Thursday, 5 November 2026, 15:00 CET
Modern HPC systems gain their computational power through accelerators and ever-increasing parallelism. However, making efficient use of these parallel and heterogeneous resources is challenging. Before targeted optimisations can be implemented, it is essential to understand how the application behaves on the machine. As in any other scientific field, achieving this understanding relies heavily on measurement, post-processing and visualisation tools.
To ensure that our POP tools stay “sharp” and efficient, we ran them at large scale on up to 112,000 CPU cores and 2,048 GPUs on MareNostrum 5 using two HPC applications that maintain a reasonable granularity at this scale.
In this webinar, we will explore the overheads and costs of measuring applications at scale during data collection, post-processing and visualization. This will enable you to select the most suitable tool for your needs.
About the Presenter
Valentin Seitz is a research engineer at the Barcelona Supercomputing Center and a PhD candidate at the Universitat Politècnica de Catalunya. His research focuses on parallel programming models, performance analysis, dynamic load balancing and performance tools.