Determinacy Driven Optimizations of And-Parallel Prolog Implementations

And-parallelism arises in Prolog programs when conjunctive subgoals in a query or the body of a clause are executed in parallel. In this paper we present three optimizations, namely, the last parallel call optimization, the shallow parallelism optimization, and the processor determinacy optimization that take advantage of determinacy to improve eeciency of and-parallel execution of Prolog programs. All three optimizations depend on a posteriori knowledge of determinacy rather than on a priori knowledge detected at compile-time. With the help of these optimizations, data-and parallel Prolog programs can be eeciently executed on very general and-parallel system (such as &ACE and &-Prolog) with the eeciency of dedicated data-and parallel systems (such as Reform Prolog system). These optimizations have been implemented in the &ACE system, and the results are also presented.