Coupled resolution engines for programming knowledge based systems in logic

TL;DR: The design and use of a logic programming system that reconciles three different requirements for this type of application, which couples a concurrent logic programming or CLP engine with a multi-threaded Prolog engine, and is compatible with scale enhancing approaches to resolution using knowledge base machines.

Abstract: Research on a fifth generation of computer systems aims at building high performance knowledge processing engines on the basis of the highly parallel execution of logic programs. Howev er, outstanding problems in the design and implementation of logic programming languages make it problematic to combine parallel execution, adequate support for deduction and a systems programming capability. This combination is important for programming multi-user knowledge based systems in logic. This research describes the design and use of a logic programming system that reconciles these three different requirements for this type of application. The system couples a concurrent logic programming or CLP engine with a multi-threaded Prolog engine. Concurrency among coupled CLP and Prolog computations is sustained partly by fine-grained and-parallelism in the CLP engine, and partly by coarse grained concurrent processing of multiple Prolog processes and a CLP engine under Unix. Communication among distinct computations is realised by message passing, which allows the architecture to reduce contention for shared memory during multi-processing. The CLP engine sustains the process interpretation of systems programming, which enables it to execute the management system of a multi-user knowledge based system. The Prolog engine can exhibit don’t know non-determinism, which enables it to perform exhaustive searches over a knowledge base. The coupled system allows Prolog computation threads to share or to have private databases, and is compatible with scale enhancing approaches to resolution using knowledge base machines. Only minor extensions to CLP and Prolog engines are needed to realise the coupled architecture. This leaves it free to evolve with emerging CLP and Prolog implementation technology, especially in the area of multi-processing CLP languages in andparallel and Prolog in or-parallel. Ke ywords: logic programming, parallelism, deduction, control, systems programming, knowledge base