Scalable reachability analysis via automated dynamic netlist-based hint generation

Abstract

While SAT-based algorithms have largely displaced BDD-based verification techniques due to their typically higher scalability, there are classes of problems for which BDD-based reachability analysis is the only existing method for an automated solution. Nonetheless, reachability engines require a high degree of tuning to perform well on challenging benchmarks. In addition to clever partitioning and scheduling techniques, the use of hints has been proposed to decompose an otherwise breadth-first fixedpoint computation into a series of underapproximate computations, requiring a larger number of (pre-) image iterations though often significantly reducing peak BDD size and thus resource requirements. In this paper, we introduce a novel approach to boost the scalability of reachability computation: automated netlist-based hint generation. Experiments confirm that this approach can yield significant resource reductions; often over an order of magnitude on complex problems compared to reachability analysis without hints, and even compared to SAT-based proof techniques.