POP/VI-HPS Tuning

Important note: The workshop cannot take place on the specified date (6–10 July 2026). A new date will be announced as soon as possible. Please register for the workshop here anyway to reserve your place. We will let you know once the new date is set.

On completion participants will be familiar with common performance analysis and diagnosis techniques and how they can be applied in practice on JUPITER (and similar HPC systems). Particular focus will be on MPI, OpenMP/OpenACC/CUDA and hybrid combinations of these parallelisation paradigms for applications running on multiple GH200-based compute nodes. Those who prepared their application test cases will have been coached in their execution performance measurement and analysis, and provided optimisation and tuning suggestions.

Prerequisites:

Participants should have experience with JUPITER-Booster (GH200) or similar GPU-accelerated HPC systems (at JSC or elsewhere) and have an HPC application code that runs satisfactorily on such systems. They should ideally be part of a team familiar with and working on the same code: at least one member needs to be in loco throughout the workshop, whereas others can participate remotely. In loco participants should bring their own notebook computer and be able to connect it to the JSC WLAN to access JUPITER. Jupyter-JSC or SSH with X11-forwarding are both acceptable access options.

A personal institutional email address (university/research institution, government agency, organisation, or company) is required to register for JSC training courses. If you don't have an institutional email address, please get in touch with the contact person for this course.