This event is not part of the ARCHER2 Training programme but is likely to be of interest to ARCHER2 users.
Cirrus is the high performance computing and data science service hosted and run by EPCC at The University of Edinburgh. It now has a GPU partition with 38 x 4-GPU nodes providing a fantastic resource for accelerating research. The goal of the two sessions is to give an overview of some of the ways and best practices to take advantage of GPUs via both traditional HPC simulation and new approaches to research using artificial intelligence and data science. No previous GPU knowledge is assumed, target audience is research software engineers and researchers interested in using GPUs.
Each session is self-contained, i.e. attendance at session 1 not necessary for session 2.
Session 1: Making efficient use of GPUs for High Performance Computing
This session will provide an overview of the GPU hardware available on the Cirrus system, and highlight some tools and techniques to make efficient use of those GPUs. Topics will include a brief introduction to GPU computing, hardware features of the Cirrus nodes, approaches to programming GPUs, tools to support developers and resources for further information.
Session 2: AI & Data Science – tools and frameworks
This session will highlight the GPU accelerated frameworks available for a broad range of applications in deep learning, data science and deployment of AI-based applications. This session will include several examples of what can be achieved using the many tools and frameworks available to researchers.