Introductory: 60% Intermediate: 25% Advanced: 15%

SC Conference - Activity Details

M06: Productive Performance Engineering of Petascale Applications with POINT

Sameer Shende  (University of Oregon)
Allen Malony  (University of Oregon)
Shirley Moore  (University of Tennessee, Knoxville)
Nick Nystrom  (Pittsburgh Supercomputing Center)
Rick Kufrin  (University of Illinois at Urbana-Champaign)
Tutorials Session
Monday,  08:30AM - 05:00PM
Room TBD
This tutorial presents the state-of-the-art tool developments for leading-edge HPC, parallel and cluster systems, focusing on scalable profile- and trace-based program measurement, automatic performance analysis, performance databases, and open generic interfaces for tool integration. To evaluate the performance of parallel, scientific applications, we present four open source parallel performance evaluation tools - PerfSuite, PAPI, TAU, Scalasca/Kojak - which are being integrated in the POINT (Performance Productivity through Open Integrated Tools) project. POINT is being supported through the NSF Software Development for Cyberinfrastructure (SCDI) program. Program execution and performance questions of what, where, when, and why form the center of the tutorial, guiding our presentation of tools' characteristics and features. Through methodical treatment of performance engineering including data collection, analysis, and optimization in the context of production applications, this tutorial will prepare participants to apply modern tools for diagnosing and locating typical performance bottlenecks in real-world parallel programs at scale.
   IEEE Computer Society  /  ACM     2 0   Y E A R S   -   U N L E A S H I N G   T H E   P O W E R   O F   H P C