首页 | 本学科首页   官方微博 | 高级检索  
     

波形板干燥器内液滴模拟的GPU加速研究
引用本文:张迪,黄伟,罗琦,王侃. 波形板干燥器内液滴模拟的GPU加速研究[J]. 原子能科学技术, 2015, 49(3): 491-496. DOI: 10.7538/yzk.2015.49.03.0491
作者姓名:张迪  黄伟  罗琦  王侃
作者单位:1.清华大学 工程物理系,北京100084;2.中国核动力研究设计院 核反应堆系统设计技术重点实验室,四川 成都610041;3.中国核动力研究设计院,四川 成都610041
摘    要:基于CUDA C,使用GPU模拟了波形板内大量液滴的二维运动。根据GPU硬件特性,优化了模拟算法。在CPU上实现相同算法,对比了单一液滴直径、不同液滴数量下GPU与CPU的计算时间。结果表明:液滴数量越大,GPU并行加速效果越明显,在液滴数量较大时,得到约60倍的加速。用GPU分析了典型液滴直径分布的情况,得到约30倍的加速。

关 键 词:波形板干燥器   液滴运动模拟   GPU并行加速   CUDA C

GPU Accelerated Droplets Simulation in Wave-type Vane Dryer
ZHANG Di;HUANG Wei;LUO Qi;WANG Kan. GPU Accelerated Droplets Simulation in Wave-type Vane Dryer[J]. Atomic Energy Science and Technology, 2015, 49(3): 491-496. DOI: 10.7538/yzk.2015.49.03.0491
Authors:ZHANG Di  HUANG Wei  LUO Qi  WANG Kan
Affiliation:1.Department of Engineering Physics, Tsinghua University, Beijing 100084, China;2.Science and Technology on Reactor System Design Technology Laboratory, Nuclear Power Institute of China, Chengdu 610041, China;3.Nuclear Power Institute of China, Chengdu 610041, China
Abstract:Droplets’ two-dimensional dynamic simulation was performed by GPU based on CUDA C. The simulation algorithm was optimized considering hardware characteristics of GPU parallel computation. The same algorithm was carried out by CPU. The computing time of GPU was compared to that of CPU for the cases with single droplet diameter but different droplet numbers. The results show that the larger the number of water droplets is, the more obvious the GPU parallel acceleration effect is. In the case with large number of water droplets, about 60 times speedup is achieved. And the computing time of GPU for the case with a typical distribution of multiple droplet diameters obtains about 30 times speedup.
Keywords:wave-type vane dryer  droplet dynamic simulation  GPU paraller acceleration  CUDA C
本文献已被 CNKI 等数据库收录!
点击此处可从《原子能科学技术》浏览原始摘要信息
点击此处可从《原子能科学技术》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号