(A) parallel prolog execution model

Abstract

In this thesis, a model for parallel prolog inference system is proposed in which the parallelism is realized by pipeline processing for the solutions. Moreover, to enhance the AND parallelism, the ordering of all subgoals in a clause is determined in run time or in compile time, using the producer and consumer relationships among subgoals sharing common variables. The OR parallelism is not applied to this model because of the communication complexity. This model is applied to multiprocessor system. In this system, a processor has a local processor list and this processor searches this list for a process which is ready for execution. The process is ready for execution when the producers of its consumed variables are done.