As the complexity of electronic systems increases, designers adopt a re-use methodology where new products are assembled out of components. This is a common trend in many application domains. In consumer electronics, Systems-on-Chips (SoCs) integrate many different cores to provide tens of different functions. In automotive, modern cars rely on a distributed, networked embedded system that comprises hundreds of processors to provide comfort, fuel efficiency and entertainment. In large scale systems, such as avionics and building automation, networked distributed controllers are used to provide comfort, safety and energy efficiency.

Since the system behavior depends not only on the components, but also on the way in which they interact, architecting their interconnection is a critical step in the overall design flow. Being subject to tight performance and cost constraints, the design of the interconnection architecture needs to be tailored to the specific system application. This task is too complex to be done by hand, considering also the heterogeneous nature of these systems. Therefore, there is a need for communication synthesis tools that, starting from a characterization of the communication constraints among the components and the library of available communication building blocks, automatically derive an optimal interconnection architecture.

In this thesis I argue that the essence of the communication synthesis problem is invariant to the application domain. I introduce a formal framework to capture the communication constraints, the library of communication building blocks, and the rules to compose them. Using this framework, I formulate a general communication synthesis problem. I show the generality of the approach by formulating and solving the problem in two different domains: system-on-chips and building automation systems.




Download Full History