DMCA
RCL: Reduce learnt clauses (2013)
Cached
Download Links
BibTeX
@MISC{Audemard13rcl:reduce,
author = {Gilles Audemard and Jean-marie Lagniez and Lakhdar Saïs},
title = {RCL: Reduce learnt clauses},
year = {2013}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
RCL is a SAT solver which includes all the modern enhancements of the DPLL procedure as they can be found in solvers such as RSAT [4] and MINISAT [2]. These enhancements include watched literal to the unit propagation, first-UIP learning scheme, frequent restarts (Luby strategy [3]), activity-based decision heuristics (VSIDS), and the phase learning policy is used [4]. The main improvement consists in storing the binary clauses in an adjacency list, and in reducing the learnt clauses. These learnt clauses are reduced by resolution with the binary clauses. For reducing learnt clauses database, clauses are sorted by using the phase. This step allows to associate a weight with each clause. This weight is the number of falsified literals. Once the clauses are sorted, half of them are kept. For not too large instances, we use SatElite as a pre-processor [1]. 2
Keyphrases
binary clause learnt clause large instance adjacency list learnt clause database main improvement unit propagation luby strategy sat solver activity-based decision heuristic first-uip learning scheme modern enhancement frequent restarts phase learning policy dpll procedure falsified literal
