A Hybrid Optimization Based on GPU Parallel Computing Method and its Application of Three Dimensional Large Eddy Simulations

Abstract:

Using compute Unified device architecture (CUDA), a traditional computational fluid dynamics (CFD) program is paralleled and optimized based on graphic processing unit (GPU). The calculation process is divided into two parts as serial and parallel. Their main characteristics are analyzed and different optimization schemes are given. CPU (central processing unit) and GPU work respectively as flow control and high-speed parallel computation. Bandwidth between devices is applied effectively. Data transfer between devices is moderately improved to simplify algorithm. Finally, the method is verified by simulating a three-dimensional isotropic homogeneous turbulence flow field. The calculation uses large eddy simulation (LES) method with secondary filter and solves the three-dimensional N-S equations. The maximum grid number achieves 8,000,000 and takes 33 seconds each step. All calculations are using ordinary single desktop computer, optimized acceleration ratio can reach 9.

• Periodical: Applied Mechanics and Materials (Volumes 275-277)
• Pages: 2589-2594
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.275-277.2589
• Online since: January 2013
• Authors: Yu Xuan Zhang*, Song Ping Wu
• Keywords: CUDA, Graphic Processing Unit (GPU), LES, Optimize, Parallel Compute, Secondary Filter

References

[1] Michael Resch: A Comparison of OpenMP and MPI for the Parallel CFD test Case. High Performance Computing Center Stuttgart (1999).

[2] M. Jahed Djomehri: Hybrid MPI OpenMP Programming of an Overset CFD Solver and Performance Investigations. NASA Technical Reports Server (2002).

[3] Li X M: Numerical Parallel Algorithm and Software. Wu J P. Beijing: Science Press, (2007) (In Chinese).

[4] nVidia: NVidia CUDA Getting Started for Microsoft Windows. American: nVidia, (2012).

[5] Zhang B: Parallel Computing Methods for CFD Using a GPU and Implicit Scheme. Acta Aeronoutica et Astronautica Sinica (2010), 31(2): 249-256 (In Chinese).

[6] Dong Y X: Acceleration of Computational Fluid Dynamics Codes on GPU. Computer Systems & Applications (2011), 20(1): 104-109 (In Chinese).

[7] nVidia: NVIDIA CUDA C Programming Guide. American: nVidia (2012).

[8] nVidia: SDK Code Sample Guide to New Features in CUDA Toolkit v4. 2. American: nVidia (2012).

[9] Grama A: Introduction to Parallel Computing (Second Edition). Zhang W, translated. Beijing: China Machine Press (2005).

[10] Fu D X: Direct Numerical Simulation of compressable turbulent flow. Ma Y W. Beijing: Science Press (2005) (In Chinese).

[11] Li B: The Study of Sub-Grid Model for Compressible Turbulent Flows. Beijing: School of Aeronautic Science and Engineering, Beihang University (2007) (In Chinese).

[12] Bardina J: Improved subgrid model for large-eddy simulation. AIAA (1980), paper 801357.

[13] J. Smagorinsky: General circulation experiments with the primitive equations. Monthly Weather Review (1963), 91(3): 99-164.

DOI: 10.1175/1520-0493(1963)091<0099:gcewtp>2.3.co;2

[14] Zhang Z S: Large eddy simulation of turbulent flows: Theory and Application. Cui G X. Beijing: Tsinghua University Press, (2008) (In Chinese).

[15] M. P. Martin: A bandwidth-optimized WENO scheme for the effective direct numerical simulation of compressible turbulence. Journal of Computational Physics, 2006, 220(2006): 270-289.

DOI: 10.1016/j.jcp.2006.05.009

[16] Jiang G S: Efficient implementation of weighted ENO schemes. Journal of Computational Physics (1996), 126: 202-228. (In Chinese).

DOI: 10.1006/jcph.1996.0130

[17] Shu C W: Efficient Implementation of Essentially Non-oscillatory Shock-Capturing Schemes. Journal of Computational Physics (1988), 77: 439.

DOI: 10.1016/0021-9991(88)90177-5

[18] Li X L: Direct Numerical Simulation of isotropic homogeneous turbulent flow. Science in China (2002), 32(8): 716-724 (In Chinese).