Improving Distributed Unification through Type Analysis

n distributed implementations of logic programming, data structures are spread among different nodes and unification involves sending and receiving messages to access them. Traditional implementations make remote data structures accessible to other processes by sending messages which carry either the overall data structure (infinite-level copying) or only remote references to these data structures (zero-level copying). These fixed policies can be far from optimal on various classes of programs and may induce substantial overhead. The purpose of this paper is to present an implementation scheme for distributed logic programming which consists of tailoring the copying level to each procedure. The scheme is based on a consumption specification which describes the way the procedure "consumes" its arguments locally. The consumption specification (or an approximation of it) can be automatically obtained through a static analysis inspired by traditional type analyses. The paper also describes a high-level distributed implementation that uses the consumption specification to avoid unnecessary copying and to request data structures globally. Experimental results for a network of workstations show the potential of the approach.