Sensitivity-driven co-synthesis of distributed embedded systems

Abstract

Describes a new, sensitivity-driven algorithm for the co-synthesis of real-time distributed embedded systems. Many embedded computing systems are distributed systems: communicating periodic processes executing on several CPUs/ASICs connected by communication links. We use performance estimates to compute a local sensitivity of the design to process allocation. We propose a priority prediction method to schedule processes. Based on these techniques, we develop a gradient-search algorithm which co-synthesizes heterogeneous distributed systems of arbitrary topology and the associated application software architecture. Experimental results show that our algorithm can find good implementation architectures in small amounts of CPU time.