In this 90 minute tool talk POP expert Bernd Mohr (JSC) provided an introduction to the performance measurement methods profiling and tracing, and how this is supported by the performance instrumentation and measurement framework Score-P. He also showed how the measured profile data can be analyzed and visualized with the Cube profile browser.
This presentation was part of the HPC.NRW Tool Talk series that presented introductory talks to software development tools for HPC applications. The videos are live recordings of online presentations.
- 0:00 Introduction
- 0:48 Parallel Performance Tools 101 - Intro
- 1:31 Performance Analysis Cycle
- 4:58 Measurement Techniques
- 7:27 Details on Profiling
- 11:26 Details on Tracing
- 20:17 Parallel Performance Tools 101 - Summary
- 22:38 The Score-P Ecosystem
- 37:07 Performance Analysis Workflow
- 39:07 Demo: Measurement of a Jacobi Solver
- 55:59 Profile Analysis with Cube
- 1:02:45 Demo: Profile Analysis of Tealeaf
- 1:20:55 Further information
- 1:21:25 Cube Command-line Tools
- Presentation slides
- Score-P and Scalasca overview
- Example code used in Score-P demo
- Example cube file used in Cube demo
About the Presenter
Bernd Mohr first started to design and develop tools for performance analysis of parallel programs with his diploma thesis at the University of Erlangen in Germany, and continued this in his Ph.D. work (1987 to 1992). During a three year postdoc position at the University of Oregon, he designed and implemented the original TAU performance analysis framework. Since 1996 he has been a senior scientist at Forschungszentrum Jülich. Starting in 2000, he has been the team leader of the group ''Programming Environments and Performance Analysis''. Besides being responsible for user support and training in regard to performance tools at the Jülich Supercomputing Centre (JSC), he is leading the Scalasca performance tools efforts in collaboration with Prof. Felix Wolf of TU Darmstadt. From 2007, he has also served as deputy head for the JSC division ''Application support''.