GPU-Based Evaluation to Accelerate Particle Swarm Algorithm (ACM)
With the advent of the cards GPU, many computational problems have suffered from a net increase of performance. Nevertheless, the improvement depends strongly on the usage of the technology and the porting process used in the adaptation of the problem. These aspects are critical in order that the improvement of the performance of the code adapted to GPU is significant. This article focus on the study of the strategies for the porting of Particle Swarm Algorithm with parallel-evaluation of Schwefel Problem 1.2 and Rosenbrock function. The implementation evaluates the population in GPU, whereas the other intrinsic operators of the algorithm are executed in CPU. The design, the implementation and the associated issues related to GPU execution context are evaluated and presented. The results demonstrate the effectiveness of the proposed approach and its capability to effectively exploit the architecture of GPU.
Paper available at ACM.