Abstract of PARSOLPS Project

The importance of heat exhaust for ITER and future fusion devices has been extensively recognized. One of the more widely used tools for modelling heat exhaust in current and future tokamaks is the SOLPS code, but a significant drawback to its use is the 1 – 12 weeks that it can take for a converged run for ITER or DEMO. Parallelization in current versions of the code is limited: the EIRENE component of SOLPS has been parallelized using MPI, and the B2 component with OpenMP (but only achieving a scale-up of a factor of 2). The proposed work would address the issues of (1) improving the parallelization in B2; (2) ensuring that the OpenMP parallelization of B2 and the MPI parallelization of EIRENE function efficiently together; and (3) explore other methods of speeding up the B2 part of SOLPS, possibly by using better matrix solvers.