Overview of a parallel reduction machine project

Bevan, D. I.; Burn, G. L. and Karia, R. J. (1987). Overview of a parallel reduction machine project. In: PARLE Parallel Architectures and Languages Europe. Volume 1, Parallel Architectures, Eindhoven, The Netherlands, June 15-19, 1987; Proceedings (de Bakker, J. W.; Nijman, A. J. and Treleaven, P. C. eds.), Lecture Notes in Computer Science, Springer-Verlag, New York, pp. 394–413.

DOI: https://doi.org/10.1007/3-540-17943-7_141

Abstract

ESPRIT Project 415 has taken what are considered to be good programming language styles and is developing parallel architectures to support them. Here we describe the part of the project which is developing a distributed memory architecture for functional languages. Designing parallel architectures for evaluating functional languages presents many challenging problems. Firstly a model for the parallel reduction of such languages must be found. An abstract interpretation has been developed which leads to a parallel reduction model. It can be implemented in a compiler so that programs can automatically be annotated with parallelism information. The original COBWEB, a novel distributed memory architecture, is described, along with the conclusions we have drawn from our simulation work. We also briefly describe some of the architectural features of the architecture we are designing to support the parallel reduction model. Many programming languages including functional ones require automatic storage allocation which has to be garbage collected. We present another piece of work from our project which has resulted in the discovery of a distributed reference counting garbage collection algorithm which has very low overheads.

Viewing alternatives

Metrics

Item Actions

Export

You can export this page using these formats Digital Object Identifier - DOI