Introduction to LUMI-G hardware and programming environment for LUMI-G pilot projects

This one day on-line course serves as a detailed introduction to the LUMI-G architecture and programming environment. It is intended exclusively for LUMI-G pilot projects to enable them make good use of LUMI-G during the upcoming pilot phase (similar courses will be held for other projects/users in the future). For doing the hands-on exercises on the course day, participants need an active account on LUMI (username and project with remaining compute/GPU hours and storage hours). The course will be jointly conducted by the LUMI HPE Centre of Excellence (HPE CoE), AMD and the LUMI User Support Team (LUST).

All registrations from LUMI-G pilot projects will be admitted to attend the course. We will send information about connection details shortly before the course (latest on August 22nd 2022).

Attendance is free of charge.

Tentative schedule of the course.

• 9:00-9:10: Introduction
• 9:10-10:15: HPE topics
  • Hardware outline
  • Quick recap of Cray Programming Environment
  • Slurm usage
  • binding (cpu/GPU/NIC) considerations
  • MPI update for GPUs/SlingShot11 (G2G)
  • How to use available compilers (CCE/ROCm) for OpenMP/OpenACC/HIP etc.
  • Tools (debugging and profiling)
• 10:15-10:45: Break
• 10:45-12:00: Continuation of HPE topics
• 12:00-14:00: Lunch break
• 14:00-15:45: AMD topics
  • GPU Hardware intro
  • Introduction to ROCm and HIP
  • Porting Applications to HIP
  • ROCm libraries
  • Profiling
  • Debugging
• 15:45-16:15: Break
• 16:15-17:00: Continuation of AMD topics + examples & hands-on exercises
• 17:00-17:30: Questions

Note, the LUMI User Support Team cannot create any accounts. If you belong to a LUMI-G pilot project, please contact its PI.