三维Navier-Stokes方程的差分-谱方法混合法在GPU上的实现与优化

J4 ›› 2012, Vol. 34 ›› Issue (8): 53-58.

三维Navier-Stokes方程的差分-谱方法混合法在GPU上的实现与优化

徐莹，徐磊

(上海超级计算中心，上海 201203)

收稿日期:2012-04-28 修回日期:2012-06-11 出版日期:2012-08-25 发布日期:2012-08-25
基金资助:
国家863计划资助项目（2012AA01A308）;自然科学基金青年基金资助项目（10902063）

Implementation and Optimization of the Spectralfinite Difference Hybrid Scheme for 3D NavierStokes Equations on GPU

XU Ying,XU Lei

(Shanghai Supercomputing Center,Shanghai 201203,China)

Received:2012-04-28 Revised:2012-06-11 Online:2012-08-25 Published:2012-08-25

摘要/Abstract

摘要：

差分-谱方法通常在槽道湍流的直接数值模拟中使用，本文主要研究差分谱方法在单GPU卡上的实现。由于GPU的硬件发展十分迅速，不同的GPU硬件对双精度计算的支持有所不同，本文首先验证GPU上数值计算的精度，用差分谱混合法求解标量扩散方程，并将GPU和CPU上获取的数值结果与解析解进行对比，以确定GPU上数值算法实现的精确度。标量扩散方程在Nvidia S2050单GPU卡上求解，获得接近20倍的加速比，三维不可压缩NavierStokes方程达到了25倍的加速比。

关键词: 异构平台, GPGPU, 并行算法, 不可压缩NavierStokes方程, 差分谱方法混合法

Abstract:

The approach of accelerating the applications with GPUs already delivers impressive computational performance compared to the traditional CPU.The hardware architecture of GPU is a significant departure from CPUs,hence the redesign and validation of the numerical algorithm are required.The spectralfinitedifference scheme is usually used in the direction, and the numerical simulation (DNS) of turbulent channel flows is studied.In order to validate the numerical accuracy,the scalar diffusion equation is first solved with this scheme,and the results from GPU and CPU are validated with the analytical solution.The performance study of the scalar diffusion equation shows at least 20X speedup.For the 3D full NavierStokes equation,the performance on GPU shows a 24X speedup.

Key words: heterogeneous platform;GPGPU;parallel algorithm;incompressible NavierStokes;spectralfinite difference hybrid scheme

徐莹，徐磊. 三维Navier-Stokes方程的差分-谱方法混合法在GPU上的实现与优化[J]. J4, 2012, 34(8): 53-58.

XU Ying,XU Lei. Implementation and Optimization of the Spectralfinite Difference Hybrid Scheme for 3D NavierStokes Equations on GPU[J]. J4, 2012, 34(8): 53-58.

参考文献

［1］TOP 500 List［EB/OL］.［20120410］.http://www.top500.org/lists/2011/06.
［2］nVIDIA［EB/OL］.［20120410］.http://www.nvidia.com/object/cuda_home_new.html.
［3］http://www.khronos.org/registry/cl/.
［4］Cant S.HighPerformance Computing in Computational Fluid Dynamics:Progress and Challenges［J］.Philosophical Transactions:Mathematical,Physical and Engineering Sciences.2002.360:12111225.
［5］Stam J.Stable Fluids［C］∥Proc of SIGGRAPH’99,1999:121128.
［6］Harris M, Baxter W, Scheuermann T, et al. Simulation of Cloud Dynamics on Graphics Hardware［C］∥Proc of the ACM SIGGRAPH/EUROGRAPHICS Conference on Graphics Hardwarej,2003:92101.
［7］Liu Y,Liu X,Wu E.RealTime 3d Fluid Simulation on Gpu with Complex Obstacles［C］∥Proc of the 12th Pacific Conference on Computer Graphics and Applications,2004:247256.
［8］Crane K,Llamas I,Tariq S.RealTime Simulation and Rendering of 3D Fluids［M］.AddisonWesley Professional, 2007.
［9］Zhao Y.Lattice Boltzmann Based Pde Solver on the GPU［J］. The Visual Computer,2008,24(5):323333.
［10］Li W,Wei X,Kaufman A. Implementing Lattice Boltzmann Computation on Graphics Hardware［J］.The Visual Comput, 2003,19(78):444456.
［11］Kuznik F,Obrecht C,Rusaouen G,et al.LBM Based Flow Simulation Using GPU Computing Processor［J］. Journal of Computers & Mathematics with Applications, 2010,59(7):23802392.
［12］Obrecht C, Kuznik F, Tourancheau B, et al. A New Approach to the Lattice Boltzmann Method for Graphics Processing Units［J］.Computers & Mathematics with Applications, 2010.61(12):36283638.
［13］Goodnight N,Woolley C,Lewin G,et al.A Multigrid Solver for Boundary Value Problems Using Programmable Graphics Hardware［C］∥Proc of the ACM SIGGRAPH/EUROGRAPHICS Conference on Graphics Hardware,2003:102111.
［14］Bolz J,Farmer I,Grinspun E,et al.Sparse Matrix Solvers on the GPU:Conjugate Gradients and Multigrid［C］∥Proc of SIGGRAPH’03, 2003:917924.
［15］Brandvik T,Pullan G.Acceleration of a 3D Euler Solver Using Commodity Graphics Hardware［C］∥Proc of the 46th AIAA Aerospace Sciences Meeting and Exhibit,2008,AIAA2008607.
［16］Elsen E, LeGresley P, Darve E. Large Calculation of the Flow over a Hypersonic Vehicle Using a GPU［J］.Journal of Computational Physics, 2008,227(24):1014810161.
［17］Moin P,Kim J. Numerical Investigation of Turbulent Channel Flow［J］.Journal of Fluid Mechanics, 1982,118:341377.
［18］Cortese T A, Balachandar S. High Performance Spectral Simulation of Turbulent Flows in Massively Parallelmachines with Distributed Memory［R］.Technical Report TAM Rep. 765,University of Illinois at UrbanaChampaign,1994.
［19］Garg R,Ferziger J H,Monismith S G.Hybrid Spectral Finite Difference Simulations of Stratified Turbulent Flows on Distributed Memory Architectures［J］.International Journal Numerical Methods in Fluids, 1997,24(11):11291158.
［20］Kim J,Moin P,Moser R D.Turbulence Statistics in Fully Developed Channel Flow at Low Reynolds Number［J］.Journal of Fluid Mechanics,1987,177:133166.
［21］http://www.nscctj.gov.cn/example/example_12.asp.
［22］Bell N,Garland M.Efficient Sparse MatrixVector Multiplication on CUDA［R］.NVIDIA Technical Report NVR2008004,NVIDIA Corporation,2008.
［23］NVIDIA:Cufft library［EB/OL］.［20120410］.http://developer.nvidia.com/cufft.

