Intermediate uniformly distributed tuple space on transputer meshes

Abstract

Linda is a paradigm for high-level parallel programming that offers a novel alternative to conventional models of process coordination and communication. The basis of the Linda model is a global, logically shared tuple space through which process interaction and synchronization are achieved, providing a conceptually simple approach to parallel programming. There is evidence that suggests a need for a new programming methodology to support transputer-based applications, and Linda, as an attractive and elegant alternative to existing methodologies, has great potential for this role. This paper describes the implementation of a particular tuple space model, intermediate uniform distribution, on transputer meshes and discusses the nature of the overheads inherent in the implementation. It is concluded that although the specific tuple space model is not ideally suited to transputer-based systems and the implementation, as it stands, is too inefficient to be of practical use, the approach requires further exploration in order to exhaust its full research potential.