Acceleration of Dijkstra's algorithm on multi-core processors

Abstract

The Single Source Shortest Path (SSSP) problem has been solved using various algorithms. We focus on accelerating a well known SSSP algorithm, the Dijkstra's algorithm using a multi-core CPU. We achieve acceleration by using the iParallel kernel, a hybrid kernel that runs on the sequential as well as parallel kernels intelligently. Our experimental results help find an optimal threshold to exploit parallelism for different sizes of the search terrain. We have achieved an overall acceleration of −48% and −51% on a dual-core ARM A9 processor and a 16-core Epiphany co-processor respectively.