Coping with Moore’s law (and more): Supporting arrays in state-of-the-art model checkers

Abstract-State-of-the-art hardware model checkers and equivalence checkers rely upon a diversity of synergistic algorithms to achieve adequate scalability and automation. While higher-level decision procedures have enhanced capacity for problems of amenable syntax, little prior work has addressed (1) the generalization of many critical synergistic algorithms beyond bit-blasted representations, nor (2) the issue of bridging higherlevel techniques to problems of complex circuit-accurate syntax. In this paper, we extend a variety of bit-level algorithms to designs with memory arrays, and introduce techniques to rewrite arrays from circuit-accurate to verification-amenable behavioral syntax. These extensions have numerous motivations, from scaling formal methods to verify ever-growing design components, to enabling hardware model checkers to reason about software-like systems, to allowing state-of-the-art model checkers to support temporallyconsistent function-and predicate-abstraction