Abstract of REFMULIO Project

This proposal aims at improving the I/O capability of the code REFMUL3, a 3D parallel FDTD full wave Maxwell code for simulation of reflectometry.

Simulation of reflectometry using a Finite-difference Time-domain (FDTD) code is one the most popular numerical techniques used, as it offers a comprehensive description of the plasma phenomena. This method requires however a fine spatial grid discretization to keep the error to a minimum, which in turn implies a high-resolution time discretization to comply with CFL stability condition. Memory demands become also constringent, especially if with full-size simulations of large devices, like JET, ASDEX Upgrade or next generation machines like ITER are considered. Simulations in 3D are therefore very demanding computationally and are only possible if the problem can be distributed over large numbers of resources. REFMUL3, a 3D full-wave code using a FDTD Yee scheme [1] with full polarization, manages that by having a hybrid MPI/OpenMP parallelisation using explicit 3D domain decomposition. It copes simultaneously with o- and x-modes, supports a general external magnetic field and a dynamic plasma. REFMUL3 is able to address a wider range of scenarios than the present day codes can (which are 2D, either o- or x-mode). The code is being developed at Instituto de Plasmas e Fusão Nuclear (IPFN) and its parallelisation was made during the HLST-REFMUL3 project (2016). The final results of the HLST-REFMUL3 project showed that (i) the parallelization employed yields very good scalability of the code over few thousands of cores; (ii) I/O has become a major bottleneck since the code parallelism allows much larger grids and higher numbers of iterations.

The development of a 3D code represents a highly envisaged effort that is much needed in the EU fusion programme and a major new tool for the reflectometry community.