gpu hackathon 2019 - Training
|Date:||16.10.2019 9:00 - 18.10.2019 17:00|
|Location details:||The event is organised at the CSC Training Facilities located in the premises of CSC at Keilaranta 14, Espoo, Finland. The best way to reach us is by public transportation; more detailed travel tips are available.|
|The fee covers all materials, lunches as well as morning and afternoon coffees.|
CSC is in collaboration with Nvidia and the E-CAM European HPC center of Excellence arranging a 3-day GPU hackathon. The GPU hackathon is a coding event in which teams of developers port their applications or kernels to run on GPUs, or optimize their applications that already run on GPUs. In particular the hackathon focuses on applications that can scale up to multiple GPU nodes.
We are looking for teams of 3-4 developers. Collectively the team should know the application intimately. Please keep in mind that we are looking for teams with plans to develop GPU code – not to just run their code on GPUs. During the hackathon each team is supported by one mentor with in-depth GPU programming expertise.
At CSC the new Puhti-AI partition provides 80 nodes with 4 NVidia Volta GPUs each. This system provides in total more than 2 petaflops of performance. This system is available during the course and accepted teams will also have access to the system beforehand to do some initial porting of the applications to Puhti.
This event helps the teams to jumpstart acceleration or optimization of their code on GPUs. By the end of the event each team should have their code running on GPUs, or at least have a clear roadmap of how to get there.
Basic level of GPU programming proficiency is enough. If the team members do not know GPU programming, they should attend the OpenACC course that is arranged at CSC during Monday and Tuesday the same week (https://www.csc.fi/web/training/-/gpu-openacc-2019). The participants should also be familiar with HPC programming in general, and in using clusters and supercomputers.
There are a variety of tools available to program GPUs (e.g. CUDA, OpenACC, OpenMP4.5+, etc.) and we encourage you to explore the best one for your needs.