Efficient Application Mapping onto Three-Dimensional Network-on-Chips Using Multi-Objective Particle Swarm Optimization

Abstract

Network-on-chip (NoC) is considered the next generation of communication in embedded system. In this case, an application is implemented by a set of collaborative intellectual propriety blocks (IPs). The selection of the most suited block from a library of IPs as well as their physical mapping onto the three-dimensional Network on-chip infrastructure to implement efficiently the application are two NP-complete problems. In this paper, we propose to use Multi-Objective Particle Swarm Optimization (MOPSO) to yield the best selection of IP and there physical mapping of a given application on three-dimensional topologies. In this purpose, MOPSO is exploited to obtain the personalized result for the application at hand. Only the used resources, switches and channels by the application mapping are part of the customized implementation platform. The optimization is driven by the minimization of required hardware area, the imposed execution time and the necessary power consumption of the final implementation.