Resource management in dataflow

  • Authors:
  • A. J. Catto;J. R. Gurd

  • Affiliations:
  • Departamento de Computacao e Estatistica, Universidade Federal de Sao Carlos, Caixa Postal 384, 13560 Sao Carlos SP, BRASIL;Department of Computer Science, University of Manchester, Oxford Road, Manchester M13 9PL, ENGLAND

  • Venue:
  • FPCA '81 Proceedings of the 1981 conference on Functional programming languages and computer architecture
  • Year:
  • 1981

Quantified Score

Hi-index 0.00

Visualization

Abstract

Recent proposals for nondeterministic facilities in high-level dataflow programming systems have stopped short of giving details of low-level implementation. The underlying machine is assumed to provide basic nondeterministic operations which lead to the required high-level effects. This paper gives details of a practical implementation of one such high-level language, Id {3}, for a specific dataflow computer, the Manchester prototype {11}. It adds to previous work by the authors {7, 8} in which implementations of Communicating Processes {12} and Distributed Processes {5} were proposed. Id is based on an unravelling dataflow interpreter which closely resembles the labelled token scheme used in the Manchester prototype. Thus translation of Id programs into suitable machine code is relatively straightforward. However, instead of requiring complex nondeterministic operators to support resource managers as in {1}, the existing simple matching functions of the Manchester system {8} prove to be adequate. For the non-specialist reader, the Manchester labelled dataflow schema and the resource management constructs of Id are outlined before details of implementation are given.