subject: Winners of Householder Prize

This year the Householder Prize for the best thesis in numerical
algebra was awarded to two people. They were

Maria Elizabeth Go Ong
Thesis Title: Hierarchical Basis Preconditioners for Second Order
Elliptic Problems in Three Dimensions
Advisor: Loyce Adams
Insitutuion: Applied Math. Dept., University of Washington

and

Alan Edelman
Thesis Title: Eigenvalues and Condition Numbers of Random Matrices
Advisor: Nick Trefethen
Insitutuion: MIT

I hope you will join with me in congratulating them on their
accomplishment. The competition was fierce. Sixteen theses were
Submitted, all of high quality and several of them outstanding. With
new people like this coming up, the field is in no danger of
stagnation.

The prize committee consisted of Chandler Davis, Beresford Parlett,
Axel Ruhe, Pete Stewart (chair), and Paul Van Dooren. Pete Stewart
will rotate off the committee and his place will be taken by Olof
Widlund. Beresford Parlett will be the new chairman.

Pete Stewart


------------------------------

From: Jerzy Wasniewski <JERZY%YALEVM.BITNET@Forsythe.Stanford.EDU>
Date: Wed, 27 Jun 90 07:48:18 EDT
Subject: More About Choleski

I checked in the Polish Armorial. The name "Cholewski" is
a polish name. His family emigrated from Poland to France.
His ancestry were good polish patriots. The correct
writing is

C H O L E W S K I

The pronansiation can be express in english

H O L E S K Y

Best regards,
Jerzy


------------------------------

From: John G. Aspinall <JGA@ALDERAAN.SCRC.Symbolics.COM>
Date: Fri, 29 Jun 90 17:11 EDT
Subject: Error Bounds for Tensor Splines

Has anyone generalized the Hall & Meyer results ("Optimal Error Bounds
for Cubic Spline Interpolation" J Approx Th. 16 (1976)) to tensor splines?

(reply to me, please, don't swamp the list; I'll provide a summary for
anyone who wishes)
John Aspinall, Symbolics Inc.


------------------------------

From: Dave Dodson <convex!dodson@uxc.cso.uiuc.edu>
Date: Tue, 5 Jun 90 16:33:51 -0500
Subject: Re: "Scaling, Underflow, and Rank Deficiency" and "Extending the BLAS"

I would like to respond to Roger Grimes' criticism ("Extending the BLAS,"
NA Digest, Volume 90: Issue 19, May 13, 1990) of Cleve Moler's submission
("Scaling, Underflow, and Rank Deficiency," NA Digest, Volume 90: Issue 18,
May 6, 1990).

Moler proposed creating another Basic Linear Algebra Subprogram, DSCALD,
which scales a vector by dividing by a scalar, in contrast to the existing
BLAS subprogram DSCAL which scales a vector by multiplying by a scalar.
Moler gave ample justification for such an addition and I agree with him
that it would be advantageous to extend the Basic Linear Algebra Subprograms
to include such a capability. I agree with Grimes that Moler did not do a
good job of integrating his proposal into the BLAS framework. However,
I don't think Grimes did an adequate job of stating his case, which would
have been stronger had he given a specific criticism and a detailed
counterproposal. This submission is intended to correct these shortcomings.

The problem Grimes was addressing is that the family of BLAS subroutines
_SCAL can have either a one- or a two-character prefix. The allowed
prefixes are S, D, C, and CS. Two additional prefixes, Z and ZD, are
common nonstandard extensions applicable to those computer systems with
Fortran compilers that can handle a double precision complex data type.
The point Grimes was trying to make is that you cannot add a suffix
character D to a name such as CSSCAL without stepping outside the bounds
of the current ANSI Fortran Standards, which allow names only as long as
six characters.

