Warwick Harvey and Peter J. Stuckey. Constraint representation for propagation. In Michael Maher and Jean-Francois Puget, editors, Proceedings of the Fourth International Conference on Principles and Practice of Constraint Programming | CP'98, LNCS 1520, pages 235-249. Springer, 1998.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....benchmark the propagation performance of our constraint solver. x f = n (2) x f = 1 (3) We have implemented the constraint model one to one with Mozart Oz finite domain constraints and used disjunctive combinators producing choice points to obtain the same behavior as the program used in [4]. The search strategy is nave, i.e. it picks from the left most finite domain variable x l the minimum element m and creates a choice point x l = m x l m. Deriving an Early Failure Criterion. Deriving a criterion is a creative process and it is hard to give any guidelines. But it is ....

....programming language CLP(R) is proposed in [5] They use quote and eval functions which are analogous to the corresponding Lisp functions. Solvers dedicated to a certain set of constraints as well as dedicated constraints can of course do the same analysis as discussed in this paper. In [4] early failure detection as described in Sect. 3 has been proposed as a by product of analyzing the impact of simplifications for equational constraints on the propagation behavior. ILOG Solver [7] is a C library for constraint programming in C . It does not support first class constraints as ....

Warwick Harvey and Peter J. Stuckey. Constraint representation for propagation. In M. Maher and J.-F. Puget, editors, Proceedings of the Fourth International Conference on Principles and Practice of Constraint Programming (CP98), Lecture Notes in Computer Science, pages 235--249, Pisa, Italy, October 1998. Springer-Verlag.

....have written using the Objective Caml system [Leroy, 1999] This strongly typed functional language provides well documented libraries for various data structures. A fast compiler produces ecient native code. Our constraint library includes standard nite domain variables, arithmetic constraints [Harvey and Stuckey, 1998], ecient arc consistency propagation on binary constraints (AC6) Bessiere and Cordier, 1993] and some global constraints: di erence with ltering based on a bipartite graph matching algorithm, sorting, The search is controlled in a Prolog way with goals (success and failure continuations) and ....

Harvey, W. and Stuckey, P. J. (1998). Constraint representation for propagation. In Maher, M. and Puget, J.-F., editors, Principles and Practice of Constraint Programming, pages 235{ 249. Springer-Verlag.

.... i2T (x i = x f 1) x f = n (2) 8 arc(f; T ) 2 Arcs : i2T (x i 1 = x f ) x f = 1 (3) We have implemented the constraint model one to one with Mozart Oz finite domain constraints and used disjunctive combinators producing choice points to obtain the same behavior as the program used in [3]. The search strategy is nave, i.e. it picks from the left most finite domain variable x l the minimum element m and creates a choice point x l = m x l 6= m. Deriving an Early Failure Criterion. Deriving a criterion is a creative process and it is hard to give any guidelines. But it is ....

....constraint programming language CLP(R) is proposed in [4] They use quote and eval functions which are analogous to the corresponding Lisp functions. Solvers dedicated to a certain set of constraints as well as dedicated constraints can of course do the same analysis as discussed in this paper. In [3] early failure detection as described in Sect. 3 has been proposed as a by product of analyzing the impact of simplifications for equational constraints on the propagation behavior. ILOG Solver [6] is a C library for constraint programming in C . It does not support first class constraints as ....

W. Harvey and P. J. Stuckey. Constraint representation for propagation. In M. Maher and J.-F. Puget, editors, Proceedings of the Fourth International Conference on Principles and Practice of Constraint Programming (CP98), Lecture Notes in Computer Science, pages 235--249, Pisa, Italy, October 1998. Springer-Verlag.

....issues for implementing advanced propagators. It discusses also the related work and future work. In this paper, we presented a simple interface between the CP system and the filtering algorithms. The interface cannot handle advanced filtering algorithms that can perform constraint reasoning [HS98,Mul00]. Such filtering algorithms can impose new constraints, replace itself with another constraint or even transform a set of constraints into another set of constraints. In addition, it can drop old variables from a constraint and add new variables to a constraint. In Mozart, constraints can unify ....

