Skip to content. | Skip to navigation

EUROfusion

Abstract of PARFS2 Project

In 2012 the Frascati theory and modelling group has applied for the use of HLST resources in order to get a distributed version of the field solver (MARST) of the code HYMAGYC (ParFS project): the project has been approved and has given promising results, although not conclusive. In 2013 this group has applied for the continuation of the project (ParFS2 project), but due to limited human resources of HLST the project has been rejected. Within this HLST project the profound restructuration of the MARST module, i.e. the parallelization of the original computation of the matrices elements of the linear system representing the MHD linear system has been successfully completed, mainly by the Frascati group. The first choice of the parallel solver (MUMPS) and its implementation by the HLST team has also been completed, allowing MARST to face the investigation of relevant mode numbers (n ≈ 40) for the class of Alfvénic modes in ITER like configurations. Nevertheless, the scalability, in terms of memory and time execution, of MUMPS is not yet satisfactory.

In the present application we propose to extend the investigation on other parallel solvers with regard to robustness, scalability and resource usage characteristics, in order to find an alternative to MUMPS.

As a side application of the present activity, we propose to apply a similar parallel solver to the original code from which the field solver of HYMAGYC was derived, i.e., the linear MHD eigensolver MARS, extending it to the study of "free boundary modes".