I would like to propose an extension to the BLAS that was co-developed by
John Lewis and me. The operation is called "vector reciprocal scaling"
and has names of the form _RSCL with the same prefixes allowed as for SCAL.
We purposely chose a root name which could be abbreviated in four characters
that would not begin with S, D, C, or Z. Thus, the acceptable usages are
as follows:

Purpose:

Given a real or complex scalar a and a real or complex
vector x of length n, these subprograms perform the
reciprocal vector scaling operation

x <-- x / a

Usage:

INTEGER N, INCX
REAL A, X(LENX)
CALL SRSCL (N, A, X, INCX)

INTEGER N, INCX
DOUBLE PRECISION A, X(LENX)
CALL DRSCL (N, A, X, INCX)

INTEGER N, INCX
COMPLEX A, X(LENX)
CALL CRSCL (N, A, X, INCX)

INTEGER N, INCX
REAL A
COMPLEX X(LENX)
CALL CSRSCL (N, A, X, INCX)

INTEGER N, INCX
DOUBLE COMPLEX A, X(LENX) ! the declaration the compiler allows
CALL ZRSCL (N, A, X, INCX)

INTEGER N, INCX
DOUBLE PRECISION A
DOUBLE COMPLEX X(LENX) ! the declaration the compiler allows
CALL ZDRSCL (N, A, X, INCX)


Entry: N The number of elements of vector x to be used in the
scaling operation. If N .LE. 0 the subprograms do not
reference X .

A The scalar a.

X Array of length LENX = (N-1)*ABS(INCX)+1 containing the
input x vector.

INCX Increment for the array X, INCX .GT. 0. The vector x
is stored forward in array X with stride INCX; i.e.,
x[i] is stored in array element X((i-1)*INCX+1).


Exit: X If N .LE. 0 then X is unchanged. Otherwise, x / a
replaces the input.


Notes: A divide-by-zero error may occur if A .EQ. 0.

The subroutines need not compute the result by division, but
will not produce any overflows or underflows if x[i] / a never
overflows or underflows. (This allows the usually faster
code suggested by Moler to be used.)


------------------------------

From: Daniel Howard <NAHOWARD%VAX.OXFORD.AC.UK@Forsythe.Stanford.EDU>
Date: Wed, 27 Jun 90 12:30 GMT
Subject: ASE 91 Conference

CALL FOR PAPERS
ASE 91 Conference
on
Application of Supercomputers
in Engineering

Sponsored by Wessex Institute of Technology and NSF (pending)


BOSTON, MASSACHUSETTS
August 13-15 1991

ORGANIZATION AND EDITORIAL COMMITTEE:

Dr. Carlos Brebbia,
Director of Wessex Institute of Technology,
Southampton, UK

Professor Avi Lin,
Department of Mathematics,
Temple University and ICOMP NASA, USA

Dr. Daniel Howard,
Numerical Analysis Group,
Oxford University and Rolls Royce PLC, UK

Dr. Ing. Alex Peters
IBM Deutschland GMbH
Heidelberg, West Germany


INTERNATIONAL SCIENTIFIC ADVISORY COMMITTEE:

Dr. R. D. Ciskowski, IBM Kingston, USA
Prof. D. J. Evans, U. of Technology, Loughborough, UK
Prof. Dr. W. Gentzsch, FM Regensburg, West Germany
Dr. F.K. Hebeker, IBM Heidelberg, West Germany
Dr. R. Hiromoto, Los Alamos National Laboratory, USA
Prof. A. Jameson, Princeton University, USA
Prof. Y. Paker, Poly. of Central London, UK
Mr. E. Schnepf, Siemens AG, Munich, West Germany
Dr. T. Takeda, Ohbayshi Corporation, Tokyo, Japan


CONFERENCE THEMES:

The first conference, ASE 89, took place in Southampton University and
resulted in the publication of two volumes of proceedings. At that conference
the emphasis was mainly on the impact of supercomputer architectures on the
engineering community and on the meaning of benchmark tests for these
computers.

The themes of ASE 91 will have a stronger emphasis on parallel algorithms
for the efficient solution of partial differential equations, on examples of
large scale computation which have had an impact on an engineering design,
as well as on hardware and software aspects of supercomputing which result
in more efficient indirect memory addressing, ie. for numerical solution
methods on unstructured grids. Invited speakers on these and other relevant
topics will be disclosed in a future announcement.

Contributors should consider four main subject categories when submitting
an abstract:

(1) New and better algorithms for parallel engineering computation:

(a) multigrid and vector extrapolation schemes
(b) conjugate gradient methods
(c) operator split and domain decomposition
(d) the mathematics of parallel computation
(e) Finite and Boundary Element algorithms

(2) Examples of engineering applications on vector and parallel computers:

(a) structural dynamics, rock and ice mechanics
(b) fatigue, impact, and crash simulations
(c) Computational Fluid Dynamics, heat transfer and combustion
(d) turbulence and environmental modelling
(e) shallow water equations
(f) soil mechanics
(g) CAD and CIM interacting with Finite/Boundary Elements

(3) Three dimensional visualisation of engineering problems using the latest
algorithms, hardware configurations and distributed systems.

