Accelerating the Vlasov code kobra with GPUs

• Promoter: Prof. Sven Van Loo
• Supervisor: Prof. Sven Van Loo
• Study programs: Master of Science in Engineering Physics, Master of Science in Physics and Astronomy, Master of Science in Teaching in Science and Technology (Physics and Astronomy), European Master of Science in Nuclear Fusion and Engineering Physics
• Location: Technicum, at home

In order to achieve enough energy output viable for commercial exploitation of fusion reactors, fusion plasmas need to be ridden from global instabilities in the magnetic structure of the plasma. To understand the dynamics and arising instabilities, theoretical modelling needs to be tightly integrated with the experiments. While a fluid approach  has been successfully applied to many problems in fusion devices, it is restricted because it is based on the assumption that collisions between plasma particles cause the velocity distribution function to be Maxwellian. However, in tokamak plasmas, the electrons also often exhibit markedly non-Maxwellian distributions. Thus, in many cases the details of the kinetics cannot be neglected and the plasma needs to be described using the Vlasov equation. Therefore we have developed the Vlasov code (written in C) that can be run using multiple computation nodes or processors. This is necessary for large models to be executed in an acceptable amount of time. In the last few years  GPU accelerators have had a notable impact on high-performance computing across many disciplines. They provide high performance with low cost/power, and therefore have become a primary compute resource on many of the largest supercomputers.

In this project the aim is to port the Vlasov code kobra from CPUs to GPUs. First you will get familiar with the current code. Then you will make modifications in order for kobra to be run on a single GPU. You will run test models to identify the speed-up. As a single GPU's memory is limited, you will then attempt to get the code working on multiple GPUs. You will learn about kinetic plasma theory, numerical simulation techniques and CPU and GPU computing during this project.