This three-half-day program is aimed at HPC scientists, researchers, and practitioners who want to port, optimize, and profile their codes on AMD systems.
Program highlights:
- Day 1 (Beginner): HIP and ROCm introduction, OpenMP offloading, profiling basics.
- Days 2–3 (Advanced): advanced HIP/OpenMP, GPU-aware MPI, performance portability frameworks, AMD profiling tools, latest utilities (trace decoder, advanced profilers).
- Each module includes hands-on exercises.
Audience: researchers and engineers from academia or industry with prior HPC experience.
Prerequisites: Linux command-line, HPC software stack familiarity, compiling/running codes on HPC systems.
Date: October 28–30, 2025 (from 1pm to 5pm each day CET)
Hybrid format:
- Online: remote participants will access AMD infrastructures via the AMD Accelerator Cloud.
- On-site at CINES (Montpellier, France): limited seats, with the unique opportunity to run exercises directly on the ADASTRA supercomputer. Accommodation recommendations are provided below. On-site registration deadline on September 28, 2025. You will be asked to bring your ID/Passport.
This training will equip participants with the skills to efficiently exploit AMD GPUs and CPUs for high-performance workloads.