Developing HPC applications with AMD GPUs

EuroCC2 is collaborating with AMD to organize a workshop from 2 May to 5 May 2023. The workshop will concentrate on creating HPC applications for AMD GPUs. It will begin by offering a developer's viewpoint on AMD system GPUs and the related hardware. It will also cover various programming options such as HIP and OpenMP. The workshop will include presentations on optimizing application performance and an introduction to AMD profiling and debugging tools. No prior GPU programming experience is necessary.

Material from the workshop

2 May 2023

• AMD presentation roadmap, George Markomanolis
• Introduction to LUMI for exercises, George Markomanolis
• Introduction to the AMD CDNATM 2 architecture, George Markomanolis
• Introduction to HIP programming, George Markomanolis
• Developing Fortran applications: HIPFort, OpenMP, and OpenACC, Bob Robey
• Porting applications to HIP, Bob Robey

3 May 2023

• Getting started with OpenMP offload applications on AMD accelerators, Jose Noudohouenou

• Developing OpenACC applications, Bob Robey

• Directive based GPU programming on AMD systems, Justin Chang

• Kokkos and performance portable languages, Bob Robey

4 May 2023

• AMD node memory model, Bob Robey

• GPU-aware MPI with ROCm, Mahdieh Ghazimirsaeed

• AFFINITY – placement, order and binding, Gina Sitaraman

5 May 2023

• Introduction to ROC-Profiler, Bob Robey

• Introduction to OmniTrace, George Markomanolis

• Introduction to Omniperf, George Markomanolis

• AMD debugger: ROCgdb, Bob Robey