Parallel and Distributed Computing
My research spans several areas centered on parallel and distributed computing.
One major focus of the research has been reconfigurable computing,
particularly reconfigurable models and algorithms.
We have proposed an important model, the Reconfigurable Multiple Bus
Machine (RMBM), that reveals the relative powers of segmenting and fusing
buses. We used this to place reconfigurable models in a hierarchy of computing
powers relative to other reconfigurable models and classical models of
computation. Another significant direction of our work has been in identifying
model and algorithmic features that affect the scalability of reconfigurable
models. Our book "Dynamic
Reconfiguration: Architectures and Algorithms" is a
brings together, in one unified framework, the wide
range of models, techniques and algorithms proposed during a decade of research
in the area.
A recent focus has been on reconfigurable hardware and architectures that
facilitate fast partial reconfiguration of devices such as FPGAs.
increased interest in the use of reconfigurable fabrics as integrated
accelerators, this direction of research is important.
We have designed a method that can efficiently select a subset of frames to
reconfigure and rapidly route the configuration bits to the selected frames.
Specifically, we have constructed a that simultaneously enables multiple elements and automatically forms a
scan-path through them (for sending in configuration bits, if the selected elements are frames to be
reconfigured). This work could have application beyond partial reconfiguration, for example in
simultaneous access to multiple (independent) memory locations.
It may also be useful for matching optical and
electrical bandwidths by serializing multiple pieces of data to a single optical waveguide).
This approach could lead to a big improvement over conventional methods that select one element (frame or memory location)
at a time (for reconfiguration or for access). This work has been
In another work on reconfigurable architectures, we have proposed the idea of
combining branching programs (that represent
solutions on bus-based reconfigurable models) and Boolean gates to generate faster circuits. If this
work can be generalized, then it could bring advantages of theoretical reconfigurable models to
Another major thrust of the research involves various aspects of distributed
One direction in this thrust is the study of coordination
problems for swarms of mobile autonomous robots. In particular, we have
examined the problem of arranging the robots in positions that make all robots
visible to each other; this "Complete Visibility" problem is a fundamental
one that leads to solutions to other coordination problems. A key result we
proved is that Complete Visibility can be achieved in constant time
(independent of the number of robots in the swarm). This result is significant
in itself, but perhaps more importantly, our work has introduced the idea of
time as a performance metric; previous approaches for oblivious autonomous
robots only guaranteed algorithm completion in finite time.
Another direction is to use information-theoretic
approaches in distributed computing. For instance, we derived message and time
bounds on the well-studied problem of scheduling access to a shared channel.
Our approach showed a marked improvement over conventional "group-testing"
based approaces. The main intuition
to our result was an observation that it is not necessary to identify all
competitors for the shared channel (as many current solutions do); just
distinguishing them suffices. This approach brings a fresh perspective to a
well-studied, but important, problem.
We are also working on distributed optical mutual exclusion and distributed
algorithms for load balancing on an electrical grid. This work is at a more
Our work is also in other areas
including computational geometry and multiple bus-based interconnection networks.
My current work aslo involves pedagogical research. Recently we received an
NSF grant through which we organized the
"Workshop on Connecting Concepts
Across the Curriculum"
that brought together faculty from across the USA and Canada .
The key contribution of the workshop and subsequent study was to produce a
useful template for reinforcing foundational concepts across a sequence of
courses. We are currently looking to implement some of these ideas in our own
curriculum. A recent
chapter details some initial work in this direction.
I have also served as the program chair for the NSF/IEEE-TCPP Workshop on
Parallel and Distributed Computing Education (EduPar), 2016. I am a lead member
of a multi-university team to update the NSF/IEEE-TCPP Curriculum Guideline for
Parallel and Distributed Computing (PDC) in Undergraduate Education. NSF has
recently funded this activity. We believe that this is very important for
Computer Engineering and Science education, as parallel and distributed systems
are everywhere in modern computing platforms, whereas coverage of these topics
has lagged in most modern undergraduate curricula.
Elaine T. and Donald C. Delaune Distinguished Associate Professor
School of Electrical Engineering and Computer Science
Louisiana State University
Baton Rouge, LA 70803-5901
Phone: (225) 578-5238
Fax: (225) 578-5200