To My Family Summary

Embedded systems are becoming increasingly complex and have tight competing con-straints in terms of performance, cost, energy consumption, dependability, flexibility, security, etc. The objective of this thesis is to propose design methods and tools for supporting the tradeoff analysis of competing design objectives during the early design phases, which are characterized by uncertainties. We consider safety-critical real-time applications modeled as task graphs, to be implemented on distributed heterogeneous architectures consisting of processing elements (PEs), interconnected by a shared com-munication channel. Tasks are scheduled using fixed-priority preemptive scheduling, and we use non-preemptive scheduling for messages. As a first step, we address the problem of function-to-task decomposition. In this con-text we have assumed that the application functionality is captured by a set of functional blocks, with different safety requirements. We propose a Genetic Algorithm-based metaheuristic to solve the function-to-task decomposition problem. Our algorithm also decides the mapping of tasks to the PEs of a distributed architecture and the reliabil-ity of each PE in the architecture, such that the safety and integrity constraints ar