Fluids Engineering Division Summer Meeting

Hilton Waikoloa Village, Waikoloa, Hawaii

Jul 30-Aug 3, 2017

Jul 31-Aug 2, 2017


Program - Workshop


An Introduction to the NASA FUN3D Computational Fluid Dynamics Software

Instructor: Dr. Eric Nielsen, NASA Langley Research Center
Day/Date: Sunday, July 30, 2017
Time: 8:00 am – 12:00 pm
Price: $150 ($250 for both workshops)

NASA Langley Research Center's FUN3D software is an unstructured-grid computational fluid dynamics suite used to tackle complex aerodynamics problems in support of major national research and engineering efforts at NASA and among groups across US industry, other government agencies, and academia. This workshop will provide participants with guidance on how to install and execute FUN3D for common aerospace applications. Detailed instructions will be provided for compilation and installation of the software, gridding guidelines, solution and visualization basics for steady flows, as well as adjoint-based design optimization for such applications. A broad overview of more advanced features will also be included, covering topics such as error estimation and mesh adaptation, unsteady flow simulations, dynamic and overset grid applications, adjoint-based design optimization of unsteady flows, and multidisciplinary simulations. The material included here is a subset of a two-day short course offered periodically by the FUN3D development team.

Please note that FUN3D is export-controlled software and may only be provided to US persons. See more information.

8:00 - 8:15 am Introductions
8:15 - 8:45 am Welcome and Overview
8:45 - 9:00 am Compilation and Installation
9:00 - 10:30 am Gridding, Solution, and Visualization Basics
10:30 - 10:45 am Break
10:45 - 12:00 pm Adjoint-Based Design for Steady Flows

Computational Fluid Dynamics on GPUs: Architecture, Algorithms and Implementation

Instructor: Prof. Pratap Vanka, University of Illinois, Urbana-Champaign
Day/Date: Sunday, July 30, 2017
Time: 1:00 pm – 5:00 pm
Price: $150 ($250 for both workshops)

With the explosive growth of the need for computing power, many core processors are becoming the norm in the new computing architectures. One particular advance has been the use of Graphics Processing Units (GPUs) for scientific computing. A GPU is a very compact processor containing thousands of cores with specialized architecture and memory access. Theoretically, a GPU is rated to provide teraflop speeds in a small footprint. However, in order to exploit the full speed, various constraints are encountered, particularly the design of the algorithm and the access of the data from the memory. Practically speed-ups of up to 80 or 100 can be obtained by a small desk side box housing 4 GPUs.

In this short course, details of the architecture, CFD algorithms and implementations, including multiple GPU programming will be covered. Typical CFD algorithms will be explained and their implementations will be shown. The total course is currently planned for 4 hours, as follows.

  1. What is a GPU? GPU architecture, theoretical performance and memory layout
  2. Programming a GPU: CUDA and Fortran
  3. Data parallel algorithms for CFD
  4. Examples of CFD algorithms for GPUs
  5. Multi-GPU programming
  6. Other emerging processors (Intel PHI)
  7. Summary

About the Instructor

Pratap Vanka is Professor Emeritus and Research Professor in the department of Mechanical Science and Engineering at the University of Illinois. He has more than 40 years of experience in developing CFD algorithms and implementing them on advanced parallel computers. He recently gave the Freeman scholar lecture, titled "Computational Fluid Dynamics on GPUs". His group has developed several CFD methodologies (finite volume, LBM, molecular dynamics, multigrid algorithms) on GPUs, including multiple GPUs. Prof. Vanka is a Fellow of ASME and the recipient of several teaching and research awards. A similar course was taught in 2014 and was very well received by the attendees.

Workshop on Extract-based and In Situ Methods for HPC Enabled CFD

Instructor: Steve M. Legensky – Founder and General Manager of Intelligent Light
Day/Date: Thursday, August 3, 2017
Time: 1:30 pm – 4:00 pm
Price: Complimentary

This workshop will assess the state-of-the-art in CFD Visualization and Post-Processing in High Performance Computing (HPC) environments. Specifically the issues of data transfer from remote HPC resources, a comparison of remote graphics vs. extract methods, and the trend toward in situ methods. Real world examples of industrial CFD workflows will be presented that offer a path to greater fidelity and faster throughput enabling an increased use of unsteady simulation.


Prof. Kozo Fujii, Tokyo University of Science/ ISAS-JAXA Revisit the Role of Research Graphics in HPC Post Processing

Yves-Marie Lefebvre, FieldView Product Chief, Intelligent Light Unblocking Data Bottlenecks in HPC CFD with Extract-based Workflows

Steve Legensky, Founder and General Manager, Intelligent Light In Situ: Real and in Production