Abstract of SFINCS Project

In view of the forthcoming start of Wendelstein 7-X, collisional transport calculations are timely and important for the stellarator community. In non-axisymmetric devices the neoclassical transport can often be of the same order as the turbulent transport, particularly in the reactor-relevant low collisionality regime. The SFINCS code is a novel drift-kinetic solver which can be used to predict neoclassical flows in 3D magnetic configurations. The code is written in FORTRAN and built around the PETSc libraries which employ MPI as well as BLAS and LAPACK. At the moment, the time and memory requirements for running SFINCS with experimental-like input are significant, and opportunities exist for making the code more efficient and robust. The most demanding part of the code is related to solving large sparse linear systems and there should be room for improvements in this part, e.g. by finding another solver or by modifying the preconditioning.