(4) We would like to see hardware manufacturers, industrialists and applied
academic researchers tackling the debate on the merits of unstructured,
multiblock and body fitted grids for solution of discrete engineering
problems by the engineer. `Philosophical' papers from people with a
wealth of practical experience are welcome under this subject category.
Papers dealing with hardware or software approaches to an efficient
indirect addressing memory management in current/future supercomputers
will be particularly welcome.

Papers are invited on the topics outlined above and on other topics which
will fit within the general scope of the conference.

TIME SCHEDULE:

Submission of abstracts: 1st November 1990 (deadline)
Preliminary acceptance: 15th December 1990
Submission of final paper: 5th April 1991
Final acceptance: 17th May 1991
Conference venue: 13th-15th August 1991

ABSTRACTS should be no longer than 300 words and should clearly state
the purpose, results and conclusions of the work to be described in the
final paper. Final acceptance will be based upon review of the full
length paper.

ALL ABSTRACTS must be submitted to the Conference Secretary:

Liz Neuman,
Conference Secretary,
W.I.T., Ashurst Lodge
Ashurst, Southampton,
SO4 2AA, England, UK

Tel: 44-703-293223
FAX: 44-703-292853

For further information on ASE 91 please contact the Conference Secretary
above. Specific conference theme queries may be addressed to the following:

E-mail: avilin@edu.temple.math.euclid or howard@na.ox.ac.uk


------------------------------

From: Raphaele Herbin <raphaele@masg1.epfl.ch>
Date: 27 Jun 90 18:39 -0600
Subject: Chair at University of Grenoble

The Univerity Joseph Fourier (Grenoble 1) and the Control Data Company
have created a chair of Industrial Mathematics in October 1989.
Applications are invited for high level specialists of computational
mathematics and numerical simulation. Salary is about $45000 a year.
Candidates from University or Industry are encouraged to apply.
This is a six month or one year appointment.
For further information please contact :
Jacques Blum
Laboratoire de Modelisation et Calcul
BP 53X
F-38041 Grenoble Cedex (France)
tel (33) 76 51 48 60
or
E. Tournier, e-mail : tournier@frgren81.bitnet
To apply, send curriculum vitae to :
M.A. Nemoz, President of the Joseph Fourier Univerity,
BP 53XF-38041 Grenoble Cedex (France)



------------------------------

From: David Kahaner <kahaner@cs.titech.ac.jp>
Date: 3 Jul 90 10:56 +0200
Subject: List of Reports on Scientific Computing in Japan

All except those noted are available by electronic mail.
Questions should be addressed to me, electronically if possible. For
other correspondence use the address below.
From within the USA...
Dr. David K. Kahaner
Office of Naval Research, Far East
APO SF 96503-0007
From other locations...
Dr. David K. Kahaner
Office of Naval Research, Far East
Roppongi 7-23-17
Minato-ku, Tokyo 106 Japan
Tel: (03) 401-8978, Fax: (03) 403-9670

Apr 11 12:16 trprpt.290
(1) Trip report 20-24 Feb 1990, Hiroshima and Ehime Universities
(2) Brief reports:
9 Feb, Quantum Magneto Flux Project
13-14 Feb, UTokyo
15 Feb, Hitachi Central Research Lab
26 Feb, Nat Res Lab Metrology, Tsukuba
1 Mar, Parallel computing meeting, Keio University

Apr 11 12:19 quality.sw
The Second International Workshop on Software Quality Improvement
was held in Kyoto, Japan, from 22-24 January 1990. The workshop was
jointly organized by Professor Torii (Osaka University) and
Professor Basili (University of Maryland). The workshop is
summarized from the perspective of a numerical analyst. The major
conclusions are that some Japanese companies are better at managing
the software development process than comparable U.S. companies and
that more interaction is needed between numerical and nonnumerical
software developers.

