Index matching hardware for selective clause retrieval in large Prolog knowledge bases

摘要

While interest in the use of Prolog for knowledge base applications is growing, the size of these applications is limited by the capabilities of current Prolog translators for handling disc resident clauses. A major impediment is the inordinate amount of time required for retrieval and unification of clauses from a large set stored on disc. The CLAuse Retrieval Engine (CLARE) is a hardware engine designed specially to enhance the performance of Prolog systems when dealing with large clause sets. It is based on selective retrieval of disc resident clauses using a superimposed codeword indexing strategy. In this article, the principle of the index matching strategy is explained and the design of corresponding hardware is described.