The Demeter Seminar meets in the fall quarter from 3.15-5 on Monday
afternoon in 206 or 306 Egan Center.

Today in 306 Egan.
============================================================

We are exploring connections between the TOP-C parallel programming model
and COOL and other abstractions used in AP.

		Low Performance Parallel Computing
		         Gene Cooperman

Much of the early effort in parallel computing was devoted to high
performance parallel computing:  using complex software constructs
in conjunction with specialized, expensive hardware.  More recently,
there has been much effort in making parallel computing easy to carry
out on cheap, common hardware, such as networks of workstations (NoW's)
and multi-processor shared memory architectures (SMP).  The title
is meant to contrast with the earlier, high-performance efforts.

I will discuss a philosophy that stresses task-oriented parallel
programming.  I refer to the model as the TOP-C (Task-Oriented Parallel C)
model, although implementations also include LISP/MPI, and GAP/MPI (GAP =
Groups and Programming).  The TOP-C model provides a higher level
programming abstraction than message-based programming or shared memory
programming.  The higher level abstraction makes it easier to write many
parallel programs.  It has also been adapted to include many features.
Some of the features are: small command set; easy tracing for debugging;
latency hiding; ease of checkpointing; robustness as processors are added
and removed; meta-computing over the Internet; easy implementation on top
of existing interactive languages (as well as on top of standard C, and
related languages); and re-targetability on top of standard parallel
systems (e.g.: MPI, PVM, POSIX threads).  Finally, the nicest feature is
that by building on top of other systems, the TOP-C model has been
implemented in only a few thousand lines of code.

======================

Mira Mezini and Karl Lieberherr:
A study of interactions between the building blocks of AP
Rondo proposes a better way to separate and connect visitors or
adjustments, strategies and class graphs.

Karl Lieberherr:
Environmental acquisition and AP

In an OOPSLA '96 paper, Gil and Lorenz introduce the concept of
environmental acquisition. Adaptive programming has some good answers
to dealing with environmental acquisition.

======================

view graphs 9-50

Viewgraphs are in
http://www.ccs.neu.edu/research/demeter/course/f97/lectures/powerpoint

lec10.ppt

http://www.ccs.neu.edu/research/demeter/course/f97/lectures/postscript

lec10-color-2-per-page.ps

=============================================================

Planned in future weeks:
Josh Marshall and Binoy Samuel: Introduction to RMI