NA Digest Sunday, April 26, 1992 Volume 92 : Issue 17

Today's Editor:

Cleve Moler
The MathWorks, Inc.

Submissions for NA Digest:

Mail to

Information about NA-NET:

Mail to


From: David Bailey <>
Date: Tue, 21 Apr 92 10:46:38 -0700
Subject: Multiprecision Computation Translator

Some of you may be aware of a multiprecision computation package and a
multiprecision Fortran translator that I have developed during the
past year or two. I am now pleased to announce that this software has
been approved by NASA Ames for unlimited international distribution,
at no cost. A brief description of this software is as follows:

The first package contains approximately 85 Fortran subroutines that
perform a variety of arithmetic operations and transcendental
functions. The basic arithmetic routines are very efficient, and for
extra high levels of precision (about about 1000 digits), some
additional routines are available that employ advanced algorithms,
such FFT-based multiplication and the Borwein transcendental
algorithms, for additional speedup. On some problems these routines
are over 200 times faster than certain other multiprecision tools.

However, as with other such packages, manually converting a program to
use these routines is a tedious and error-prone process. To
facilitate such conversions, I have now developed a multiprecision
translator for Fortran programs. By means of source directives
(special comments) placed in a standard Fortran-77 program, the user
declares the precision level and specifies which variables in each
subprogram are to be treated as multiprecision. The translator
program reads this source program and outputs a program with the
appropriate multiprecision subroutine calls, which can then be linked
with the multiprecision package and executed.

This translator program supports multiprecision integer, real and
complex datatypes. The required array space for multiprecision data
types is automatically allocated. In the evaluation of computational
expressions, all of the usual conventions for operator precedence and
mixed mode operations are upheld. Furthermore, most of the Fortran-77
intrinsics, such as ABS, MOD, NINT, COS, EXP are supported and produce
true multiprecision values.

Both the multiprecision package and translator are written in highly
portable Fortran-77 and have been successfully implemented on a
variety of systems, ranging from parallel supercomputers to

If anyone is interested in obtaining this software, plus the
accompanying documentation, please send me a note. I eventually
intend to place this software on public media such as netlib, but for
the time being I prefer to keep a list of users, so that any bug fixes
or enhancements can be promptly distributed.

David H. Bailey
NASA Ames Research Center


From: Eildert Groeneveld <>
Date: 23 Apr 92 14:42
Subject: Multivariate Optimization

We are performing multivariate (dimensions of 60 and more) numerical
optimization on functions which are very costly to evaluate: each
evaluation might take 30 minutes CPU time or more. So far we have used
some Downhill Simplex and QuasiNewton procedures. However, there may be
recent algorithms out there that I am not aware of. Thus, I am looking
for an optimization algorithm that meets the following conditions:

1. very efficient in regard to number of function evaluations
2. able to recover from non smooth function values (indicating a set
of function parameters that are not defined)
3. constrained or unconstrained optimization.

Efficiency is of great importance, but also robustness. Any pointers would be
greatly appreciatated.

Address: Eildert Groeneveld
Institute of Animal Husbandry
and Animal Behaviour
Federal Agricultural Research Centre (FAL)
W-3057 Neustadt 1
Hoeltystr. 10
Tel : 05034 871 155
Fax : 05034 871 143
e-mail :


From: Tina Flores <>
Date: Tue, 21 Apr 92 15:15:20 EST
Subject: SIAM Meeting Advance Registration Deadlines

Just a friendly reminder.....

Deadlines for advance registration:

Fourth SIAM Conference on Optimization ------ May 4, 1992
Sixth SIAM Conference on Discrete Mathematics-- June 1, 1992

For your convenience, you can register either by phone, FAX
or e-mail. Call 215-382-9800. FAX 215-386-7999. E-mail
VISA, MC and AE credit cards are welcomed. Contact SIAM now
and furnish the following information: registrant's
name, affiliation and address; name of conference; SIAM
member or non-member; type of credit card, number and
expiration date.

We look forward to your presence at these conferences.

Thank you.

Society for Industrial and Applied Mathematics


From: Herbert Muthsam <A8131DAA%AWIUNI11.BITNET@AEARN.EDVZ.UNI-Linz.AC.AT>
Date: Wed, 22 Apr 92 09:12:20 MEZ
Subject: Position at University of Vienna

University of Vienna/AUSTRIA

Institute of Mathematics

The Institute of Mathematics invites applications for a newly created
tenured full professorship in computational mathematics.

Requirements include successfull scientific activity in the development
resp. analysis of computer-aided solution methods for mathematical,
scientific or technological problems. The successful candidate will be
ready to enhance pertinent teaching activities and cooperate with
neighbouring scientific disciplines at the university. Knowledge of the
German language is necessary.

Candidates should send applications (including curriculum vitae,
publication list, etc.), to: University of Vienna, FN Dekanat, Karl
Lueger-Ring 1, A-1010 Vienna, Austria. Applications should arrive
before June 15, 1992.


From: Jack Dongarra <>
Date: Thu, 23 Apr 92 16:45:29 -0400
Subject: Contents: Journal of Supercomputer Applications

International Journal of Supercomputer Applications

Table of Contents Volume 6.2 summer 1992

The Frankenstein Project: Building a Man in the
Machine and the Arrival of the Computational Physician
M. Witten

Protein Sequence and Structure Comparison on Page 138
Massively Parallel Computers
by R. Jones Thinking Machines Corporation

Parallel Computation for Medicine and Biology: Page 147
Applications of Linda at Yale University
D.F. Sittig, M.A. Shifman, P. Nadkarni, P.L. Miller
Yale University

Computational Methods for Bone Mechanics Studies Page 164
R.T. Hart, Z.M. Oden, S.W. Parrish, *D.B. Burr
Tulane University, *Indiana University

Solid Modeling on a Massively Parallel Processor Page 175
By D. Strip, M. Karasick Sandia National Laboratories,
IBM T.J. Watson Research Center

History Matching for Multiphase Reservoir Models Page 193
on Shared Memory Supercomputers
by J. Zhu, Y.M Chen Mississippi State University


End of NA Digest