Revisiting a BSP/CGM transitive closure algorithm

Abstract

We present a new BSP/CGM parallel algorithm for the transitive closure problem. Our algorithm uses O(n/sup 3//p/spl alpha/) local computation time with O(p//spl alpha/) communication rounds, where /spl alpha/ is the size in bits that can be stored in a primitive data item. For all the randomly generated graphs that were used in the tests, the number of communication rounds was bounded by log p/spl bsol//spl alpha/+1. Our algorithm, even for the worst case, improves the previous results. The algorithm was implemented and the results show the efficiency and scalability of the presented algorithm and compare favorably with other parallel implementations.