Apr 11 12:23 trp12-89
Accompanying Professor Gene Golub, Computer Science Department,
Stanford University to Institute of Statistical Mathematics, PAX
computer project at University of Tsukuba, IBM's Tokyo Research
Laboratory, Kyoto University's Research Institute for Mathematical
Sciences, and Ryukoku University.

Apr 11 12:26 speech.etl
Summary of the report, Researches of the Electrotechnical Laboratory,
Number 905, December 1989. "Studies on a Vocal Tract Model for
Speech Synthesis and Analysis" (126 pages including 124 item
bibliography) by
Hiroshi Ohmura
Speech Processing Section
Machine Understanding Division
Electrotechnical Laboratory

Apr 11 12:52 sna90
The First International Conference on Supercomputing in Nuclear
Applications (SNA90) was held from 12-15 March 1990 in Mito City,
Japan. This paper summarizes the significant presentations and
vendor exhibits associated with Japanese software and hardware.

Apr 11 14:56 jifip (Not available electronically)
Titles and authors of March 1990 meeting of Japan Information
Processing Society (more than 1000).

Apr 12 08:55 pax
The PAX parallel processing project developed at the University of
Tsukuba by T. Hoshino is summarized. PAX is a two dimensional array
of processors which operate in MIMD form. QCDPAX has a fast floating
point unit, specially programmed for QCD calculations. Peak
performance is 12.5 Gigaflops.

Apr 25 10:01 crc
This report illustrates the breadth of applications and research
areas of a "service bureau" organization, Century Research
Corporation (CRC), Tokyo.

Apr 25 13:52 scj90
Supercomputing Japan 90, 27-29 March 1990, Tokyo, Japan, and visit
to NEC to see SX/3 supercomputer.

Apr 25 15:46 fujitsu.3d
Fujitsu's 3D movie generated partially via CAP 256 parallel computer.

Apr 28 11:31 anritsu
Commercialization of PAX by Anritsu Corporation.

May 1 14:55 yosh
Report by H. Yoshihara on Japanese Supercomputer Performance on
specific computational fluid dynamics benchmark program and
comparison with Cray.

May 9 10:55 isdn
Papers on ISDN (Integrated Service Digital Network).

May 24 10:09 j-supers.590
List of supercomputers in Japan.

May 24 15:22 ohbayashi
Computer Modeling in the Construction Industry--
Ohbayashi Corporation's Research Lab

May 30 09:30 sony
An overview of Sony's Computer Science Laboratory (CSL) is given.
CSL specializes in research on object-oriented programming and
distributed operating systems.

Jun 4 12:22 ibm-nic
The IBM Tokyo Numerically Intensive Computation Center is reviewed.
This center mainly supports users who wish to develop vectorized
versions of programs on IBM's 3090 VF computers. The Tokyo Research
Laboratory performs longer term research in computer science,
mathematics, graphics, and languages.

Jun 5 11:33 pp90
Titles of papers presented at Information Processing Society of
Japan Symposium on Parallel Processing '90, 17-19 May 1990.

Jun 14 09:57 atr
The Advanced Telecommunication Research Institute (ATR)

Jun 21 13:39 japgovt
A brief outline of Japanese Government ministries and agencies is
shown with an emphasis of those that support science and technology.
Several major projects related to computing are described.

Jun 26 10:33 nipt
A proposed new MITI national project "New Information Processing
Technology" (NIPT) merging soft information processing with massive
parallelism is described. The proposed project is also compared to
related work in other countries. I am told that the project, in some
form, has a good chance of being supported in the near future.

Jul 2 07:36 10ssymp
The tenth meeting of Software Symposium, held in Kyoto Japan 6-8
June 1990 is briefly summarized. The most interesting aspects were
discussions by the Chinese of their national project.

Jul 2 07:37 etl
A visit to the Data Flow project at ETL is summarized. The Dataflow
SIGMA-1 computer project is ending. The new project EM-4, will have
1,024 processors and is designed to have less overhead. EM-4 was
originally proposed for symbolic rather than numeric computation,
but the designers now feel that with the inclusion of floating point
hardware it will also be used for numerical computation. Currently
an 80 processor version of EM-4 is running at 997 MIPs.

Jul 2 09:53 spice
We review research at Toshiba in vectorizing the circuit simulation
program SPICE, and research at NEC at building a special purpose
multiprocessor for circuit simulation modeling.



------------------------------

End of NA Digest

**************************
-------