Run Time Optimization using a novel implementation of Parallel-PSO for real-world applications

Abstract

The majority of optimization algorithms and methods generally necessitate a considerable run time to reach their goal. Most of them are used mainly in real-world applications. This article concentrates on an efficient and well-known algorithm to solve optimization problems: the Particle Swarm Optimisation algorithm (PSO). This algorithm needs a considerable run time to solve an optimization problem with a high dimension space and data. The article also concentrates on OpenCL, which defines a common parallel programming language for various devices such as GPU, CPU, FPGA, etc. In order to minimize the run time of PSO, this paper introduces a new implementation of PSO in OpenCL. By decomposing the PSO code into two fragments, each one can run simultaneously. The experimental results covered both the sequential and parallel implementations. Furthermore, show that the PSO’ OpenCL implementation is faster than the Sequential-PSO implementation. The OpenCL profiling results show the timing of each part of the executing of PSO in OpenCL.