Automatic distribution of reactive systems for asynchronous networks of processors

Abstract

The paper addresses the problem of automatically distributing reactive systems. We first show that the use of synchronous languages allows a natural parallel description of such systems, regardless of any distribution problems. Then, a desired distribution can be easily specified, and achieved with the algorithm presented here. This distribution technique provides distributed programs with the same safety, test, and debug facilities as ordinary sequential programs. Finally, the implementation of such distributed programs only requires a very simple communication protocol ("first in first out" queues), thereby reducing the need for large distributed real time executives.