摘要
在CUDA(Compute Unified Device Architecture)架构下将GPU(Graphic Processing Unit)计算首次应用到海洋环流模式POP(Parallel Ocean Program)中。测试结果表明:无论高分辨率还是低分辨率,GPU都能够提高海洋环流数值模式POP的计算速度,GPU加速比最低都在1.5倍以上,最高可以超过2.2倍;并且随着模式使用线程数目的增多,GPU的加速比在降低,但是GPU利用效率在增长。
In this paper,the GPU（Graphic Processing Unit） based on the CUDA（Compute Unified Device Architecture） framework is applied at first time to ocean general circulation mode POP（Parallel Ocean Program）.Testing results show that the GPU can improve calculation speed of the ocean general circulation value mode POP,no matter the resolution is high or low.The lowest speedup ratio is higher than 1.5 and the maximum speedup ratio can exceed 2.2.Moreover,along with the increase of numbers of thread used in the mode,the utilisation efficiency of GPU is increasing while the speedup ratio is decreasing.
出处
《计算机应用与软件》
CSCD
2010年第10期27-29,共3页
Computer Applications and Software
基金
国家自然科学基金(40906018)
