CUDA and GPUs for CFD Applications

16 May 2011

Tutorial for Parallel CFD 2011, Led by Mr. Tom Reed, NVIDIA

09:00 – 10:30            1. Introduction to CUDA

1.1  Overview of GPU computing

1.2  Fundamentals of a CUDA C program

10:30 – 11:00    Break

11:00 – 12:30            2. CUDA Optimization Techniques

2.1  Overview of CUDA development tools

2.2  Optimization fundamentals for  Fermi

2.3  Global and shared memory utilization

2.4  CPU-GPU interaction

12:30 – 14:00   Lunch

14:00 – 15:00            3. Overview of CUDA 4.0 with Code Examples

3.1  Ease in application porting

3.2  New and improved development tools

3.3  Improved multi-GPU programming

3.4  Select CUDA code examples

15:00 – 15:45            4. CUDA Strategies for Legacy CFD Software

4.1  Case Study I: PORTING OF FEFLO TO MULTI-GPU CLUSTERS (45 min)

-      Prof. Dr. Rainald Lohner, George Mason University, USA

15:45 – 16:00  Break

16:00 – 18:00            4. CUDA Strategies for Legacy CFD Software (Cont.)

4.2  Case Study II: WARIS: A General strategy for Computational Mechanics on GPUs (45 min)

-      Dr. Mauricio Hanzich and F. Rubio, Barcelona Supercomputing Center, Spain

4.3  HMPP WORKBENCH– Directive-based compiler for hybrid computing

-      CAPS Enterprise, France (30 min)

4.4  PGI Accelerator Compilers– Directive-based compiler for GPU computing

-      Portland Group, USA (30 min)

4.5  Discussion and Closing Remarks

Mr. Tom Reed

HPC and Visualization Solution Architect, NVIDIA, Houston, TX, USA

Tom Reed currently works with NVIDIA’s key customers and OEMs to optimize the benefits of GPU technology. Prior to joining NVIDIA, Mr. Reed spent 20 years with Silicon Graphics where his tenure included roles in systems engineering, benchmarking & performance engineering, software development, and professional services. Mr. Reed’s career has always involved some combination of high performance computing (HPC) and visualization and he’s long been an advocate for the fusion of best-of-breed technologies to drive sustainable and efficient advances in IT capability. Past research work has included initiatives in interactive parallel computing, data-centric systems design, and hybrid computing architectures. In addition to SGI, Mr. Reed has worked for McDonnell Douglas, General Dynamics, and Linuxcare as well as research contracts with Kirkland Air force Base, Arnold Engineering Development Center (AEDC), and NASA. Mr. Reed attended Iowa State University for both undergraduate and graduate studies, majoring in aerospace engineering with graduate research in computation fluid dynamics (CFD).

Mr. Douglas Miles

Director, PGI

Mr. Miles is responsible for all business and technical operations of The Portland Group (PGI). He has worked in various positions over the last 25 years in HPC applications engineering, math library development, technical marketing and engineering management at organizations FPS, Cray Research, Superservers, PGI and STMicroelectronics.

Prof. Dr. Rainald Löhner

Distinguished Professor of Fluid Dynamics, School of Computational Sciences, George Mason University

Rainald Löhner is the head of the CFD center at the department of computational and data sciences of George Mason University in Fairfax, VA, in the outskirts of Washington, D.C. He received a M.Sc. in Mechanical Engineering from the Technische Universität Braunschweig, Germany, as well as a Ph.D. and D.Sc. in Civil Engineering from the University College of Swansea, Wales, where he studied under Prof.’s Ken Morgan and Olgierd Zienkiewicz. His areas of interest include numerical methods, solvers, grid generation, parallel computing, visualization, pre-processing and fluid-structure interaction. He is the author of more than 600 articles covering the fields enumerated above, as well as a textbook on Applied CFD Techniques. The codes developed by Prof. Löhner have been used extensively for high-fidelity simulations of blast-structure interaction, store separation, ship and submarine flows, contaminant transport in urban areas, flows in the brain and the bronchial system, as well as optimum shape and process design in the automotive manufacturing industries.