Multi-Grain Reconfiguration for Advanced Adaptivity in Cyber-Physical Systems

Abstract

The advent of Cyber-Physical Systems (CPS) is putting designers to the test: during operation, such a kind of systems has to meet multiple and variable requirements coming from the environment, the user or the system itself. Runtime adaptivity allows addressing mutable requirements over time by modifying system behavior, and it can be implemented in hardware leveraging on reconfigurable devices operating at different granularities: mainly coarse- and fine-grain. The former is faster but less flexible (minor modifications are applied) than the latter. For modern CPS an advanced adaptivity support providing both the advantages of the two main hardware reconfiguration approaches could be extremely useful. In this work, a novel approach for multi-grain reconfiguration is presented in order to deliver advanced adaptive hardware accelerators to CPS.