Domain independent ask approximation in CCP

Abstract

The main difficulty in the definition of a static analysis framework for CC programs is probably related to the correct approximation of the entailment relation between constraints. This approximation is needed for the abstract evaluation of the ask guards and directly influences the overall precision of the analysis. In this paper we provide a solution to this problem by stating reasonable correctness conditions relating the abstract and the concrete domains of computation. The solution is domain independent in the sense that it can be applied to the class of downward closed observations. Properties falling in this class (e.g. freeness) have already been studied in the context of the analysis of sequential logic programs. We believe that the same abstract domains can be usefully applied to the CC context to provide meaningful ask approximations.

Strategic directions in constraint programming

“…Other theoretical efforts focus on the possibility of analyzing CC-like programs at compile time, thus deriving properties to be used at run time. This holds, for example, for the works on abstract interpretation [Zaffanella 1995;Codognet et al 1990], which execute CC programs on an abstract constraint domain with the hope of deriving some useful knowledge for program simplification, for those on suspensión analysis [Codish et al 1994], whose aim is to understand the conditions under which CC programs deadlock, and for those on relating CC and CLP languages [Bueno et al 1994], which try to parallelize CLP programs using CCbased techniques or to sequentialize CC programs via an analysis of their inherent concurrency.…”

Concurrent and Reactive Constraint Programming

“…Another approach to the analysis of ccp was pursed in [65,66] where it was extended to ccp languages the generalized semantics approach to static analysis, initially proposed in [41] for sequential CLP languages. [65] shows that such an extension can be easily achieved for approximations that are closed under antientailment: applications include analyses that can identify definite suspensions, e.g., to compute upper bounds to the degree of concurrency in a ccp program.…”

“…[65] shows that such an extension can be easily achieved for approximations that are closed under antientailment: applications include analyses that can identify definite suspensions, e.g., to compute upper bounds to the degree of concurrency in a ccp program. For the more common case of entailment closed properties (that are of interest for, e.g., proving suspension freeness), it is shown in [66] that correctness can only be achieved by modifying the generalized semantics approach so as to introduce a domain-dependent approximation of the synchronization primitive, which cannot be modeled as an entailment test on the abstract domain.…”