Description
The refurbished ASTRA8 suite of codes [1] is a powerful framework for interpretive and
predictive modelling of transport in magnetic confinement plasmas.
Thanks to its flexibility and to the new json/NetCDF output file format, ASTRA8 can now
run with IMAS input, and also convert its output files to IMAS, currently the core profiles
and equilibrium IDSs. However the aim is to make ASTRA8 a workflow actor in the strict
sense, e.g. in IMAS python workflows. For this purpose, a restart option has been im-
plemented, which allows to resume an interrupted ASTRA8 run using the last-written
output json file as initial condition for the new simulation. In this way, ASTRA8 can
be run for single time steps and be included in external workflows which orchestrate the
time-advancing themselves. This is useful also for stand-alone ASTRA8 simulations in
computing environment with a constraining wall-time limit.
A new parallelisation scheme, based on Inter Process Communication (IPC), has been
recently implemented in ASTRA. Thanks to its clean separation of gradients calculation,
the new IPC interface allows to parallelise the computation of medium-fidelity transport
models down to the single radial point, for a significant gain in computation speed - a
factor 5. This makes ASTRA8 particularly suited for theory-based transport modelling.
The second ingredient for a successful approach to medium-fidelity transport modelling
is the so-called Pereverzev-Corrigan algorithm which stabilises the convergence of stiff
transport models [2]. In this paper we assess the levels of additional artificial diffusivities
of such algorithm. We discuss the minimum values required for a stable simulation with
stiff models like TGLF and QuaLiKiZ.
[1] G. Tardini et al., submitted to PPCF
[2] G. Pereverzev, G. Corrigan, Comp. Phys. Comm 179 (2008) 579