Evaluation of Hierarchical Pincers Attack Search on Distributed Memory Systems

This paper evaluates a distributed memory parallel search algorithm based on hierarchical pincers attack search (HPAS). HPAS is a parallel tree search algorithm with depth first search that uses a master processor and some slave processors. This algorithm is efficient because it can implicitly share data frequently with all processors using shared memory. The communication overhead costs of HPAS on a distributed memory system will increase because of the need for frequent inter-processor communication. Thus, few studies have reported implementations of HPAS on a distributed memory system, and the effectiveness of this method is unclear. Therefore, this paper proposes an HPAS implementation method on a distributed memory system and evaluates its effectiveness. As a result of the evaluation, the maximum speedup ratio of the proposed method compared with the branch and bound method is approximately 52.78 times.