Meeting Agenda
Wednesday, February 21, 1996
- Location:
- DC 1304
- Time:
- 12:30 PM
- Chair:
- Rick Kazman
1. Adoption of the Agenda - additions or deletions
2. Coffee Hour
- Coffee hour this week:
- ???
-
- Coffee hour next week:
- ???
3. Next meeting
- Date:
- February 28, 1996
- Location:
- DC 1304
- Time:
- 12:30 PM
- Chair:
- Gilles Khouzam
- Technical presentation:
- Ryan Gunther
4. Forthcoming
- Chairs:
- John Kominek
- Rob Kroeger
- Iain Little
- Tech Presenters:
- Fabrice Jaubert
- Rick Kazman
- Gilles Khouzam
5. Technical Presentations
- Presenter:
- Saar
Friedman
- Title: Visualization application builders
- Abstract:
-
Visualization is the process of extracting information from data via graphical
representation. Currently the best commonly availabe, general purpose vis.
systems are application builders based on the dataflow model. I'll give
a brief introduction to these using AVS as an example, and then move to talk
about some shortcomings and possible solutions/improvements.
6. General Discussion Items
7. Action List
- Richard, Rob: Side FX/U of T visit; March 2nd 1996
8. Director's Meeting
9. Seminars
CSC Talk
Wednesday, 21 February, 1996
16:30
MC 4060
CORBA: A Standard for Distributed Objects
by
Mike McCool
Abstract
Distributed objects under CORBA are object-oriented "components" that are
language and location independent. They are an abstraction that hides
the network layer, so applications can be written using a collection of
distributed objects as if they were all local objects.
CORBA is a standard which allows the abstract definition of the interface
to an object to be mapped into multiple languages, and which defines a
set of runtime services that provide location independence, persistence,
locking, naming, introspection, and so forth.
In this talk, I will outline the basic concepts in CORBA, and give a
small example of its use.
**********************************************************************
COMPUTER SCIENCE SEMINAR
-Thursday, Feb. 22, 1996
Charles Clarke, Dept. Comp. Sci., University of
Waterloo, will speak on ``An Algebra for Structure Text
Search.''
TIME: 4:00-5:00 p.m.
ROOM: DC 1302
ABSTRACT
Under the shortest substring model for text search, the
solutions to a user's query are the shortest intervals
of the target text that satisfy the query conditions.
This talk describes the application of this search
model to structured text databases, relevance ranking
of documents for information retrieval, and pattern
matching in flat text.
Structured text search is based on an algebra for
combining and manipulating sets of text intervals. The
algebra includes operators for expressing containment
relationships, ordering relationships and combination
relationships. The algebra has been implemented as
part of a distributed structured text database system
developed at the University of Waterloo. The database
permits storage and search of documents in a mixture of
formats, including documents formatted under SGML or
HTML, and less formally structured documents such as
netnews articles or email messages. Properties of the
algebra provide a novel and efficient technique for
ranking the results of a query by likely relevance to
the user. In practice, this technique has compared
favorably with standard techniques for relevance
ranking. Finally, we look briefly at the application
of the shortest substring model to searching flat text
files with a regular-expression-based version of the
search algebra.
**********************************************************************
COMPUTER SCIENCE COLLOQUIUM SERIES
-Tuesday, February 27, 1996
David Lorge Parnas, Software Engineering Research
Group, Department of Electrical and Computer
Engineering, McMaster University, Hamilton, Ontario,
Canada L8S 4K1, will speak ``Software Engineering: An
Unconsummated Marriage''.
TIME: 4:00-5:00 p.m.
ROOM: DC 1302
ABSTRACT
Although the first of many conferences on "Software
Engineering" was held in Munich nearly three decades
ago, communication between those who study software and
those who work as Engineers has not been effective.
Today, the majority of Engineers understand very little
of the "science of programming". On the other side,
the scientists who study programming understand very
little about what it means to be an Engineer, why we
have such a profession, how the profession is
organised, and what Engineers learn during their
education. In spite of this mutual ignorance, today's
Engineers spend much of their time writing and using
software, and an increasing number of people trained in
Computer Science or Mathematics pontificate about "what
Engineers do".
This talk attempts to explain each field to the other
and to suggest why and how the two groups must learn to
work together. The educational and societal
implications of a more interactive marriage will be
discussed.
**********************************************************************
SCIENTIFIC COMPUTATION SEMINAR
-Friday, February 23, 1996
Dr. Marie-Gabrielle Vallet, Institut des Materiaux
Industriels, CNRC, Boucherville, Quebec, will speak on
``Anisotropic Mesh Adaptation, Applications to 2D Flow
Computations''.
TIME: 3:30-4:30 p.m.
ROOM: DC 1304
ABSTRACT
The goal of mesh adaptation is to increase the ratio
(accuracydiscretization based on a mesh. It involves
both an posteriori local error estimator and a mesh
----------
optimization algorithm.
In the case of external flow computation, the gain is
significant because the computational domain presents
some large regions where the solution is smooth and
some small ones where it has rapid variations.
Moreover, directional features arise in Computational
Fluid Dynamics, typically shock waves and boundary
layers, which are better discretized using stretched
elements. In order to obtain stretched adapted meshes
the error estimator and the optimization technique must
both be sensitive to anisotropic features.
The originality of the proposed approach is that the
error is estimated along the edges instead of inside
the elements. An adapted mesh is then defined as a mesh
having nearly the same error on each edge. The mesh
optimizor performs successive local modifications
tending to equally distribute the estimator on edges.
It combines refinement and coarsening with reconnection
and relocation of the nodes.
When coupled with the adaptive algorithm, a fluid
solver becomes mesh independent because mesh and
solution are computed together. Through results on
several steady flow test-cases, it will be demonstrate
that, as far as the mesh is adapted to the solution,
- absolutely no artificial viscosity is needed if the
discretization scheme is stable (LBB condition),
- the final solution does not depend on the scheme
nor on the method involved (Finite Element or
Finite Volume).
**********************************************************************
ICR Seminar
Parallelism and C++: Maybe the Unknowns Will Cancel Out
by: Dr. Gregory V. Wilson, Research Scientist
of: Center for Advanced Studies
IBM Canada Ltd.
North York, Ontario
Date: Tuesday, March 5, 1996
Time: 2:30 pm.
Place: William G. Davis Computer Research Centre, Room 1304
Abstract:
C++ is fast becoming the most widely used systems programming
language. It is also starting to be employed by scientists and
engineers for numerical programming. At the same time, the most
powerful high-performance computers are now multiprocessors with
either shared or distributed memory. While extensions for For-
tran programming on such platforms have emerged, there is no con-
sensus on how, or even whether, to extend C++ to handle and/or
disguise such things as multiple address spaces, asynchronous ex-
ecution, symmetric multiprocessing, and communication topology.
This talk will compare and contrast the pros and cons of several
different approaches, as exemplified by a variety of current pro-
ducts and research projects. It will also describe in detail the
way these problems are addressed in ABC++, IBM's parallel class
library product.
**********************************************************************
10. Lab Cleanup (until 1:30 or 5 minutes)