SC Conference - Activity Details

Can Developing Applications for Massively Parallel Systems with Heterogeneous Processors Be Made Easy(er)?

Tarek El-Ghazawi  (George Washington University)

David Patterson  (University of California, Berkeley)
Marc Snir  (University of Illinois at Urbana-Champaign)
David Bader  (Georgia Institute of Technology)
Vivek Sarkar  (Rice University)
John Shalf  (Lawrence Berkeley National Laboratory)
Panels Session
Tuesday,  03:30PM - 05:00PM
Room Ballroom G
HPC systems of the near future will be quite heterogeneous containing multi-core processors and application-specific many-core co-processors. Developing applications for such systems is a major challenge as the currently widely-used programming models and tools are not suitable to support the diverse computing resources that will be available in these systems. This panel will look at what needs to be done in order to make the application development for massively parallel systems with heterogeneous processors easier. The panel will address the following questions: Are there better ways to develop algorithms and applications for heterogeneous multicore systems than today's existing approaches? Can an efficient unified programming model(s) be developed to support heterogeneous multicore systems? What performance measurement, analysis, and prediction models and tools are needed to support application development on heterogeneous multicore systems? How applications that spread across multiple multi-core processors and many-core co-processors can be monitored and debugged in a scalable way?
   IEEE Computer Society  /  ACM     2 0   Y E A R S   -   U N L E A S H I N G   T H E   P O W E R   O F   H P C