Atelier T8.A05

Tutoriel parallèlisme hybride Multi GPU avec OpenMPI et OpenACC .

Arnaud Renard et Jean-Matthieu Etancelin Centre de calcul ROMEO et Francois Courteille/Nvidia.

GPU, OpenMP, openACC, Cuda, openCL, librairie, simulations parallèles, parallèlisme, calcul intensif, HPC, GPU.

• In this lab you will learn how to program multi GPU systems or GPU clusters using the Message Passing Interface (MPI) and OpenACC. Basic knowledge of MPI and OpenACC is a prerequisite.

The topics covered by this lab are:

• Exchanging data between different GPUs using CUDA-aware MPI and OpenACC.
• Handle GPU affinity in multi GPU systems.
• Overlapping communication with computation to hide communication times
• Optionally how to use the NVIDIA performance analysis tools

• La connaissance d’un langage de programmation (Fortran, C, …) est souhaitable.
• Connaissance basique de MPI et OpenACC.
  • Les ateliers T8.AP02 et T8.A03 permettent de s'initier à OpenACC.