Seminars & Colloquia
The College of William and Mary
"Enhancing Data Locality for Massive Parallelism"
Wednesday March 20, 2013 09:30 AM
Location: 3211, EBII NCSU Centennial Campus
(Visitor parking instructions)
Massive parallelism is exhibited at various levels of a system, ranging from a data center to a manycore processor. It is the foundation for providing the computing power required by Big Data applications. As parallelism continues increasing fast, bandwidth expansion in memory and network lags behind, causing an ever growing gap between bandwidth and computing power. Consequently, effectively bringing data to cores is one of the most critical challenges for tapping into the potential of future systems. It is also a key to power efficiency as data movements will be responsible for the major power consumption in future computing systems. This talk discusses the important role of data locality enhancement in meeting the challenges. It examines the implications massive parallelism brings to data locality, and presents some recent findings in addressing them. In addition, this talk will describe some novel inter-disciplinary approaches to enabling scalable data processing.
Xipeng Shen is the Adina Allen Term Distinguished Associate Professor in the College of William and Mary, an IBM Center for Advanced Studies (CAS) Faculty Fellow, a Visiting Researcher at Microsoft Research, Intel Lab, and MIT. His research in data locality and massive parallelism won the prestigious Early Career Research Award from the US Department of Energy in 2011 and the Best Paper Award at ACM PPoPP 2010. His research in input-centric program dynamic optimizations won the CAREER Award from the US National Science Foundation in 2010.
Xipeng Shen's research lies in the broad field of programming systems, with an emphasis on enabling extreme-scale data-intensive computing and intelligent portable computing through innovations in both compilers and runtime systems. He has been particularly interested in capturing large-scale program behavior patterns, in both data accesses and code executions, and exploiting them for scalable and efficient computing in a heterogeneous, massively parallel environment.
Host: Frank Mueller, Computer Science, NCSU