关键词: 布谷鸟搜索算法, 并行计算, 图形处理器（GPU）, 统一计算设备架构（CUDA）

Abstract: Cuckoo search (CS) is an intelligent meta-heuristic algorithm which has developed in recent years and has been successfully applied to a number of optimization problems. In order to solve the time consuming problem of CS when dealing with the large amounts of data and large-scale complex problems, this paper proposes a parallel version of Cuckoo Search algorithm based on CUDA (compute unified device architecture). The GPU (graphic processing unit) thread blocks and threads are mapped to each individual and the dimensions of individual, respectively. Task parallelism and data parallelism combination is adopted to implement nest position update, individual fitness evaluation, rebuilding nest and searching for best individual operations of CS algorithm. The whole optimization process can be parallel implementation by GPU to reduce the communication overhead between CPU and GPU. The experimental results on 4 classical benchmark functions show that the parallel CS algorithm obtains 110 times speedup with the similar quality compared to a standard sequential CS algorithm.

Key words: cuckoo search, parallel computing, graphic processing unit (GPU), compute unified device architecture (CUDA)