High-Scalability workshop brings researchers to MareNostrum 5

The Barcelona Supercomputing Center (BSC) hosted research teams and advanced users for the High-Scalability Workshop on MareNostrum 5. The event brought together professionals looking to scale their applications to extreme levels and fully leverage modern high-performance computing (HPC) architectures on one of Europe’s most powerful supercomputers.

The three-day workshop focused on the development, optimisation, and execution of highly scalable codes on MareNostrum 5. Combining technical talks with practical hands-on sessions, participants worked directly on optimising and benchmarking their codes using a substantial fraction of the MareNostrum 5 system. Applications with strong scalability and robustness were executed on the full machine to validate extreme-scale performance.

Topics covered during the workshop included large-scale parallelisation strategies, advanced debugging techniques, performance profiling, GPU optimisation, and best practices for executing jobs across thousands of nodes. Participants also gained access to the accelerated partition, enabling experimentation with GPU-enabled applications, including AI workloads, large-scale simulations and large language model (LLM) training.

According to Federico Pedron from BSC, the event exceeded expectations: “The event was great; with some unexpected BSC attendants, we reached 35–36 people, and all of them were very happy with the experience.”

The High-Scalability Workshop strengthens the importance of hands-on collaboration between HPC experts and research teams, with the mission to help participants prepare their codes for supercomputing.