Article in arXiv:2212.03036, Computational Physics

L. Einkemmer, A. Moriggl: A semi-Lagrangian discontinuous Galerkin method for drift-kinetic simulations on GPUs

By Alexander Moriggl

We have demonstrated the efficiency of a novel numerical scheme to solve the drift-kinetic equation, which
is a simplifed model to simulate the evolution of plasmas with strong magnetic fields, for example, in
tokamak devices. This method (the 2d semi-Lagrangian discontinuous Galerkin method) is used in
this setting to resolve the dynamics in the cross-section of a tokamak. It is able to conserve mass up
to machine precision, which is an important property to get physically relevant long-time simulations.
Additionally, it allows us to take large time steps. That is essential to reduce the overall computation
time. This numerical scheme also provides local data dependency and can thus be implemented efficiently
on modern computer architectures. Our implementation obtains excellent performance using graphic
processing units (GPUs), which outspeed central processing units (CPUs) significantly.

 

 


Nach oben scrollen