Hybrid workshop for vital nuclear fusion supercomputer code
UKAEA personnel and external grantees gathered at Culham and via Zoom on 7th October 2021 to review progress in and consolidate future plans for ExCALIBUR project NEPTUNE (NEutral and Plasma TUrbulence Numerics for the Exascale).
As part of the wider ExCALIBUR portfolio, the project aims to produce a next-generation computer code/library capable of describing the physics of the so-called Scrape Off Layer or ‘edge’ region of the tokamak plasma where superheated particles come into contact with the reactor vessel first wall. The project will deliver capability in the form of actionable tools for the design and operation of a future fusion power plant capable of delivering clean and reliable electrical power to the grid.
The plasma edge region is subject to extreme variation in temperature – it connects the hotter-than-the-Sun, rarefied plasma core with the solid walls of the confinement vessel. The underlying physics ranges from highly-energetic particles produced in the core region to turbulent fluid phenomena further out, bounded by magnetic ‘sheaths’ at the walls themselves. UKAEA’s Tokamak Science Programme, leading members of the Oxford University Plasma Theory group and the York Plasma Institute are all contributing world-class physics expertise to the project.
A morning session surveyed work to date, including the production of the first executable codes (or proxyapps): a solver for highly-anisotropic diffusion problems, a solver for the drift-kinetic plasma equations (both using state-of-the-art spectral / hp finite element methods), and a particle simulation code using a novel implicit time-advancing method. These early outputs are publicly available from the project’s github repository (https://github.com/ExCALIBUR-NEPTUNE/). Access to reports produced by the project is available on request to firstname.lastname@example.org.
The range of designs proposed for exascale supercomputers is more diverse than ever before; such systems demand new programming techniques in order to extract the maximum performance (indeed, today’s codes, if not carefully modified, may actually run slower on the new machines than they do on current hardware). The workshop therefore included UK experts in ‘performance portability’ as well as developers of leading high-performance, scalable particle and fluid simulation codes, augmented by in-house activities by members of the UKAEA’s Advanced Computing Department.
Afternoon sessions focussed on software engineering aspects of the project and on relevant new developments in particle and fluid simulations (including their integration), with lively discussion resulting in a clear picture of forthcoming directions.
The workshop took place in hybrid format with 25 participants making good use of new digital conferencing technology in the CCFE Pease Room installed over the COVID-19 lockdown.