Computational Models for Parallel Computers
Members of the project team
Project description
A major reason for the lack of practical use of parallel
computers is the absence of a suitable computational
model for parallel computers (or model of parallel
computation). Many existing models are either overly
simplistic (notably the PRAM) or are tied to a particular
parallel architecture. Models of the first kind do not
correspond closely to the behaviour of existing or
conceivable architectures. Models of the second kind
yield programs that are not portable.
Recently, a number of parallel computation models have
been proposed that are more reflective of realistic
architectures than the PRAM and also general enough
to model several architectures. However, in our
opinion none of these models accurately reflects
all resources of a parallel computer.
The goal of this research project is to develop a model
of parallel computation which can be used to design
fast and portable parallel software. We will explore
which features a model should possess in order to meet
these requirements, and to which design techniques the
use of the model will lead. Also, we will explore which
primitive operations (e.g. broadcast) frequently occur
in practice. It can be advantageous to implement these in
hardware.
Recent Publications
- B.H.H. Juurlink and H.A.G. Wijshoff,
Experiences with a model for parallel computation,
Proc. 12th Annual ACM Symp. on Principles of Distributed Computing,
August, 1993.
- B.H.H. Juurlink and H.A.G. Wijshoff,
The Parallel Hierarchical Memory Model,
Proc. 4th Scandinavian Workshop on Algorithm Theory,
1994, LNCS 824.
Last modified on July 2, 1996 by Lex Wolters.