An improvement in decomposed reachability analysis for symbolic model checking

Abstract

Even though BDD-based reachability analysis can handle many large designs, the BDD sizes still often explode, and its performance is hard to predict. To address this problem, several decomposed approaches are presented in [6] to perform approximate state space traversal on a partitioned design. These approaches use reachable states as invariants to constrain the inputs of each partition to reduce unreachable states. This paper presents an improvement such that the input behavior of each partition is constrained by the state transitions allowed by the source partitions where the inputs are defined. Experimental results show that this type of input constraints helps to produce tighter state space for each partition. Furthermore, the effect of design partitioning on the performance is also discussed.