[1]	许晓阳, 王斯棋. 基于粒子分解的SPH并行算法研究与应用[J]. 计算机工程与科学, 2022, 44(06): 964-970.
[2]	刘杰, 龚春叶, 杨博, 郭晓威, 甘新标, 李胜国, 李超, 陈旭光, 肖调杰, 穆利安, 宋敏, 赵冬勇, 鞠羽中. YH-ACT：热工流体力学并行应用程序[J]. 计算机工程与科学, 2021, 43(01): 58-69.
[3]	李雯, 迟利华, 张会, 张哲, 刘杰, . 潜艇使用自航式声诱饵防御鱼雷模型并行计算方法研究[J]. 计算机工程与科学, 2020, 42(12): 2163-2168.
[4]	沈立，杨耀华，王志英. 通过部分Warp重组消除GPGPU控制流的不一致性[J]. 计算机工程与科学, 2019, 41(08): 1335-1342.
[5]	彭林，张鹏，方建滨，黄春，唐滔. 异构平台多流编程机制的性能模型研究[J]. 计算机工程与科学, 2019, 41(07): 1145-1154.
[6]	王俊吉1,2,朱朝艳1,2,4,陈建军1,2,郑澎3，5,徐权3. 基于无锁原子操作的多线程并行Delaunay三角化算法[J]. 计算机工程与科学, 2018, 40(05): 773-779.
[7]	章子凯，武继刚，姜文超，刘竹松. 容错处理器阵列的多逻辑列并行重构算法[J]. 计算机工程与科学, 2018, 40(01): 24-33.
[8]	李娟1，汤德佑1，傅娟2，3. 一种基于蚁群算法的生物序列并行比对方法[J]. 计算机工程与科学, 2017, 39(09): 1610-1616.
[9]	刘子骏1，何炎祥1,2，张军1,3，李清安1,2，沈凡凡1. 一种面向GPGPU的行为感知的存储调度策略[J]. 计算机工程与科学, 2017, 39(06): 1011-1021.
[10]	宋威，吉红蕾，李晋宏. 一种高效用项集并行挖掘算法[J]. J4, 2015, 37(03): 422-428.
[11]	舒兵，任秀江，张清波，陈芳园. 数据密集型应用在NVIDIA Fermi片内存储结构上的适应性分析[J]. J4, 2014, 36(04): 601-606.
[12]	刘云华，刘巍. 异构装配信息的统一表达方法研究[J]. J4, 2014, 36(01): 145-149.
[13]	杨菊，吴卓锋，王刚，刘晓光. 多核平台PAML并行算法研究[J]. J4, 2013, 35(9): 15-19.
[14]	周琛，陈振杰，张帅. 基于边界代数法的矢量栅格化并行算法设计与实现[J]. J4, 2013, 35(4): 37-41.
[15]	王多雷，龙汉，陈璇. 组合打击下一类面目标易损性的并行计算研究[J]. J4, 2013, 35(4): 42-46.