Warwick Harvey and Peter Stuckey. Constraint representation for propagation. In Jean-Francois Puget and Michael Maher, editors, Proceedings of the Fourth International Conference on Principles and Practice of Constraint Programming(CP98), Lecture Notes in Computer Science, pages 235--249, Pisa, Italy, October 1998. SpringerVerlag.

No context found.

Warwick Harvey and Peter J. Stuckey. Constraint representation for propagation. In Michael Maher and Jean-Francois Puget, editors, Proceedings of the Fourth International Conference on Principles and Practice of Constraint Programming | CP'98, LNCS 1520, pages 235-249. Springer, 1998.

....equality constraints. Since we wished to support full equality constraints, this required implementing a true uni cation based solver which interacted gracefully with the Mercury run time system. This integration is described more fully in [3] The integer solver is the same as that described in [10]. It is a bounds propagation solver which keeps linear constraints in a tableau form, and simpli es them during execution to improve further propagation. It was originally embedded in CLP(R) s compiler and runtime system, yielding the language CLP(Z) It has since been interfaced to Mercury, and ....

W. Harvey and P.J. Stuckey. Constraint representation for propagation. In Procs. of PPCP98, pages 235-249, 1998.

....performance benefits that arise. Previous authors [13, 15] have noted the di#erence in e#ciency in bounds and domain propagation, for particular primitive constraints but not considered when di#erent propagators lead to the same search space. The closest related work is that of Harvey and Stuckey [7], who consider the relative propagation strengths of di#erent equivalent forms of constraints. Although they consider both domain and bounds propagation, they never compare bounds propagation to domain propagation. While there has been considerable success in optimizing constraint programs over ....

W. Harvey and P. J. Stuckey. Constraint representation for propagation. In M. Maher and J.-F. Puget, editors, Proceedings of the Fourth International Conference on Principles and Practice of Constraint Programming, volume 1520 of Lecture Notes in Computer Science, pages 235--249. Springer-Verlag, Oct. 1998.

.... classic problem; mycien m which colors a 5 colorable graph (taken from [16] with n nodes and m edges with 4 colours; and fulladder which searches for a single faulty gate in a n bit adder (see e.g. 14] for the case of 1 bit) 8 More exactly, we use the integer propagation solver described in [7] which is implemented in C and interfaced to HAL using the methodology described in Section 5. 9 It is much easier to build highly exible but inecient mechanisms for de ning solvers. Benchmark Var Con Search Dynamic Scheduling dyn t dyn i SICSw SICS b mycie23 71 184 583 19717 1855 1816 34769 ....

W. Harvey and P.J. Stuckey. Constraint representation for propagation. In Procs. of PPCP'98, LNCS, pages 235-249. Springer-Verlag, 1998.

....performance benefits that arise. Previous authors [13, 15] have noted the di#erence in e#ciency in bounds and domain propagation, for particular primitive constraints but not considered when di#erent propagators lead to the same search space. The closest related work is that of Harvey and Stuckey [7], who consider the relative propagation strengths of di#erent equivalent forms of constraints. Although they consider both domain and bounds propagation, they never compare bounds propagation to domain propagation. While there has been considerable success in optimizing constraint programs over ....

W. Harvey and P. J. Stuckey. Constraint representation for propagation. In M. Maher and J.-F. Puget, editors, Proceedings of the Fourth International Conference on Principles and Practice of Constraint Programming, volume 1520 of Lecture Notes in Computer Science, pages 235--249. Springer-Verlag, Oct. 1998.

....the primitive constraint c 1 j P n i=1 a i x i op d where op 2 f= 6=g, Let c 2 j P n i=1;i6=j a i x i op d Gamma a j d j . Then for solv 2 fdsolv ; bsolvg and any domain D where D(x j ) fd j g we have that solv (c 1 ; D) solv (c 2 ; D) 1 We omit proofs due to space reasons, see [4] for details. Often large linear constraints are broken up into smaller parts, in order to reduce the kind of constraints to a small number of forms by introducing new variables. For example the constraint x 1 x 2 Gamma 2x 3 Gamma 2x 4 = 5 can be decomposed into v x 2 Gamma 2x 4 = 5 v = x ....

