GPU performance impact of the Durkin's radio propagation algorithm

Abstract

Network simulators are important tools when used by researchers. However, complex simulations for largescale networks (especially if the scenarios are realistic, i.e. introducing terrain details) results in unreasonably long response from the simulator. High performance parallel resources like GPU devices are solution for obtaining results in reasonable time. Additionally, the latest GPU generations have architecture that enables the developer to configure different input parameters, in order to achieve better performance. In this paper we evaluate the performance impact of these configurations for various resolutions of terrains. The obtained results show that there is an impact on the performance, however it is not significant.