摘要
为量化分析GPU加速技术在城市雨洪过程模拟计算中效率提升效果,采用基于GPU技术的数值模型,对不同降雨及网格分辨率情形下的典型城市雨洪过程进行模拟,并对比分析GPU加速计算效率及其规律。结果表明:城市雨洪过程模拟中,GPU加速技术较CPU技术具有优势,不同情形下可提速23.88~158.72倍;在大暴雨和高分辨率网格条件下,GPU模型的提速效果更加明显,相较于短重现期降雨及低分辨率网格情形提升0.29%~8.43%,表明该技术适用于大范围高分辨率问题的模拟。该成果可为城市洪涝过程快速模拟预报提供技术支持。
In order to quantitatively analyze the computational efficiency of GPU technology for urban rain-flood process numerical simulation,the numerical model based on GPU technology was used to simulate the process of rain and flood in typical cities with different rainfall and grid resolution,and to compare and analyze the efficiency and law of GPU acceleration.The results showed that GPU technology had more advantages than CPU for the simulation of rain-flood process in cities,which could increase the computational efficiency by 23.88~158.72 times,compared with CPU in different situations.In the case of heavy rainstorms and high-resolution grids,the speed-up effect of the GPU model was more obvious,and it kept about 0.29%~8.43%higher than the short return period rainstorms and low-resolution grids conditions.And this technology was proved suitable for simulation of large-scale and high-resolution problems.The model was thus considered as a technical support for rapid simulation forecasting of floods processes.
作者
龚佳辉
侯精明
薛阳
张大伟
张兆安
杨少雄
韩浩
GONG Jia-hui;HOU Jing-ming;XUE Yang;ZHANG Da-wei;ZHANG Zhao-an;YANG Shao-xiong;HAN Hao(State Key Laboratory Base of Eco-hydraulic Engineering in Arid Area,Xi’an University of Technology,Xi’an 710048,China;POWERCHINA Huadong Engineering Corporation Limited,Hangzhou 311122,China;China Institute of Water Resources and Hydropower Research,Beijing 100038,China)
出处
《环境工程》
CAS
CSCD
北大核心
2020年第4期164-169,175,共7页
Environmental Engineering
基金
国家自然科学基金项目(51609199)
陕西省水利科技项目(2017slkj-14)
中国水科院基本科研业务费项目(JZ0145B772017)。
关键词
GPU
城市雨洪过程
网格数
计算效率
GPU
urban rain-flood process
grid numbers
computational efficiency
