Conflict Driven Learning in a Quantified Boolean Satisfiability Solver Lintao Zhang Sharad Malik Department of Electrical Engineering Department of Electrical Engineering Princeton University Princeton University
[email protected] [email protected] ABSTRACT Within the verification community, there has been a recent without learning. Though methods based on traditional DPLL increase in interest in Quantified Boolean Formula evaluation algorithm do not have the memory blow up problem, they (QBF) as many interesting sequential circuit verification require significant CPU cycles and are unable to handle problems can be formulated as QBF instances. A closely related practical sized problems as of now. research area to QBF is Boolean Satisfiability (SAT). Recent A closely related problem to QBF is the well-known Boolean advances in SAT research have resulted in some very efficient Satisfiability problem (SAT). SAT can be regarded as a SAT solvers. One of the critical techniques employed in these restricted form of QBF. In a SAT problem, only existential solvers is Conflict Driven Learning. In this paper, we adapt quantifiers are allowed. As SAT is very important both in conflict driven learning for application in a QBF setting. We theory as well as in practice, it has attracted significant research show that conflict driven learning can be regarded as a attention. Recent years have seen significant advancements in resolution process on the clauses. We prove that under certain SAT research, resulting in some very efficient complete SAT conditions, tautology clauses obtained from resolution in QBF solvers (e.g. GRASP [12], rel_sat [13], SATO [14], Chaff [15]). also obey the rules for implication and conflicts of regular (non- These solvers are also based on the DPLL algorithm.