In DiRAC’s latest collaboration with Intel, we presented three half days of virtual introductory training on Intel’s oneAPI Base Toolkit, designed for users of all experience levels. This was followed by a three-day in-person hackathon targeted at experienced researchers aiming to develop their codes for diverse GPU architectures.
The remote training sessions took place from 21st to 23rd May and introduced the oneAPI Base Toolkit. This toolkit is a core set of tools and libraries for developing high-performance, data-centric applications on various GPU platforms. The course covers essential concepts and basic skills such as:
Following the virtual training, a three-day in-person hackathon was held at Codeplay‘s Edinburgh offices, with six research teams participating. The event focused on using Intel’s oneAPI Base Toolkit to refactor CPU-centric codes for GPU architectures and adapt GPU-based codes for new platforms.
Participants were given access to the DAWN Cluster, part of the University of Cambridge’s Open Zettascale Lab, providing a high-performance computing environment for their development and benchmarking work.
Participants were given access to the DAWN Cluster, part of the University of Cambridge’s Open Zettascale Lab, providing a high-performance computing environment for their development and benchmarking work.
Each team was supported throughout the event by four expert software engineers from Codeplay and two DiRAC Research Software Engineers (RSEs).
One-to one and onboarding sessions were held which not only addressed access and connectivity to the DAWN Cluster but also introduced the Codeplay support team to each project’s codebase. This early engagement helped assess the complexity and scope of the work and set realistic goals for the three-day event.
A highlight of the event was the networking dinner, which fostered collaboration and conversation between researchers and industry experts.
The event culminated in an afternoon of project presentations from the research teams demonstrating the wide-ranging scope of applications for GPU technologies.
