NA Digest Sunday, August 13, 1989 Volume 89 : Issue 31

Today's Editor: Cleve Moler

Today's Topics:


From: Peter Mayes <>
Date: Mon, 7 Aug 89 17:05 BST
Subject: NAG Replies to Proskurowski

We would like to thank Professor Proskurowski for his comments on the
NAG multigrid routine D03EEF and its documentation.

Firstly a few words on the philosophy behind the example programs to be
found in the NAG Library manual. These programs are really intended to
give a simple example of how to set up a problem, call the routine and
interpret the results. They are not designed to show the full
capabilities of any particular routine, nor are they intended to show
the behaviour of the routine on particularly "difficult" or "special"
examples. In addition, they are not intended to test the implementation
of the underlying numerical algorithm. For this we have a separate set
of "stringent" test programs.

Turning to the particular routine D03EEF, the example chosen has a
quadratic solution, with zero truncation error for central differences.
This has the unintentional advantage that it gives a demonstration of
the correctness of the implementation in this case. In fact, the results
given by Professor Proskurowski are in good agreement with the multigrid
theory. However, the example has the disadvantage that it gives the
wrong impression about what the user is likely to see from his/her own
problem. It would have been better to give as an example a problem for
which second-order differences gave only second-order accuracy.

We are currently in the process of finalising code and documentation for
Mark 14, and we will change the example to treat a less "special" case.

Thank you very much

Peter Mayes, NAG


From: Harry Nelis <dutentb!>
Date: Wed, 9 Aug 89 14:35:45 met
Subject: Algorithms and Parallel VLSI Workshop

International Workshop on
Algorithms and Parallel VLSI Architectures
June 10--16, 1990
Les Houches, France

Announcement and Call for Papers

Conference Committee

Ed Deprettere Patrick Dewilde
Dept. of Electrical Engineering Dept. of Electrical Engineering
Delft University of Technology Delft University of Technology
email: ed@dutentb email: dewilde@dutentb

Thomas Kailath Sun-Yuan Kung
Information Systems Laboratory Dept. of Electrical Engineering
Stanford University Princeton University
email: email:

Henk van der Vorst Prabhakar Rao
Dept. of Mathematics Dept. of Electrical Engineering
Delft University of Technology Drexel University
email: numan@dutinfd email: rao@drexel

The workshop will consist of a number of courses and workshops.
The following topics will be covered:


Schur's Algorithm and its Applications
Array Forms of Fast Signal Processing Algorithms
Architecture Design
Multiscale Signal Processing
Large Scale Modeling
Solving Large Systems of Linear Equations


Computer Graphics Modeling with Finite Elements
Signal Estimation Adaptive and Orthogonal Filtering
Video Architectures Architectures for Control and Communications
Microcoded Architectures Systolic Architectures
Artificial Neural Nets Eigenvalues and Singular Values

The courses will be given by eminent lecturers in the field (names will be
announced shortly).
The workshops will consist of plenary talks by invited speakers and poster
We plan to have one course and two workshops per day.
Papers are solicited for contributions to the workshops, and will have to be
presented in poster format.
Authors are invited to submit four copies of a 4-page summary of the paper to
the workshop secretariat for review.
Papers will be severely reviewed --- the number of accepted papers is limited.
Authors of accepted papers will be asked to prepare a version for publication.

Authors' Schedule

Submission of 4-page summary: Februari 19, 1990
Notification of Acceptance: March 30, 1990

Workshop Secretariat

Harry Nelis
Dept. of Electrical Engineering
Delft University of Technology
2628 CD Delft, The Netherlands
email: harry@dutentb


From: Zahari Zlatev <>
Date: Wed, 9 Aug 89 11:18:20 -0500
Subject: Address Change for Zlatev

My stay at the Center for Supercomputing Research and Development
at the University of Illinois at Urbana-Champaign is finishing
soon. After August 17 I shall be back in Denmark at the old

Air Pollution Laboratory
Danish Agency for Environmental Protection
Risoe National Laboratory
DK-4000 Roskilde, DENMARK

E-mail address: LFLZL@NEUVM1.BITNET


From: Jim Walker <>
Date: 10 Aug 89 13:21 -0600
Subject: Where is George Marsaglia

Does anyone have the email address and FAX number for George Marsaglia
(random numbers etc.)

Jim Walker.


From: John D. McCalpin <>
Date: 11 Aug 89 12:38:35 GMT
Subject: FFT's for Solving Elliptic Equations

I would like to be able to use an FFT-based method for solving a
finite-difference version of a linear, separable elliptic equation
with Neumann boundary conditions. I routinely use this technique to
solve the same problem with periodic boundaries in one or both

The one unusual aspect of this problem is that the derivative of the
solution is not specified at a grid point, but exactly midway between
two grid points. (This is a natural specification of a derivative for
the finite-difference part of the equation).

It seems that it should be possible to "twiddle" with the transformed
variables to introduce the appropriate phase shift, but I can't figure
out how to do it.

Solutions which are only first-order accurate in space are not
interesting. On the other hand, the value of the derivative at the
boundaries is _almost_ zero, and I may be able to get away with
calling it zero and trying to use a cosine transform -- if I can
figure out how to apply the appropriate phase shift.

Any pointers wil be appreciated!

John D. McCalpin - -
Supercomputer Computations Research Institute


End of NA Digest