Power and Energy Management in Coarse-Grained Reconfigurable Systems: methodologies,automation and assessments

In the era of Cyber-Physical Systems (CPS), designers need to cope with several constraints that have to be met at the same time. CPS are complex systems composed of different interactive and deeply intertwined components that have to change their behavioural modalities according to several factors as the environment status, requests from user and even their internal status, thus requiring high flexibility and performance, possibly with a low power consumption. The spectrum of existing computing systems ranges from general purpose to application specific systems. General purpose systems as CPUs, GPUs, DSPs offer high flexibility but are not able to provide high performance, due to their poor specialization. On the other side, Application Specific Integrated Circuits (ASICs) offer high performance but they do not provide flexibility at all, being designed for computing a single, specific application. In the middle between general purpose systems and ASICs lie the reconfigurable systems that provide a valuable solution to challenge simultaneously different requirements. Reconfigurable systems offer a certain level of flexibility, while guaranteeing high performance. However, two major issues still limit their wide applicability: high design complexity, implying huge engineering effort, as well as power inefficiencies. The activities behind my thesis address both these issues, with the primary focus on power consumption. The starting assumption is the definition of a set of strategies that, depending on the considered scenario and the chosen target device (ASIC or FPGA), may enable power/energy awareness and consumption optimization. In parallel, these strategies have been automated within different extensions of a dataflow to hardware design suite for coarse-grained reconfigurable systems.