RESEARCH

Short Vitas

Representative Publications

Patents

Editorial Board Membership

International Journal on Parallel Programming (IJPP)

 

PARALLEL COMPUTING IN THE EXAFLOPS ERA

Computer simulation continues to be an efficient tool to model and study complex phenomena which are otherwise too expensive, too dangerous, or impossible to experience. In many computational science and engineering disciplines scientists are interested in exploring and discovering biological, climate, and high energy physical phenomena, which require immense computing power. Recent successful advances in massively-parallel computing culminated in integrating thousands of compute nodes in a single supercomputer, each node typically made up of several processor cores which communicate using high speed interconnection networks. To perform massively parallel simulations, nodes typically use massive thread-level parallelism to hide the latency of memory and inter-node communication.  By working together, the nodes can perform Petaflop calculations that would require millennia to perform on personal computers. The next challenge is to achieve the Exaflops supercomputing power that will be used to provide deeper insights in many disciplines. The approach to achieve the challenge is a converging strategy through advances in programming a class of hierarchical algorithms as well as architectural challenges. The COE research on parallel computing is committed to contribute to the above challenges in all areas of computer architecture and high-performance computing.

HIGH-PERFORMANCE COMPUTING RESEARCH GROUP

DESIGN OF AN INTELLIGENT TELEROBOTIC SYSTEM

GROUP OF INTEREST ON ROBOCUP