A practical course presented by PDC in collaboration with ENCCS, HPE and AMD
- 12:50 – 13:00 Soft start and icebreaker question on HackMD
- 13:00 – 13:10 Welcome to the course (Johan Hellsvik)
- 13:10 – 13:45 Why GPUs? What problems fit to GPU? The GPU hardware and software ecosystem (Jonathan Vincent)
- 14:00 – 14:20 GPU programming concepts (Jonathan Vincent)
- 14:20 – 14:45 Introduction to GPU programming models (Andrey Alekseenko)
- 15:00 – 15:20 Introduction to Dardel (Johan Hellsvik)
- 15:20 – 15:45 OpenMP (Johan Hellsvik)
- 16:00 – 16:45 OpenMP exercises (Johan Hellsvik)
- 16:45 - 17:00 Buffer time
- 12:50 – 13:00 Soft start and icebreaker question on HackMD
- 13:00 – 13:20 HIP (Jonathan Vincent)
- 13:20 – 13:45 HIP, incl Calling C from Fortran (Jonathan Vincent)
- 14:00 – 14:20 HIP exercises (Jonathan Vincent)
- 14:20 – 14:45 SYCL (Andrey Alekseenko)
- 15:00 – 15:20 SYCL exercises (Andrey Alekseenko)
- 15:20 – 16:45 AMD OmniTrace and Omniperf (Samuel Antao, AMD)
- 16:45 – 17:00 Information on homework exercises (Johan Hellsvik)
- 09:00 – 11:00 Compiler techniques, Parallel environment and modules, Advanced placement (Jean Pourroy, HPE)
- 11:00 – 12:00 Own work on Himeno and own codes with tutoring
- 13:30 – 15:00 Own work on Himeno and own codes with tutoring
- 15:00 – 16:00 Q&A section: Highlight things seen during the onsite help. Hand out template solutions for Himeno
- 09:00 – 11:00 Optimization tools Introduction to perftools, Advanced performance analysis, Debugging at scale (Pierre-Eric Bernard, HPE).
- 11:00 – 12:00 Continue on example codes: analyze codes
- 13:30 – 15:00 Walk through of the Himeno template solutions.
- 15:00 – 15:45 Q&A section: Highlight things seen during the onsite help
- 15:45 – 16:00 Concluding session.