A special-purpose parallel system for predictable real-time systems

Abstract

Parallel systems usually increase the complexity of evaluating the program execution time, so as to verify timing constraints into real-time environments. For this reason, the approach described in this paper is based on a massively parallel architecture using VLSI to implement in a fine-grain mode the data-flow paradigm. The system employs the static data-flow execution model that does not need to use a control token. A cluster of functional units has been simulated in order to validate the proposed architecture and the first results confirm that the overall behaviour is predictable.<>