Direct Numerical Simulations of Hypersonic Turbulent Flows Accelerated by GPU

Abstract

Hypersonic turbulent flows, such as hypersonic turbulent boundary layer transitions and hypersonic shock wave/turbulent boundary layer interactions, are typical phenomena in high-speed vehicles. As a powerful tool, Direct numerical simulation (DNS) plays an important role in the study of turbulent flows. However, the demand of DNS for computing resource is very large, especially for hypersonic flows. The Graphics processor unit (GPU) system offers strong computing performance, enabling accelerated scientific computing. In recent years, the author’s group has developed a set of high-order heterogeneous parallel code, OpenCFD-SCU, which has achieved a speed-up ratio of over 700x compared with the CPU system (a single GPU card versus a single CPU core). In this seminar, Prof. Li will present the DNS works of hypersonic turbulent flows by using the GPU-based code. The following are the details:
1) Development and optimization of high-order GPU code OpenCFD-SCU; 
2) DNS studies of hypersonic turbulent boundary layer transitions;
3) DNS studies of hypersonic shock wave/turbulent boundary layer interactions.

Speaker

Prof. Xinliang Li obtained his PhD from the Institute of mechanics, Chinese Academy of Sciences in 2000. Following that, he served as a postdoctoral researcher at Tsinghua University and Tokyo University of Technology. Currently, Prof. Li holds the position of Professor at the State Key Laboratory of High-Temperature Gas Dynamics (LHD), Institute of Mechanics, Chinese Academy of Sciences (CAS).  Prof. Li is a council member of China Aerodynamics Society, and he is also an Associate Editor of Computers & Fluids. Additionally, Prof. Li is a member of editorial board for the Chinese Journal of Aerodynamics. His primary research interests include computational fluid dynamics (CFD), turbulent flows and high-performance computing (HPC).