An Abstract Machine for Tabled Execution of Stratified Logic Programs

Click on the title to obtain a gzip-ed PostScript version of a paper almost identical to the journal version (146K).


SLG resolution uses tabling to evaluate non-floundering normal logic programs according to the well-founded semantics. The SLG-WAM, which forms the engine of the XSB system, can compute in-memory recursive queries an order of magnitude faster than current deductive databases. At the same time, the SLG-WAM tightly integrates Prolog code with tabled SLG code, and can execute Prolog code with minimal overhead compared to the WAM. As a result, the SLG-WAM brings to logic programming important termination and complexity properties of deductive databases. This article describes the architecture of the SLG-WAM for a powerful class of programs, the class of fixed-order dynamically stratified programs. We offer a detailed description of the algorithms, data structures, and instructions that the SLG-WAM adds to the WAM, and a performance analysis of engine overhead due to the extensions.