Proceedings of the Workshop on Parallelism and Implementation Technologies for (Constraint) Logic Programming Languages

Incremental stack-copying was used to implement OR-parallel Prolog efficiently in the MUSE system. The idea of incremental stack-copying is to only copy the difference between the WAM data areas of two processors, instead of copying them entirely. In order to further reduce the communication during stack-copying and make its implementation efficient on distributed memory machines, a new technique, called stack-splitting, has recently been proposed. In this paper, we describe a scheme to implement incremental stack-splitting for OR-parallelism on distributed memory machines. The incremental stacksplitting scheme presented is based on a procedure which labels parallel choice-points and then compares the labels to determine the incremental WAM areas to be copied. Furthermore, we describe a scheduling scheme for the incremental stack-splitting strategy. Both the incremental stack-splitting and the scheduling schemes described are currently being implemented in the PALS system. Preliminary results from this implementation are also included.