No abstract

Abstract. This paper introduces a novel approach which combines SLS SAT solving and unit propagation to realize fast performance on huge satisfiable uniform random 3-CNF formulas. Currently, applying unit propagation within local search solvers was only beneficial for structured formulas. We show how unit propagation can be combined with SLS solvers to realize speed-ups on random formulas as well. Our novel approach includes techniques of systematic SAT solving, like a variable selection heuristic that originates from a look-ahead solver. Additionally, we developed a scheme using the Cauchy probability distribution to decide when to perform unit propagation. We present our SLS solver Eagle, which is three times faster compared to TNM, the winner of the last SAT competition on satisfiable random formulas. We also show that EagleUP, the enhanced version with unit propagation, experiences an additional speed-up of about 15% to 22%.

No abstract

No abstract