In July 2024, we organized another technical training in Hamburg in collaboration with DKRZ.
Presentations and results
Our recent technical training provided an in-depth exploration of GPU and OpenACC technologies. The focus was on the implementation and benefits of heterogeneous setups, as well as NVIDIA's profiling tools and CUDA. Participants engaged in hands-on sessions, where they started working with GPUs and ran their own kernels.
The training began with an overview of GPU architecture and the OpenACC programming standard, highlighting how they can be leveraged to accelerate computations and optimize performance. Attendees also learned about NVIDIA’s profiling tools, essential for performance tuning and debugging.
- GPU and OpenACC Intro with Examples by Claudia Frauen and Dominik Zobel (DKRZ)
- GPU on LEVANTE by Jan Frederik Engels (DKRZ) and Harald Braun (ATOS)
- Hands-on I: Getting started on GPUs - Exercises
- Hands-on II: Running first kernels on GPU - Exercises
- Specialities of Nvidia: Debugging and profiling, CUDA by Dmitry Alexeev (NVIDIA)
- Hands-on III: Debugging - Exercises
The training was highly beneficial, equipping participants with the skills to leverage GPU and OpenACC in their projects. The hands-on approach allowed attendees to directly apply their newfound knowledge, fostering a deeper understanding of the material. Overall, the session marked a significant step forward in our ongoing efforts to harness the power of heterogeneous computing.
The final agenda can be found here.
