(A) study on the execution mechanisms of a functional logic language with lazy evaluation

Abstract

This thesis aims at developing a cost-effective mechanism incorporating functional programming into logic programming, so that both programming styles can be expressed naturally and executed efficiently in a single framework. For achieving this goal, we first propose a functional logic language, called Lazy Aflog, which is an extension of Prolog with the functional language features such as infinite data structures and high order functions. Second, an abstract machine for Lazy Aflog, called FWAM-II, is proposed as an implementation method on a sequential machine. It is an extension of WAM with the instructions and run-time structures to provide the suspension/reactivation of functional closure. We have shown by experimental results that Lazy Aflog and FWAM-II pair is a good compromise between the expressiveness and efficiency of the combination. Finally, to overcome the speed limitation caused by the sequential nature of FWAM-II, a parallel execution model for Lazy Aflog and its abstract machine, called $PR^3$ (Parallel Resolution and Reduction with RAP) and PFWAM-II (Parallel Functional WAM-II), are also proposed. They are extensions of Degroot``s RAP and Hermenegildo``s RAP-WAM with the primitives for parallel function reduction, respectively. Upon the simulation results on $PR^3$ and PFWAM-II, the speedup behaviours of parallel reduction is almost the same as the parallel reduction is almost the same as the parallel resolution. It is claimed from this analysis that not only logic programming but also functional programming can be supported in $PR^3$ and PFWAM-II simultaneously with almost the same efficiency of single paradigm programming.