Reuse and Integration of Heterogeneous Components for Efficient Embedded Software Generation

Nowadays, our world is more and more saturated with computing and communication capability, that is integrated with human users and with the environment. Electronic systems are no more stand-alone entities isolated from the physical reality. On the contrary, electronic systems permeate home automation, health care management, together with every day human life. As a result, modern embedded systems are highly heterogeneous, as they are composed of a mix of analog and digital HW, as well as embedded SW. Furthermore, the tight bound with the physical environment implies to take into account physical evolution during the design and verification phases. In this context, reuse is a very difficult task, as the components to integrate are highly heterogeneous. On the other hand, reuse is a winning approach to save design cost and time. Indeed, top-down approaches allow to optimize and configure each step of design, but any time that a component must be added or changed, the whole design flow must be undergone again. This thesis focuses on three main techniques for supporting reuse and integration in the context of embedded systems: co-simulation, interface generation and homogeneous formal representation. Co-simulation and interface generation preserve the high level of heterogeneity of the system, by allowing communication between heterogeneous components with a framework (co-simulation) or via generation of the necessary interfaces. A computational model is then proposed to model the heterogeneity in a homogeneous way. The starting heterogeneous components are automatically converted to the computational model, with transformations that preserve the starting behavior, thus supporting a fully bottom-up flow. The homogeneous description is then used as a starting point for automatic generation of code for simulation and validation of the integrated system, or for efficient execution as SW on massively parallel architectures. The whole flow is supported by automatic code generation tools, that enhance the effectiveness of the proposed approach.