As the HPC community enters the new decade, it is also approaching an important crossroads where the systems, models, and methods that have driven exponential performance growth for the past 20 years will likely yield to innovative alternatives in order to achieve Exascale performance before 2020. While the introduction of multicore and heterogeneous system architectures have started this shift, these are just the first of many new developments in parallel computing that will make possible the needed multi-billion-way concurrency at the application, architecture, and system software levels. More than ever, parallelism will dominate progress in delivered performance as clock rates and processor complexity reach their peak - or even decline in the face of power constraints.
This session will present likely driving issues, technology trends and future architecture characteristics along with possible impacts on programming models that will characterize the next ten years in parallel computing. The four session speakers will address the critical problems and offer novel solutions both of which may determine the future of parallel computing in the years to come:
- Prof. Dr. Thomas Sterling, Arnaud & Edwards Professor of Computer Science, Louisiana State University, USA:
"Parallel Computing in the Years to Come"
More information is available at www.isc10.org
- Prof. Dr. Satoshi Matsuoka, Professor, Tokyo Institute of Technology, Japan: "Massively Parallel Computing for the Future: Custom 'Formula One' or Cloud 'Prius'?"
- Prof. Dr. Thomas Schulthess, Director, Swiss National Supercomputing Center, Switzerland: "Millions of CPU Cores & Beyond: A User's Wish List for Future Architectures"
- Steve Wallach, Chief Scientist, Convey Computer, USA: "Software: The 'Trojan Horse' of HPC"