SMT-based optimization for synchronous programs

Abstract

In this paper, we present several optimization techniques to improve the runtime and size of the code generated from synchronous programs. These optimizations work on extended finite state machines (EFSMs) that can be used as intermediate representation for any synchronous system. Our optimizations consists of two phases: First, local optimization guides the EFSM generation and considers the states and edges separately. Second, global optimization is based on a dataflow analysis of the entire EFSM. For both phases, we employ an SMT (Satisfiability Modulo Theories) solver to verify the individual optimization steps. Our experiments show the potential of the presented optimizations: optimized programs generally have a smaller size and a better run-time performance.