Main navigation

Loading Events

« All Events

  • This event has passed.

Computing Fluid Flows on Multi-GPU Cluster using Lattice Boltzmann Method

May 23, 2016 @ 12:00 pm - 1:00 pm

 IDRE Seminar- GPU/Many cores program
Computing Fluid Flows on Multi-GPU Cluster using Lattice Boltzmann Method

Chao-An Lin
Department of Power Mechanical Engineering,
National Tsing Hua University, Hsinchu 30013 Taiwan

Date and Time: May 23, 2016 at 12:00 PM (Lunch will be served*)
Location: 5628 Math Sciences Bldg.
RSVP**: http://cfapps.ats.ucla.edu/cfapps/events/rsvp/RSVPNow.cfm?EveID=3531&SecID=3518

Abstract:
Due to its enhanced computational capability, the graphic processing unit has drawn attention for non-graphic applications. The lattice Boltzmann method (LBM) as an explicit numerical scheme requiring only neighboring operations, is very suitable for parallel or GPU computations. Thus, GPU has been successfully used for lattice Bolztmann computations, which demonstrated that the computational power of GPU far exceeds that of PC-based CPU. There are several strategies to further improve the GPU performance, such as reducing the data transaction between host and device, and using efficient memory management. Utilizing shared memory was shown to increase the performance of GPU. Another way to increase the performance is adopting different streaming strategy to optimize the data transfer between the GPU global memory and shared memory. On the other hand, multi-GPU computation can certainly elevate the performance. This can be achieved by using multi GPUs on a single node through OpenMP. Alternatively, for cross node GPU computations, Message Passing Interface (MPI) on cluster of GPUs can be employed. Here, issues, such as the memory management and the latency during multi-GPU computations are addressed to seek possible enhancement of the computational efficiency. Numerical examples investigated are turbulent duct and channel flows and two phase flows of colliding liquid droplets.

Speaker’s Bio:
Professor Chao-An Lin received his BSc in Mechanical Engineering from National Chiao Tung University, Taiwan and M.Sc. and Ph.D. in Mechanical Engineering from University of Manchester Institute of Science and Technology (UMIST), UK in 1986 and 1991. After one year postdoctoral work at UMIST, in 1991 Dr. Lin joined the faculty of the department of Power Mechanical Engineering at National Tsing Hua University in Taiwan, where he also served as the deputy chair (2007-2009) and department chair (2009 -2015).  He is also the honorary visiting professor at University of Liverpool, UK (2015-2019).
Professor Lin’s research interests include flow physics and turbulence modeling, bio-medical fluid dynamics and development of efficient numerical methods and in particular the development of immersed boundary method, lattice Boltzmann method and simulations using multi-GPU cluster. Professor Lin is the associate fellow of American Institute of Aeronautics and Astronautics and is also members of the executive committees of Taiwan society of mechanical engineer and Taiwan society of Aeronautics and Astronautics.  Dr. Lin was the recipient of Shila Mo memorial prize of UMIST (UK) in 1991, and was awarded the distinguished teacher award from ministry of education of Taiwan in 2008. He also serves in different committees of international conferences (Parallel CFD, Turbulent Shear Flow Phenomenon, ERCOFTAC Symposium on Engineering Turbulence Modeling and Measurements, Asian Symposium on Computational Heat Transfer and Fluid Flow, Asia CFD). He edited special issues in Computers and Fluids, and Computers and Mathematics with Applications, and Energies. Prof. Lin is in the advisory board of Journal of Computational Mechanics and serves as associate editor of Journal of mechanics.

*Lunch will be ready at 11:45 AM.
**To ensure you have a space at the seminar, please RSVP ONLINE by May 18, 2016.

Venue

Kerckhoff – Charles E. Young Grand Salon
308 Westwood Plaza
Los Angeles, 90095 United States
Website:
http://maps.ucla.edu/campus/?locid=268

Events List Navigation

UCLA OIT