本文已被：浏览 2503次 下载 5608次
Received:April 29, 2009 Revised:May 20, 2010
Received:April 29, 2009 Revised:May 20, 2010
中文摘要: 本文将计算流体力学中的可压缩的纳维叶－斯托克斯(Navier-Stokes)，不可压缩的Navier-Stokes 和欧拉（Euler）方程移植到NVIDIA GPU 上。模拟了3 个测试例子，2 维的黎曼问题，方腔流问题和RAE2822 型的机翼绕流。相比于CPU，我们在GPU 平台上最高得到了33.2 倍的加速比。为了最大程度提高代码的性能，针对GPU 平台上探索了几种优化策略。和CPU 以及实验结果对比表明，利用计算流体力学在GPU 平台上能够得到预想的结果，具有很好的应用前景。
Abstract:Computational Fluid Dynamic (CFD) codes based on incompressible Navier-Stokes, compressible Euler and compressible Navier-Stokes solvers are ported on NVIDIA GPU. As validation test, we have simulated a two-dimension cavity flow, Riemann problem and a transonic flow over a RAE2822 airfoil. Maximum 33.2x speedup is reported in our test. To maximum the GPU code performance, we also explore a number of GPU-specific optimization strategies. It demonstrates GPU code gives the expected results compared CPU code and experimental result and GPU computing has good compatibility and bright future.
keywords: GPU computing CUDA CFD
文章编号： 中图分类号： 文献标志码：
基金项目:基金项目:中国科学院知识创新工程青年人才领域项目(0815011103)
引用文本：
董廷星,李新亮,李森,迟学斌.GPU 上计算流体力学的加速.计算机系统应用,2011,20(1):104-109
DONG Ting-Xing,LI Xin-Liang,LI Sen,CHI Xue-Bin.Acceleration of Computational Fluid Dynamics Codes on GPU.COMPUTER SYSTEMS APPLICATIONS,2011,20(1):104-109
董廷星,李新亮,李森,迟学斌.GPU 上计算流体力学的加速.计算机系统应用,2011,20(1):104-109
DONG Ting-Xing,LI Xin-Liang,LI Sen,CHI Xue-Bin.Acceleration of Computational Fluid Dynamics Codes on GPU.COMPUTER SYSTEMS APPLICATIONS,2011,20(1):104-109
