ABSTRACT. The AllDifferent constraint is a crucial component of any constraint toolkit, language or solver, since it is very widely used in a variety of constraint models. The literature contains many different versions of this constraint, which trade strength of inference against computational cost. In this paper, we focus on the highest strength of inference, enforcing a property known as generalised arc consistency (GAC). This work is an analytical survey of optimizations of the main algorithm for GAC for the AllDifferent constraint. We evaluate empirically a number of key techniques from the literature. We also report important implementation details of those techniques, which have often not been described in published papers. We pay particular attention to improving incrementality by exploiting the strongly-connected components discovered during the standard propagation process, since this has not been detailed before. Our empirical work represents by far the most extensive set of experiments on variants of GAC algorithms for AllDifferent. Overall, the best combination of optimizations gives a mean speedup of 168 times over the same implementation without the optimizations. 1.

Abstract. Table constraints play an important role within constraint programming. Recently, many schemes or algorithms have been proposed to propagate table constraints or/and to compress their representation. We show that simple tabular reduction (STR), a technique proposed by J. Ullmann to dynamically maintain the tables of supports, is very often the most efficient practical approach to enforce generalized arc consistency within MAC. We also describe an optimization of STR which allows limiting the number of operations related to validity checking or search of supports. Interestingly enough, this optimization makes STR potentially r times faster where r is the arity of the constraint(s). The results of an extensive experimentation that we have conducted with respect to random and structured instances indicate that the optimized algorithm we propose is usually around twice as fast as the original STR and can be up to one order of magnitude faster than previous state-of-the-art algorithms on some series of instances. 1

Abstract. Given an arbitrary constraint c on n variables with domain size d, we show how to generate a custom propagator that establishes GAC in time O(nd) by precomputing the propagation that would be performed on every reachable subdomain of scope(c). Our propagators are stateless: they store no state between calls, and so incur no overhead in storing and backtracking state during search. The preprocessing step can take exponential time and the custom propagator is potentially exponential in size. However, for small constraints used repeatedly, in one problem or many, this technique can provide substantial practical gains. Our experimental results show that, compared with optimised implementations of the table constraint, this technique can lead to an order of magnitude speedup, while doing identical search on realistic problems. The technique can also be many times faster than a decomposition into primitive constraints in the Minion solver. Propagation is so fast that, for constraints available in our solver, the generated propagator compares well with a human-optimised propagator for the same constraint. 1

Abstract. We present an implementation of table constraints in CLP(FD). For binary constraints, the supports of each value are represented as a finite-domain variable, and action rules are used to propagate value exclusions. The bit-vector representation of finite domains facilitates constant-time removal of unsupported values. For n-ary constraints, we propose pair-wise arc consistency (AC), which ensures that each value has a support in the domain of every related variable. Pair-wise AC does not require introducing new problem variables as done in binarization methods and allows for compact representation of constraints. Nevertheless, pair-wise AC is weaker than general arc consistency (GAC) in terms of pruning power and requires a final check when a constraint becomes ground. To remedy this weakness, we propose adopting early checks when constraints are sufficiently instantiated. Our experimentation shows that pair-wise AC with early checking is as effective as GAC for positive constraints. 1

Abstract. This paper presents a novel domain-consistency algorithm which does not maintain supports dynamically during propagation, but rather maintain forbidden values. It introduces the optimal NAC4 (negative AC4) algorithm based on this idea. It further shows that maintaining forbidden values dynamically allows the generic algorithm AC5 to achieve domain consistency in time O(ed) for classes of constraints in which the number of supports is O(d 2) but the number of forbidden values is O(d). The paper also shows how forbidden values and supports can be used jointly to achieve domain consistency on logical combinations of constraints and to compute validity as well as entailment of constraints. Experimental results show the benefits of the joint exploitation of supports and forbidden values. 1

Special-purpose constraint propagation algorithms (such as those for the element constraint) frequently make implicit use of short supports — by examining a subset of the variables, they can infer support for all other variables and values and save substantial work. However, to date general purpose propagation algorithms (such as GAC-Schema) rely upon supports involving all variables. We demonstrate how to employ short supports in a new general purpose propagation algorithm called SHORTGAC. This works when provided with either an explicit list of allowed short tuples, or a function to calculate the next supporting short tuple. Empirical analyses demonstrate the efficiency of SHORTGAC compared to other general-purpose propagation algorithms. In some cases SHORTGAC even exhibits similar performance to special-purpose propagators. 1

Abstract. Constraints that are defined by tables of allowed tuples of assignments are common in constraint programming. In this paper we present an approach to reformulating table constraints of large arity into a conjunction of lower arity constraints. Our approach exploits functional dependencies. We study the complexity of finding reformulations that either minimize the memory size or arity of a constraint using a set of its functional dependencies. We also present an algorithm to compute such reformulations. We show that our technique is complementary to existing approaches for compressing extensional constraints. 1

Abstract. This paper describes the three main improvements made to the solver Abscon 109 [9]. The new version, Abscon 112, is able to automatically break some variable symmetries, infer allDifferent constraints from cliques of variables that are pair-wise irreflexive, and use an optimized version of the STR (Simple Tabular Reduction) technique initially introduced by J. Ullmann for table constraints. 1 From Local to Global Variable Symmetries In [10], we have proposed to automatically detect variable symmetries of CSP instances by computing for each constraint scope a partition exhibiting locally symmetrical variables. From this local information that can be obtained in polynomial time, we can build a so-called lsv-graph whose automorphisms correspond to (global) variable symmetries. Interestingly enough, our approach allows us to disregard the representation (extension, intension, global) of constraints. Besides, the size of the lsv-graph is linear wrt the number of constraints (and their arity). To break symmetries from the generators returned by a graph automorphism

Abstract. Most research on Constraint Programming concerns the (exponential) search space of Constraint Satisfaction Problems (CSPs) and intelligent algorithms that reduce and explore it. This work proposes a different way, not of solving a problem, but of storing the domains of its variables, an important—and less focused—issue especially when they are large. The new data structures that are used are proved theoretically and empirically to adapt better to large domains, than the commonly used ones. The experiments of this work display the contrast between the most popular Constraint Programming systems and a new system that uses the data structures proposed in order to solve CSP instances with wide domains, such as known Bioinformatics problems.

This document is the case for support for Grant number EP/H004092/1, funded by the EPSRC in the UK. Our aim is to improve dramatically the scalability of constraint technology, while removing its reliance on manual tuning by an expert. This aim will be realised through the development of a constraint solver synthesiser, the principal components of which are a Model Analyser, a Constraint Solver Generator, and a Generator Tuner. As a research proposal, the reader should bear in mind that this is a proposal for future work, rather than a report on completed research.