Advanced course: LUMI Profiling and Optimization Workshop (On-site)
Start
Slut
Sted
University of Copenhagen Panum and Mærsk Tower, Blegdamsvej 3B, 2200 Copenhagen, Denmark
Arrangør
LUMI User Support Team (LUST) and University of Copenhagen, DeiC
This comprehensive advanced-level 3-day on-site course provides an in-depth exploration of the LUMI supercomputer architecture and programming environment.
Designed for experienced HPC users, the course combines theoretical knowledge with practical hands-on exercises to maximize your efficiency on both CPU and GPU partitions.
Prerequisites:
- Solid knowledge of HPC cluster computing
- Familiarity with Linux command line
- Programming experience in C, Fortran, or Python
- Good understanding of LUMI supercomputer, as obtained from, e.g., the LUMI introductory courses and some active experience
- Active LUMI project or application in progress
What you’ll learn
- LUMI Architecture: Deep dive into system design and components
- Performance Tools: Compiler optimization, performance analysis, and profiling
- Environment variables: Understanding Cray MPI on Slingshot, GPU support and
rank reordering - I/O Optimization: Efficient data handling and storage strategies
- HPE tools for profiling
- AMD GPU Tools: Specialized profiling and optimization for GPU workloads
