- Location:
- DC1304
- Time:
- 12:30 PM
- Chair:
- Randall Reid
1. Adoption of the Agenda - additions or deletions
2. Coffee Hour
- Coffee hour this week:
- Any volunteers?
-
- Coffee hour next week:
- Any volunteers?
3. Next meeting
- Date:
- May 1, 1996
- Location:
- DC 1304
- Time:
- 12:30 PM
- Chair:
- Navid Sadikali
- Technical presentation:
-
Mike McCool/Dan Milgram ??
4. Forthcoming
- Chairs:
-
- Balasingham Balakumaran
- Richard Bartels
- John Beatty
- Tech Presenters:
-
- Thomas Pflaum
- Randall Reid
- Navid Sadikali
5. Technical Presentation
- Presenter:
-
Luiz Henrique de Figueiredo
- Title:
- Surface intersection using affine arithmetic
- Abstract:
-
-
We describe a variant of a domain decomposition method proposed by
Gleicher and Kass for intersecting and trimming parametric surfaces.
Instead of using interval arithmetic to guide the decomposition, the
variant described here uses affine arithmetic, a tool recently proposed
for range analysis. Affine arithmetic is similar to standard interval
arithmetic, but takes into account correlations between operands and
sub-formulas, generally providing much tighter bounds for the computed
quantities. As a consequence, the quadtree domain decompositions are
much smaller and the intersection algorithm runs faster.
6. General Discussion Items
7. Action List
8. Director's Meeting
9. Seminars
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
THEORY SEMINAR
-Wednesday, April 24, 1996
Yuan Ma, Stanford University, will speak on "Optimal
Fault-Tolerant Sorting Networks''.
TIME: 2:30-3:30 p.m. *NOTE TIME*
ROOM: DC 1304
ABSTRACT
Sorting networks have been intensively studied for
several decades, and they have proved to be very useful
for a variety of applications, including circuit
switching and packet routing. With the rapid advance of
computer technologies, the study of the fault-tolerance
properties of sorting networks has gained increasing
importance since the presence of faulty elements is
inevitable in any large system.
In this talk, I will present optimal networks and
parallel algorithms for sorting that work correctly
even when each comparator/comparison is independently
faulty with a constant probability. These results
settle several long-standing open questions in the
literature. Both theoretical and simulation results
will be presented.
Some of the results are joint work with Tom Leighton
and Greg Plaxton.
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
MASTER'S THESIS PRESENTATION
-Friday, April 26, 1996
Ngai-Hung Jimmy Ng, graduate student, Dept. Comp.
Sci., Univ. of Waterloo, will speak on ``Efficient
Intersection Algorithms for Geometric Objects''.
TIME: 1:30-2:30 p.m.
ROOM: DC 3540
ABSTRACT
Recently, the importance of a Spatial Database
Management System (SDBMS) is being recognized. A SDBMS
is a database management system which allows us to
retrieve and manipulate geometric data. QL/G, a query
language for SDBMS, is being developed at the
University of Waterloo. One major contribution of QL/G
will be the design of a number of geometric operators.
These geometric operators are designed for manipulating
geometric objects.
One proposed geometric operator, namely intersection
operator, will be of interest throughout this thesis.
This proposed intersection operator is a generic binary
operator which can be applied on any two geometric
objects. The design and implementation of a number of
efficient intersection algorithms will cover the entire
thesis. The main goal is to make the intersection
operator generic and complete. Other than that, the
robust design of the algorithms allows us to modify and
obtain other geometric operators easily.
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
MASTER'S THESIS PRESENTATION
-Wednesday, April 24, 1996
Ada Ying Dee Cheung, graduate student, Dept. Comp.
Sci., Univ. Waterloo will speak on ``Data Transfer
Using Controlled Compression.''
TIME: 3:30-4:30 p.m.
ROOM: DC 1304
ABSTRACT
In a scenario in which a large amount of data is to be
transferred from one site to another, it is desirable
to optimize the throughput of data transfer. There are
two transmission options. One option is to compress the
data before the transfer, and to decompress them at the
receiving site. The other option is to send the data
without doing any compression. Ideally, we would like
to choose the option that accomplishes the transfer as
quickly as possible. This choice is complicated by many
factors such as the load of the machines, and the
degree of traffic congestion of the network. These
factors affect the transmission rate differently
depending on whether or not we use compression. This
thesis proposes and evaluates two algorithms for
automatically determining whether compression will be
used or not. One algorithm is based on sampling past
performance with and without compression. The other
directly exploits feedback from the network. These
algorithms have been implemented and evaluated under
different conditions. The evaluation shows that both
algorithms can pick the correct mode of data transfer
to use under different environments. The algorithm
based on network feedback is more complicated and
requires tuning to perform well.
DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES
MASTER'S THESIS PRESENTATION
NOTE DATE CHANGE -Thursday, April 25, 1996
Piotr Przybylski, graduate student, Dept. Comp. Sci.,
Univ. Waterloo will speak on ``A Type Based
Implementation for a Language with Distributed Scope''.
TIME: 2:00-3:00 p.m. NOTE TIME CHANGE
ROOM: DC 3301
ABSTRACT
Remote execution is the concept of sending code over
the network to be executed at a remote site. Recently,
much attention has been given to languages supporting
remote execution, such as Java, Telescript and Obliq.
We present the implementation of a language that
combines remote execution and distributed scoping with
parametric polymorphism and strong typing. As support
for code distribution, our implementation uses runtime
type information, and it is the first polymorphic
language using this method. The presented approach
extends the concept of transmitting code and data with
the transmission of type information between address
spaces.
10. Lab Cleanup (until 1:30 or 5 minutes)