Chaff: engineering an efficient SAT solver
Proceedings of the 38th annual Design Automation Conference
Efficient conflict driven learning in a boolean satisfiability solver
Proceedings of the 2001 IEEE/ACM international conference on Computer-aided design
BerkMin: A Fast and Robust Sat-Solver
Proceedings of the conference on Design, automation and test in Europe
Towards understanding and harnessing the potential of clause learning
Journal of Artificial Intelligence Research
Linear-Time Reductions of Resolution Proofs
HVC '08 Proceedings of the 4th International Haifa Verification Conference on Hardware and Software: Verification and Testing
Solving difficult SAT instances using greedy clique decomposition
SARA'07 Proceedings of the 7th International conference on Abstraction, reformulation, and approximation
| Hi-index | 0.00 |
The popular abstraction/refinement model frequently used in verification, can also explain the success of a SAT decision heuristic like Berkmin. According to this model, conflict clauses are abstractions of the clauses from which they were derived. We suggest a clause-based decision heuristic called Clause-Move-To-Front (CMTF), which attempts to follow an abstraction/refinement strategy (based on the resolve-graph) rather than satisfying the clauses in the chronological order in which they were created, as done in Berkmin. We also show a resolution-based score function for choosing the variable from the selected clause and a similar function for choosing the sign. We implemented the suggested heuristics in our SAT solver HaifaSat. Experiments on hundreds of industrial benchmarks demonstrate the superiority of this method comparing to the Berkmin heuristic. There is still room for research on how to explore better the resolve-graph information, based on the abstraction/refinement model that we propose.