On the Effectiveness of Thorup's Shortest Path Algorithm for Large-Scale Network Simulation

An efficient solution for a single-source problem called Thorup's algorithm has been proposed, whose computational complexity, O(N), is smaller than that of Dijkstra's algorithm, O(N log N). In this paper, we compare the performance of Dijkstra's algorithm and Thorup's algorithm for large-scale network simulation. Through extensive experiments, we show that Thorup's algorithm is slightly faster with approximately 30% larger memory consumption than Dijkstra's algorithm for a large-scale network simulation, although difference in Dijkstra's and Thorup's algorithms is not so significant.