Parallel processing model for logic programming

Book · Wed Jan 01 04:00:00 EST 1986

OSTI ID:7162944

Tung, Y W

A theoretical to the analysis of parallelism in log programming is proposed. Also, a design for a parallel logic programming compiler and architecture is offered. First, the parallel-resolution principle is formulated as an extension of SLD-resolution. Then, the computation and communication time complexities for various parallel schemes are analyzed. The parallel scheme using the non-shared-variable principle is found to be efficient for most of the declarative inference applications. Yet, there are still ways to minimize the time complexity through proper subgoal ordering. The improved parallel scheme is called NSVO (non-shared-variable with ordering), and its improved performance is demonstrated by some examples. Based on this scheme, a parallel compiler is proposed. This compiler determines the most efficient subgoal execution sequence is three steps: (1) automatic input-output mode analysis; (2) detection of data dependencies, i.e., functional dependency, internal dependency, coupling dependency and run-time dependency; (3) subgoal ordering algorithm based on data dependencies and the principles of groundability and subtree complexity. Based on this parallelism analysis, a computational model is proposed. The model consists of a multiprocessor system whose control mechanism is based on the demand-driven principle.

• OSTI ID: 7162944
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990200* -- Mathematics & Computers
ARRAY PROCESSORS
COMMUNICATIONS
ELECTRONIC CIRCUITS
FUNCTIONAL MODELS
LOGIC CIRCUITS
PARALLEL PROCESSING
PROGRAMMING