An Heuristics for Load Balancing and Granularity Control in the Parallel Instantiation of Disjunctive Logic Programs

In this paper we present a dynamic heuristics that allows for improving the performance of a parallel instantiator algorithm based on the DLV system. In this system, each rule is rewritten in several “splits” of the same size that are assigned to a number of parallel instatiator subprocesses. The new heuristics allows for dynamically determining an optimal amount of work that has to be assigned to each parallel instantiator, and, thus, it improves the overall efficiency of the parallel evaluation. We implemented our heuristics and performed an experimental analysis that confirms the efficacy of the proposed method.