Unlock the full power of accelerated supercomputers and master the techniques required to scale applications across multiple GPUs and compute nodes.
This training is organized within the EuroCC and EuroHPC initiatives in collaboration with NVIDIA and leading European supercomputing centers.
Experts from ACK Cyfronet AGH will support the event as Teaching Assistants, assisting participants during hands-on exercises and sharing their expertise in parallel programming and the use of GPU accelerators in scientific and industrial computing.
When? September 29-30, 2026
Format: Online (Zoom)
Level: Intermediate
The training focuses on scaling applications in multi-node and multi-GPU environments. Participants will learn advanced programming and optimization techniques that enable the efficient use of multiple GPUs simultaneously.
Topics covered during the bootcamp include:
- Multi-GPU programming,
- Inter-process communication using MPI,
- Profiling and identification of performance bottlenecks,
- Analysis of the root causes of performance issues,
- Using NVIDIA libraries for HPC environments,
- Communication topologies and their impact on application scaling.
Participants will be granted access to a cluster equipped with NVIDIA A100 accelerators, where they will be able to test the presented solutions in practice.
Prerequisites:
- Experience in C/C++ programming,
- Familiarity with NVIDIA CUDA,
- Familiarity with MPI.
Registration: https://events.asc.ac.at/event/358/
