Improving Coordinated SMT-Based System Synthesis by Utilizing Domain-Specific Heuristics

Abstract. In hard real-time systems, where system complexity meets stringent timing constraints, the task of system-level synthesis has become more and more challenging. As a remedy, we introduce an SMT-based system synthesis approach where the Boolean solver determines a static binding of computational tasks to computing resources and a routing of messages over the interconnection network while the theory solver computes a global time-triggered schedule based on the Boolean solver’s solution. The binding and routing is stated as an optimization problem in order to refine the solution found by the Boolean solver such that the theory solver is more likely to find a feasible schedule within a reasonable amount of time. In this paper, we enhance this approach by applying domain-specific heuristics to the optimization problem. Our experiments show that by utilizing domain knowledge we can increase the number of solved instances significantly.