Distributed processing network architecture for reconfigurable computing

Abstract

This paper introduces a set of rules and guidelines for the implementation of a Distributed Processing Network (DPN) as the basis for a dynamic reconfigurable architecture targeted at improving the performance of microprocessor (µP) based systems in computationally intensive application domains. In order to provide the computation gains needed to improve upon the performance of the µP, the DPN architecture offers: 1) A low reconfiguration overhead, 2) A simple control interface, 3) Dynamic resource allocation, 4) Concurrent execution with dynamic reconfiguration, 5) Lower power dissipation than a µP executing the same computation kernel and, 6) Scalability to tackle tasks of varying resource requirements. DPN is currently targeted at realtime computationally intensive application domains such as compression, and signal transformations.