Chaff: engineering an efficient SAT solver
Proceedings of the 38th annual Design Automation Conference
Width-Based Restart Policies for Clause-Learning Satisfiability Solvers
SAT '09 Proceedings of the 12th International Conference on Theory and Applications of Satisfiability Testing
A lightweight component caching scheme for satisfiability solvers
SAT'07 Proceedings of the 10th international conference on Theory and applications of satisfiability testing
Adaptive restart strategies for conflict driven SAT solvers
SAT'08 Proceedings of the 11th international conference on Theory and applications of satisfiability testing
Diversification and intensification in parallel SAT solving
CP'10 Proceedings of the 16th international conference on Principles and practice of constraint programming
Hi-index | 0.00 |
Restarts and activity based search are two important and correlated components of modern SAT solvers. On the one hand, updating the activity of the variables involved in conflict analysis aims to circumscribe the most relevant part of the Boolean formula. While restarts allow the solver to reorder the variables by focussing the search on this relevant subformula. This combination allows the solver to intensify the search and at the same time helps to avoid trashing. This well known strong connexion between restarts and variable ordering have also a direct consequence on clause learning. The effect of restarts on clause learning have been widely investigated (e.g. [1,5]). Our intuition is that if SAT solvers are able to solve efficiently application instances with millions of variables and clauses, this means that the most relevant part of the formula (or subset of variables) is of reasonable size. This is related to the observation previouslymade on the size of backdoor sets observed on many applications domains. In our previous work, and in the parallel portfolio ManySAT solver, we have shown how the two well known principles of diversification and intensification principles can be combined in the context of Masters/Slaves architecture [2]. The Masters perform an original search strategy, ensuring diversification, while the remaining units, classified as Slaves are there to intensify their master's strategy. By intensification we mean that the slave would explore "differently" around the search space explored by the Master.