Time: 10:30-12.:00am, August 31th, 2011 (Wednesday)

Place: 1201 Room, ICT, CAS

Speaker: Dr. Xingfu Wu

 Abstract:

The current trend in high-performance computing places a great focus on minimizing the power consumption of scientific applications on multicore systems without increasing runtime performance. Performance models can be used to provide insight into the application's performance characteristics that　 significantly impact the runtime and power consumption. As HPC multicore systems become more complex it is important to understand the relationship between performance and power consumption and the characteristics of scientific applications that influence the various levels of performance.In this talk,I will briefly discribe our ongoing NSF-funded large project MuMI (Multicore application Modeling Infrastructure),which facilitates systematic measurement, modeling, and prediction of performance, power consumption and performance-power tradeoffs for multicore systems. The MuMI framework is an integration of existing frameworks:Prophesy,PowerPack,and PAPI. Specifically, I'll discuss data collection: MAIDE system,which automatically transferes performace data collected into MuMI Database,provides automated instrumentation at the multiple levels.Then I will talk about MuMI database in detail,it was organized into 4 areas:application,executalbe,run,performance statistics. Finally,I specifically describe web-based automated power and performance modeling System,such as developing performance models,web-based modeling system,curve fitting model,parameterization model,kernel coupling model.

 bio:

Dr. Xingfu Wu is a TEES Research Professor at Texas A&M University. He is a senior ACM member and an IEEE member. His research interests are performance evaluation and modeling, parallel and cloud computing, and power and energy analysis in HPC systems. He served as session chairs and PC members for several international conferences, and was a guest editor of IEEE Distributed Systems Online Special Issue on Data-intensive Computing (Vol. 5, Issue 1, 2004). His monograph: Performance Evaluation, Prediction and Visualization of Parallel Systems was published by Kluwer Academic Publishers (ISBN 0-7923-8462-8) in 1999. He won the best paper award in the 14th IEEE International Conference on Computational Science and Engineering.