francis@hanauma.Stanford.EDU (Francis Muir) (02/28/91)
The following Advice recently appeared as a posting, and I offer a
commentary after the quote:
Postdoctoral researchers sought for developing lattice gas
models on a connection machine at the Naval Research
Laboratory. NRL is developing a lattice gas model for wave
propagation, specifically for ocean acoustics. A preliminary
2-d model has simulated reflection, transmission and scattering
processes accurately. A 3-d model is under development. The
work is unclassified and publication is encouraged. Much of
the development is being done jointly with the Center for
Nonlinear Studies at Los Alamos.
For more information on the technical program contact:
Dr. Susan K. Numrich, Code 5150
Naval Research Laboratory
Washington, DC 20375-5000
(202) 404-7345
numrich@cmsun.nrl.navy.mil
Is this connection machine a Connection Machine? A CM1 or CM2 (Logical
or Floating-point machine)? Is the Lattice Gas modeling the Navy has in
mind particle mechanical? Statistical mechanical (Boltzmann)?
Linearised Boltzmann? Are not the problems the Navy has in mind really
elastic and anisotropic (e.g. the difficult muddy bottom case)?
The casual reader of this group should be aware that there are two
quite distinct types of automata, the Lifelike and the lifelike. The
former ruled by Chaos, and the latter by Conservation. Building
Automata to model physical processes usually modeled by PDE's makes
sense and it has an attraction all it's own, but be aware that it will
not be for those drawn to the mysteries of Chaos. Modeling
conservative schemes with particles is quite instructive, but hardly
cost-efficient. The new class of massively parallel floating-point
machines (or special purpose probability machines) seem to have
relegated particle methods to the dark ages and Academia.
On the other hand, and using a diminutive CM2 with no Data Vault and
Dark Ages (256K) memory chips, I can still get 20 turnarounds a morning
on my 2- and 3-D wave modeling in heterogeneous media using a
linearised Boltzmann scheme. This makes conceptual debugging a delight,
and is largely due to the excellent implementation of an extended
(FORTRAN90-like) compiler, and the Frame Buffer - a graphics monitor
hard-wired to the CM's plane of processors.
Cheers
Philomath
UUCP: ...!decwrl!hanauma.stanford.edu!francis
INTERNET: francis@hanauma.stanford.edu
This coloured counterfeit that thou beholdest,
vainglorious with the excellencies of art,
is, in fallacious syllogisms of colour,
nought but a cunning dupery of sense;
Juana de Asbaje y Rodrigues,
Sor Ines de la Crus