SMT techniques for fast predicate abstraction

  • Authors:
  • Shuvendu K. Lahiri;Robert Nieuwenhuis;Albert Oliveras

  • Affiliations:
  • Microsoft Research;Technical Univ. of Catalonia, Barcelona;Technical Univ. of Catalonia, Barcelona

  • Venue:
  • CAV'06 Proceedings of the 18th international conference on Computer Aided Verification
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

Predicate abstraction is a technique for automatically extracting finite-state abstractions for systems with potentially infinite state space. The fundamental operation in predicate abstraction is to compute the best approximation of a Boolean formula ϕ over a set of predicatesP. In this work, we demonstrate the use for this operation of a decision procedure based on the DPLL(T) framework for SAT Modulo Theories (SMT). The new algorithm is based on a careful generation of the set of all satisfying assignments over a set of predicates. It consistently outperforms previous methods by a factor of at least 20, on a diverse set of hardware and software verification benchmarks. We report detailed analysis of the results and the impact of a number of variations of the techniques. We also propose and evaluate a scheme for incremental refinement of approximations for predicate abstraction in the above framework.