Abstract of DGGS2 Project

A new explicit algorithm, based on the Discontinuous Galerkin (DG) method, has been developed for gyrokinetic solvers. This algorithm was implemented and tested in the GS2 code under the 2010 IMPGS2 HLST project. GS2 is a time-dependent 5-D gyrokinetic code for the simulation of plasma turbulence, and has been extensively used for fusion work on HPC facilities, including HPC-FF. The new algorithm will help optimise the use of GS2 for demanding gyrokinetic simulations that require large grids and access to HPC.

Under this project we will subject the new DG algorithm to a more exhaustive set of tests than were possible under the IMPGS2 project. This will ensure that the DG algorithm is robust, and will allow its performance to be compared more extensively against GS2's original implicit method. As a second part of this project we will assess the software development required and the computational benefits of parallelising GS2 in this additional dimension.

Successful completion of the tests will demonstrate the robustness of the explicit DG algorithm in GS2, and quantify the algorithm's performance over a range of problems. Producing a road-map of the steps required to parallelise along the magnetic field direction will be extremely valuable, as this additional domain decomposition would allow significant further gains in performance. These tasks will improve the efficiency of GS2's use of HPC, and will enable larger grid gyrokinetic calculations for high fidelity predictions of plasma turbulent behaviour in devices such as JET, MAST and ITER.