Implicative Simultaneous Satisfiability and Applications

Abstract

Abstract. This paper proposes an efficient algorithm for the systematic learning of implications. This is done as part of a new search and restart strategy in the SAT solver. We evaluate the new algorithm within a number of applications, including BMC and induction with invariant strengthening for equivalence checking. We provide extensive experimental evidence attesting to a speedup of one and often two orders of magnitude with our algorithm, on a representative set of industrial and publicly available test suites, as compared to a basic version of invariant strengthening. Moreover, we show that the new invariant strengthening algorithm alone performs better than induction and interpolation, and that the absolutely best result is achieved when it is combined with interpolation. In addition, we experimentally demonstrate the superiority of an application of our new algorithm to BMC.

Boosting Verification Scalability via Structural Grouping and Semantic Partitioning of Properties

“…Much prior work has addressed methods to incrementally reuse information across muliple properties to accelerate specific algorithms. E.g., incremental SAT across proofs of different properties [20,21], and reusing verification by-products like invariants [13] and interpolants [24], can accelerate the verification of high-affinity properties.…”

Blocked Clause Decomposition

“…It is based on techniques used in formal equivalence checking of circuits. See [2,13] for a complete list of references, and further the independently derived results in [14,15]. Similar techniques have been used in the context of computing backbones, see for instance [16].…”

“…In practice, several important optimizations are required (see also [13]). First, incremental SAT solving [19] should be used, adding L as fixed formula permanently, but treating the two unit clauses added in line EE5 and EE6 as assumptions [19].…”

Model checking large design spaces: Theory, tools, and experiments

“…prior work has addressed methods to incrementally reuse information across muliple properties to accelerate specific algorithms. E.g., incremental SAT across proofs of different properties[KNPH06,KN12], and reusing verification by-products like invariants[DR17] and interpolants[MS07b], can accelerate the verification of high-affinity properties.…”