Control-Based Clause Sharing in Parallel SAT Solving

Hamadi, Youssef; Jabbour, Saïd; Saïs, Lakhdar

doi:10.1007/978-3-642-21434-9_10

Cited by 42 publications

(48 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…As seen so far, scheduling a subset of the constituent solvers of a given portfolio seems to be a good strategy (see for instance the performance of SUNNY, but also [113,92,16,160,151]) . Since often the solving time of hard combinatorial problems is either relatively short or very long, the scheduling approach naturally handles the heavy tailed nature of solving.…”

Section: Sequential Time Splitting and Bound Communicationmentioning

confidence: 99%

See 1 more Smart Citation

Portfolio approaches in constraint programming

Amadini

2015

Constraints

View full text Add to dashboard Cite

Recent research has shown that the performance of a single, arbitrarily efficient algorithm can be significantly outperformed by using a portfolio of -possibly onaverage slower-algorithms. Within the Constraint Programming (CP) context, a portfolio solver can be seen as a particular constraint solver that exploits the synergy between the constituent solvers of its portfolio for predicting which is (or which are) the best solver(s) to run for solving a new, unseen instance.In this thesis we examine the benefits of portfolio solvers in CP. Despite portfolio approaches have been extensively studied for Boolean Satisfiability (SAT) problems, in the more general CP field these techniques have been only marginally studied and used. We conducted this work through the investigation, the analysis and the construction of several portfolio approaches for solving both satisfaction and optimization problems. We focused in particular on sequential approaches, i.e., single-threaded portfolio solvers always running on the same core.We started from a first empirical evaluation on portfolio approaches for solving Constraint Satisfaction Problems (CSPs), and then we improved on it by introducing new data, solvers, features, algorithms, and tools. Afterwards, we addressed the more general Constraint Optimization Problems (COPs) by implementing and testing a number of models for dealing with COP portfolio solvers. Finally, we have come full circle by developing sunny-cp: a sequential CP portfolio solver that turned out to be competitive also in the MiniZinc Challenge, the reference competition for CP solvers.iii iv

show abstract

Section: Sequential Time Splitting and Bound Communicationmentioning

confidence: 99%

“…Recalling that in this work we focus only on sequential approaches, we would however mention some portfolio-based parallel SAT solvers like ManySAT [92], PeneLoPe [16] and ppfolio [160]. A parallel CP portfolio solver built on top of Numberjack platform [98] attended the MiniZinc Challenge 2013.…”

Section: 21)mentioning

confidence: 99%

Portfolio approaches in constraint programming

Amadini

2015

Constraints

View full text Add to dashboard Cite

show abstract

“…Because shorter clauses represent stronger constraints limiting the length of shared clauses by a constant (8 in [14]) would be a reasonable and very simple heuristic. The problem is that as the search continues the length of the clauses tends to increase, reducing the throughput of shared clauses [13]. Tarmo therefore by default shares all clauses whose length is below the running average, and this default is used in all results presented in this work.…”

Section: Clause Sharingmentioning

confidence: 99%

“…Restricting the number of learnt clauses shared between solving threads is therefore an important aspect of parallel SAT solving (see, e.g. [13]). It was stated in the introduction that incremental SAT solving "allows the solver to reuse the information it has gathered for consecutive problems".…”

Section: Clause Sharingmentioning

confidence: 99%

Asynchronous Multi-core Incremental SAT Solving

Wieringa

Heljanko

2013

Tools and Algorithms for the Construction and Analysis of Systems

View full text Add to dashboard Cite

Abstract. Solvers for propositional logic formulas, so called SAT solvers, are used in many practical applications. As multi-core and multi-processor hardware has become widely available, parallelizations of such solvers are actively researched. Such research typically ignores the incremental problem specification feature that modern SAT solvers possess. This feature is, however, crucial for many of the real-life applications of SAT solvers. Such applications include formal verification, equivalence checking, and typical artificial intelligence tasks such as scheduling, planning and reasoning. We have developed a multi-core SAT solver called Tarmo, which provides an interface that is compatible with conventional incremental solvers. It enables substantial performance improvements for many applications, without requiring code modifications. We present the asynchronous interface, a natural extension to the conventional solver interface that allows the construction of efficient application specific parallelizations. Through the asynchronous interface multiple problems can be given to the solver simultaneously. This enables conceptually simple but efficient parallelization of the solving process. Moreover, an asynchronous solver is easier to run in parallel with other independent tasks, simplifying the construction of so called coarse grained parallelizations. We provide an extensive experimental evaluation to illustrate the performance of the proposed techniques.

show abstract

“…As a result, it has received significant research attention [Du et al 1997;Mézard et al 2002], and numerous solver algorithms have been proposed and improved (e.g., by analyzing the structure of each instance [Hamadi et al 2009]). Since evaluating all these techniques requires the generation of hard satisfiable [Grégoire et al 2009], linear programming [Zimmermann and Monfroglio 1997], or tabu search [Nonobe and Ibaraki 1998].…”

Section: Introductionmentioning

confidence: 99%

Satisfiability by Maxwell-Boltzmann and Bose-Einstein Statistical Distributions

Angione

Occhipinti

Nicosia

2015

ACM J. Exp. Algorithmics

View full text Add to dashboard Cite

Recent studies in theoretical computer science have exploited new algorithms and methodologies based on statistical physics for investigating the structure and the properties of the Satisfiability (SAT) problem. We propose a characterization of the SAT problem as a physical system, using both quantum and classical statistical-physical models. We associate a graph to an SAT instance and we prove that a Bose-Einstein condensation occurs in the instance with higher probability if the quantum distribution is adopted in the generation of the graph. Conversely, the fit-get-rich behavior is more likely if we adopt the Maxwell-Boltzmann distribution. Our method allows a comprehensive analysis of the SAT problem based on a new definition of entropy of an instance, without requiring the computation of its truth assignments. The entropy of an SAT instance increases in the satisfiability region as the number of free variables in the instance increases. Finally, we develop six new solvers for the MaxSAT problem based on quantum and classical statistical distributions, and we test them on random SAT instances, with competitive results. We experimentally prove that the performance of the solvers based on the two distributions depends on the criterion used to flag clauses as satisfied in the SAT solving process.

show abstract

Control-Based Clause Sharing in Parallel SAT Solving

Cited by 42 publications

References 15 publications

Portfolio approaches in constraint programming

Portfolio approaches in constraint programming

Asynchronous Multi-core Incremental SAT Solving

Satisfiability by Maxwell-Boltzmann and Bose-Einstein Statistical Distributions

Contact Info

Product

Resources

About