基于GPU加速的Boussinesq类波浪传播变形数值模型  

A GPU accelerated Boussinesq wave propagation model

作　　者：孙家文[1,2] 朱桐 房克照[1] 刘忠波 SUN Jiawen;ZHU Tong;FANG Kezhao;LIU Zhongbo(The State Key Laboratory of Coastal and Offshore Engineering,Dalian University of Technology,DUT-UWA Joint Research Centre for Ocean Engineering,Dalian 116024,China;National Marine Environmental Monitoring Center,Dalian 116023,China;School of Traffic and Transportation Engineering,Dalian Maritime University,Dalian 116026,China)

机构地区：[1]大连理工大学海岸和近海工程国家重点实验室,DUT-UWA海洋工程联合研究中心,辽宁大连116024 [2]国家海洋环境监测中心,辽宁大连116023 [3]大连海事大学交通运输工程学院,辽宁大连116026

出　　处：《海洋工程》2020年第2期111-119,共9页The Ocean Engineering

基　　金：国家重点研发计划项目资助(2017YFC1404200);国家自然科学基金资助项目(51579034,51779022,51809053);中央高校基本科研业务费资助(No.DUT18ZD214)。

摘　　要：Boussinesq波浪模型是一类相位解析模型,在时域内求解需要较高的空间和时间分辨率以保证计算精度。为提高计算效率,有必要针对该类模型开展并行算法的研究。与传统的中央处理器(CPU)相比,图形处理器(GPU)有大量的运算器,可显著提高计算效率。基于统一计算设备架构CUDA C语言和图形处理器,实现了Boussinesq模型的并行运算。将本模型的计算结果同CPU数值模拟结果和解析解相比较,发现得到的结果基本一致。同时也比较了CPU端与GPU端的计算效率,结果表明,GPU数值模型的计算效率有明显提升,并且伴随数值网格的增多,提升效果更为明显。Boussinesq wave model is phase-resolving and requires high spatial and temporal resolution to ensure the calculation accuracy.It is thus necessary to study parallel algorithms for this kind of model in order to improve the computational efficiency.Compared with the traditional central processing unit(CPU),the graphics processing unit(GPU)has a large number of arithmetic units and can bring significant decrease in computational time.The parallel computation of Boussinesq model is implemented by CUDA C language and GPU in the present study.The computed results from GPU simulation are compared with those from CPU simulation and the analytical solutions,and they are found to be in reasonable consistence.The computational efficiency between CPU simulation and GPU simulation is also compared;the results show that the computational efficiency of GPU simulation is obviously improved with the increase of grid number.

关 键 词：BOUSSINESQ方程 图形处理器 CUDA C 并行计算 计算效率 波浪传播 

分 类 号：O353.2[理学—流体力学]