HPC Optimization and Scaling Workshop
(Course no. 1292019 in the training programme 2019 of Forschungszentrum Jülich)
Currently installed systems at JSC and the ones planned for the future are characterised by ever increasing levels of complexity. This makes it necesary to master different programming paradigms and tools for analysis. At the same time, the size of computer systems and problem sizes tackled by scientists increases. This increased potential and demand in net compute power asks for better scalability of user codes. Consequently, this workshop comes in two parts: Optimisation and Scaling. The first focuses on analysis tools and node-level performance, the latter looks into large scale executions.
This course will give users of JSC's supercomputing systems hands-on practical training with a variety of tools available to assist with performance analysis, optimisation and scaling of HPC applications on Intel architectures, to help make efficient use of allocated resources on JUWELS (Skylake) and JURECA (Haswell & KnightsLanding). Participants are encouraged to prepare representative small/short-running application test-cases convenient for experimentation in the hands-on sessions with expert coaching.
Note that assistance for porting, building and running applications will not be available during the workshop, but can be requested in advance.
Topics covered include:
- Structured parallel performance engineering
- Execution characterisation and MPI correctness validation
- Single-core performance analysis including vectorisation
- Multi-threading and node-level/roof-line analysis
- Multi-node performance/scalability analysis
For a small group of pre-qualified users, the experts will also help to investigate the scalability of their codes, aiming to run on the full supercomputer systems, each with 1000+ nodes.
Participants should belong to a currently active compute project on our systems (project-ID!) and should already have an active user account (user-id!). Test runs during the workshop will be done on the participants' own project accounts.
In order to gain maximum benefit from this workshop, participants should have an in-depth (hands-on) knowledge of the algorithms and codes used in their project.
In addition, we also ask participants to prepare a test case to evaluate the initial performance of their code and to monitor the progress.
The optimisation part of the workshop is open to all users that are asked to register via the provided link. Participation in the scaling part of the workshop requires registration and prior acceptance by the organisers. Note that in the application form we ask some information about your project and code in order to better plan the workshop program.
Due to available space, the maximal number of participants is limited.
For any questions concerning the workshop please send an e-mail to
email@example.com (Dr. Dirk Brömmel)
firstname.lastname@example.org (Dr. Brian Wylie)
Jülich Supercomputing Centre (JSC)
Institute for Advanced Simulation (IAS)
Forschungszentrum Jülich GmbH
The presentations and hands-on sessions of the workshop will take place in the Rotunda (building 16.4, room 301).
Agenda for the HPC Optimisation and Scaling Workshop (PDF, 54 kB)
Scaling Workshop Guide
The scaling part of the workshop requires a somewhat stricter engagement of particiapants. This structured approach aims for increasing the positive outcome of the workshop and is detailed in the document linked below. Please familiarise yourself with the document ahead of the workshop. Note that details may be changed as the workshop is finalised, so please check again right before the date of the workshop.
Scaling Workshop Guide (PDF, 219 kB)