Description
The architecture described in this paper achieves high performance execution of Prolog programs by exploiting fine grain parallelism. Fine grain parallelism includes unification parallelism as well as parallelism among the arithmetic and bookkeeping operations. Our implementation of the architecture has multiple functional units, each tailored to a specific task, that operate in parallel. Unification and arithmetic operations are allowed to execute out of order and are dynamically scheduled on several unification units. Simulation results for the implementation are presented and analyzed.