Add task for running simulations with `H3LAPD` solver

UCL / FabNESO

Neptune Exploratory Software plug-in for FabSim3

http://github-pages.ucl.ac.uk/FabNESO/

BSD 3-Clause "New" or "Revised" License

0 stars 1 forks source link

Add task for running simulations with `H3LAPD` solver #2

Closed matt-graham closed 11 months ago

matt-graham commented 1 year ago

From discussions with NESO team, seems like H3LAPD should be a more computationally challenging example with both relatively complex continuum / Nektar++ and particle-in-cell components.

[ ] Get H3LAPD solver running locally with example configurations.
[ ] Figure out which parameters from conditions file to expose (both 'model' parameters affecting simulation behaviour and 'control' variables affecting simulation performance / accuracy)

dleggat commented 12 months ago

The abstraction in the FabNeso.py plugin appears to work for this, but I haven't been able to get H3LAPD to actually run. I keep getting segfaults with no associated error codes, so I don't know what the problem is. Have you managed to successfully run H3LAPD @matt-graham ?

matt-graham commented 12 months ago

Just tried running H3LAPD solver with a couple of example configurations in examples/H3LAPD directory with and without mpirun and also seem to be getting segfaults in all cases 😞.

matt-graham commented 12 months ago

@dleggat - just in case you also missed this, I realised that I had not spotted the README at

https://github.com/ExCALIBUR-NEPTUNE/NESO/tree/main/examples/H3LAPD

which indicates we need to use NekMesh to convert the .geo geometry files in the examples directory to XML files that can be used with Nektar++. After installing gmsh and following the instructions to convert the geometry file using the provided script I've been able to run the examples/H3LAPD/2Din3D-hw_fluid-only example with my locally built NESO without any segfaults.

matt-graham commented 12 months ago

Owen Parry indicated in an email that reasonable choice of initial parameters to expose for doing sweeps over / calibrating would be

HW_alpha over range 0.1 – 2.0,
HW_kappa over range 0.1 – 3.5,
Te_eV over range 2.0 – 20.0.

He also indicated some configurations in this parameter space may cause solver to fail so we may need some sort of way of handling this.

As quantities of interest, the energy and enstrophy are written to a CSV file as described here.

dleggat commented 12 months ago

Ah hah right - I missed the README and went straight to the run_cmd_template.txt - woops!

matt-graham commented 11 months ago

Closing this as completed by #7