Warwick Harvey and Peter J. Stuckey. Constraint representation for propagation. Computer Science Technical Report 98/10, The University of Melbourne, 1998. Available at http://www.cs.mu.oz.au/~pjs/papers/papers.html.

....forms of equality constraints. Since we wished to support full equality constraints, this required 12 implementing a true unification based solver which interacted gracefully with the Mercury solver. This is described more fully in [1] The integer solver is the same as that described in [5]. It is a bounds propagation solver which keeps linear constraints in a tableau form, and simplifies them during execution to improve further propagation. It was originally embedded in CLP(R) s compiler and runtime system, yielding the language CLP(Z) It has since been interfaced to Mercury, and ....

....For each of the three solvers Herbrand, integer and real we have selected four to five standard benchmarks and compared them to another CLP system. For the Herbrand benchmarks we compare with SICStus Prolog 3.7. 1 (compact code) while for the integer and real we compare with CLP(Z) [5] and CLP(R) v1.02 respectively. Since CLP(Z) and CLP(R) use exactly the same underlying solvers as HAL, these comparisons are solver independent. The results are shown in Table 1. The Herbrand benchmarks are executed both with and without garbage collection. HAL s garbage collection is provided ....

[Article contains additional citation context not shown here]

W. Harvey and P.J. Stuckey. Constraint representation for propagation. In Procs. of PPCP98, pages 235--249, 1998.

....equality constraints. Since we wished to support full equality constraints, this required implementing a true uni cation based solver which interacted gracefully with the Mercury run time system. This integration is described more fully in [1] The integer solver is the same as that described in [5]. It is a bounds propagation solver which keeps linear constraints in a tableau form, and simpli es them during execution to improve further propagation. It was originally embedded in CLP(R) s compiler and runtime system, yielding the language CLP(Z) It has since been interfaced to Mercury, and ....

....For each of the three solvers Herbrand, integer and real we have selected four to ve standard benchmarks and compared them to another CLP system. For the Herbrand benchmarks we compare with SICStus Prolog 3.7. 1 (compact code) while for the integer and real benchmarks we compare with CLP(Z) [5] and CLP(R) v1.02 respectively. Since CLP(Z) and CLP(R) use exactly the same underlying solvers as HAL, these comparisons are solver independent. The results are shown in Table 1. The Herbrand benchmarks are executed both with and without garbage collection. HAL s garbage collection is provided ....

[Article contains additional citation context not shown here]

W. Harvey and P.J. Stuckey. Constraint representation for propagation. In Procs. of PPCP98, pages 235-249, 1998.

....function, if D 0 = D and D k 1 = iter( i2Jk c k ; D k ) then solv(C; D) u k1 D k The following lemmas establish simple equivalences in propagation solvers. First, we can scale constraints without affecting their propagation behaviour. 1 1 We omit full proofs due to space reasons, see [HS98] for details. Lemma 1. Consider the primitive constraints c 1 j P n i=1 a i x i op d and c 2 j b P n i=1 a i x i op bd, where op 2 f= 6=g and b 0. Then for any domain D and solv 2 fdsolv ; bsolvg we have solv (c 1 ; D) solv (c 2 ; D) Proof. For domain propagation, clearly the ....

Warwick Harvey and Peter J. Stuckey. Constraint representation for propagation. Computer Science Technical Report 98/10, The University of Melbourne, 1998.

No context found.

Warwick Harvey and Peter Stuckey. Constraint representation for propagation. In Jean-Francois Puget and Michael Maher, editors, Proceedings of the Fourth International Conference on Principles and Practice of Constraint Programming(CP98), Lecture Notes in Computer Science, pages 235-249, Pisa, Italy, October 1998. SpringerVerlag.

No context found.

Warwick Harvey and Peter J. Stuckey. Constraint Representation for Propagation. In Proc. of CP'98, pages 235--249, Pisa, Italy, October 26-30 1998.

No context found.

Warwick Harvey and Peter Stuckey. Constraint representation for propagation. In Jean-Francois Puget and Michael Maher, editors, Proceedings of the Fourth International Conference on Principles and Practice of Constraint Programming (CP98), Lecture Notes in Computer Science, pages 235-249, Pisa, Italy, October 1998. Springer-Verlag.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC