Impact of the Random Number generator quality on particle swarm optimization algorithm running on graphic processor units

Abstract

Particle swarm optimization (PSO) is a bioinspired technique widely used to solve real optimization problems. In the recent years, the use of Graphics Processing Units (GPU) has been proposed for some general purpose computing applications. Some PSO implementations on GPU were already proposed. The major benefit to implement the PSO for GPU is the possibility to reduce the execution time. It occurs due to the higher computing power presented nowadays on GPUs platform. A study on the impact of the quality of Random Number generator has been made but it only covered some variations of the algorithm on a sequential platform. In this paper, we present an analysis of the performance of the random number generator on GPU based PSOs in terms of the RNG statistical quality. We showed that the Xorshift random number generator for GPU presents enough quality to be used by the PSO algorithm.