ParCFD 2009 Conference Panel
PetaFLOPS and Beyond
Moderator: Ron Bailey, NASA Ames Research Center
- Michael Aftosmis, NASA Ames Research Center
- David Emerson, Daresbury Laboratory, UK
- John Grosh, Lawrence Livermore National Laboratory
- John Shalf, Lawrence Berkeley National Laboratory
- Suga A. Sugavanam, IBM Systems & Technology Group
The purpose of this panel is to explore the question of how CFD applications can exploit computing systems with performance potential of a petaFLOPS and beyond. Due to limits on switching speeds and power consumption, high-performance computers are moving into a new era dominated by many-core chips and heterogeneous architectures. As their speeds advance beyond petaFLOPS and toward exaFLOPS, computers can be expected to reach one million cores and perhaps more. Thus, it seems clear that new programming approaches will be needed, and perhaps, new algorithms if we are to realize significant real performance for CFD applications.
Suggested Questions to the Panel:
- What will future architectures and systems look like as performance extends from petaFLOPS toward exaFLOPS?
- What are the dominant CFD applications that could benefit from petaFLOPS and beyond computing?
- What are the algorithmic and programming challenges that must be overcome to efficiently exploit petaFLOPS and beyond hardware to solve CFD and other similar problems?
- What should the CFD community do to prepare to address challenges presented by petaFLOPS and beyond systems, and to exploit their use?