Research :: ECE :: UD


Computer Systems

Context-Aware and Context-Adaptive Code Optimization for New General High-performance Computers

Xiaoming Li

The new high-performance computing platforms, such as Chip Multi Processors (CMP) and Graphic Processing Units (GPU), designates everyone a user of high-performance computing. The new high-performance computing platforms will likely include a large number of simultaneously multi-hreaded processor cores on a chip and resources that are shared among the cores.

Unlike the traditional scientific-computing-oriented high-performance computers, the new high-performance computers will simultaneously run multiple programs and a large number of threads competing for shared resources, such as caches. This presents unprecedented challenges for code optimization of the new general high-performance computers. First, the ubiquitous existence of multi-program and multi-thread makes a program run in a runtime environment that differs significantly from that of traditional high-performance computers.

Threads have extensive resource competition among themselves. Second, the execution environment is always changing during the life of a program. Consequently, a program that is optimized in only one way will not perform equally well for the changing runtime environment. To tackle the challenges, our group will develop program optimization and runtime adaptation techniques to address the unique challenges facing the new high-performance computers that must be solved in order to ensure their successes.

Recent publications

Haiping Wu, Eunjung Park, Mihailo Kaplarevic, Yingping Zhang, Murat Bolat, Xiaoming Li, Guang R. Gao, "Automatic Program Segment Similarity Detection in Targeted Program Performance Improvement", Workshop on Performance Optimization for High-Level Languages and Libraries, in conjunction with 21st IEEE International Parallel & Distributed Processing Symposium (IPDPS), March 2007.

Bookmark and Share