Exploiting the sources of parallelism in logic programs

To achieve efficient and good performance of logic programs a proposed parallel execution model based on MUSE which was implemented by SICS of Sweden has been developed and implemented. This model is an amalgamation of two main types of parallelism AND-parallelism and OR-parallelism. The implementation of this model is done on a shared memory multiprocessor. Due to this model the logic program can be described by an AND/OR tree. The different branches of this tree are distributed to the available processing elements (PEs) during run time. Two scheduling techniques have been developed and implemented. In the first scheduler. The PEs are distributed through the AND/OR tree based on the sequential PROLOG (i.e. left-right depth first). The second scheduler is based on the static processing element allocation. Two benchmark programs have been used to assess the performance of this execution model; these are the natural language processing parser and an animal recognition expert system.<
>