A CUDA Implementation of the Standard Particle Swarm Optimization

The social learning process of birds and fishesinspired the development of the heuristic Particle Swarm Optimization (PSO) search algorithm. The advancement of GraphicsProcessing Units (GPU) and the Compute Unified Device Architecture (CUDA) platform plays a significant role to reduce thecomputational time in search algorithm development. This paperpresents a good implementation for the Standard Particle SwarmOptimization (SPSO) on a GPU based on the CUDA architecture, which uses coalescing memory access. The algorithm is evaluatedon a suite of well-known benchmark optimization functions. Theexperiments are performed on an NVIDIA GeForce GTX 980GPU and a single core of 3.20 GHz Intel Core i5 4570 CPUand the test results demonstrate that the GPU algorithm runsabout maximum 46 times faster than the corresponding CPUalgorithm. Therefore, this proposed algorithm can be used toimprove required time to solve optimization problems.