Title: System Evaluation and Application Analysis: Examples of Empirical Computer Science
This text will be replaced
Abstract
This talk discusses the nature of empirical studies in recent performance evaluation
and optimization activities. I begin with an overview of performance evaluation and
characterization studies of the Cray XT4 and IBM BG/P systems at Oak Ridge National Laboratory,
focusing in particular on how these results provide guidance in application optimization.
The second part of the talk describes the performance analysis and optimization of codes in
climate and fusion research, emphasizing the identification and removal of limitations in
performance scalability.
Pat Worley is a senior research computer scientist in the Computer Science and
Mathematics Division of Oak Ridge National Laboratory (ORNL).
Pat has a PhD in computer science from Stanford University, where he studied numerical analysis
and parallel algorithms for scientific computing. His recent work has been in parallel
algorithm design and implementation (especially as applied to models used in climate and fusion
research) and in the performance evaluation of parallel applications and computer systems.
Pat is a co-chair of the Software Engineering Working Group for the Community Climate
System Model and is the principal investigator for the Performance Engineering and Evaluation
Consortium End Station, a Department of Energy (DOE) INCITE project that provides the
performance evaluation community with access to DOE leadership class computing systems.
Pat also serves as a performance researcher and engineer for climate and fusion
computational science projects, and leads the application engagement activities of the (DOE)
SciDAC project in performance engineering. Recent system evaluation activities have focused on
the Cray X1E and XT systems and the IBM BG/P.
|
Co-sponsored by The Cyber Center, The Rosen Center for Advanced Computing, The Advanced
Computer Systems Laboratory and The Computational Science and Engineering Programs.
|
