Computer Science Research on Effective Parallel Programming

In computer science, we will focus on techniques that have immediate impact on applications being developed for petascale machines now, as well as on techniques that will change the state of the art for the future. Adaptive runtime systems (RTSs), including highly scalable dynamic load balancing will be needed as the number of processors increase and applications become more sophisticated, with adaptive techniques to handle evolving physical phenomena. New parallel programming abstractions are necessary to broaden the user-base of parallel machines, and are feasible as we better understand parallel programming and its usage patterns in CSE. With complex machines with deep memory hierarchies, auto-tuning techniques for libraries are required. Finally, to be able to debug and optimize applications at such a large scale, highly productive integrated development environments (IDEs) are essential.
 

Computer Science Information

Areas of Interest: