parallel algorithm

The time consumption in solving computationally heavy problems has always been a concern for computer programmers. Due to simplicity of its implementation, the PSO (Particle Swarm Optimization) is a suitable meta-heuristic algorithm for solving computationally heavy problems. However, despite the simplicity, the algorithm is inefficient for solving real computationally heavy problems but the presence of local interactions between particles has made this algorithm suitable for parallelization. On the other hand, by the invention of GPU (Graphical Processor Unit) and introducing the CUDA architecture as a GPU in the NVIDIA graphical processor, fundamental changes has been made in solving this type of problems. Despite all the research done in the field of implementing the algorithms through GPUs, some aspects of parallelization have not been addressed for suitable speedup and efficiency on NVIDIA GPUs. By considering the Geforce GT 525M, which is a relatively weak GPU, this paper tries to achieve the maximum speedup of the algorithm by implementing on this GPU. This experience led to reaching the acceptable efficiency on other GPUs. To reach the achievement, the multi-kernel model was used. The results show the speedup of 15.98 in solving the Rastrigin function.