This hands-on tutorial introduces you to the parallel performance analysis tools Vampir and Scalasca and how they can be effectively used together.
There will be four presentation parts plus allotted time for individual work with assistance from the presenters.
In all parts we'll start with a well-known parallel benchmark code to introduce all steps with the tools. Afterwards, we'll assist you with the tools on your own application codes on the Yellowstone system. Ideally, your application should be compiling and running correctly with an appropriate input data set on the Yellowstone system prior to the tutorial.
For the first part we'll join with the tutorial "Improving Application Performance Using TAU" since TAU also supports the common run-time measurement system Score-P.
BUFFET LUNCHES and COFFEE BREAKs are provided
Dr. Markus Geimer is a senior scientist at the Jülich Supercomputing Centre of Forschungszentrum Jülich GmbH, Germany. He is the lead developer of Scalasca's parallel trace analysis component and also involved in the design and development of the Open Trace Format 2 (OTF2) and Score-P.
Dr. Andreas Knüpfer is a senior researcher at the Center for Information Services and HPC (ZIH) at TU Dresden, Germany and involved in HPC performance analysis tools for a long time including VampirTrace, the Open Trace Formats 1 and 2 (OTF and OTF2), Score-P, and Vampir.
Markus Geimer and Andreas Knüpfer have a long history of joint tutorials about HPC performance analysis tools.