A parallel algorithm for the stack breadth-first search

Parallelization of the P-complete problem is known to be difficult. In this paper, we consider the parallelizability of a stack breadth-first search (stack BFS) problem, which is proved to be P-complete. We first propose the longest path length (LPL) as a measure for the P-completeness of the stack BFS. Next, using this measure, we propose an efficient parallel algorithm for the stack BFS. Assuming the size and LPL of an input graph are n and 1, respectively, the complexity of the algorithm indicates that the stack BFS is in the class NCk+1 if l = O(log(k) n), where k is a positive integer. In addition, the algorithm is cost optimal if l = O(n(is an element of)), where 0 < is an element of < 1.