NESO-fame proof of principle 2D simulation

[jwscook]

Intention is to use the anisotropic diffusion miniapp to test the FAME gridding vs cranking up the p refinement in terms of both accuracy and performance.

Might want to run this past John, Romain, a,nd,/o,r, Andy.

[jwscook]

Relates to https://github.com/ExCALIBUR-NEPTUNE/NESO/issues/176

[jwscook]

Estimate to have done an initial run done by monday (analysis thereafter in python and calculate effective perpendicular diffusivity (the quantity from the Giorgiani paper)).

[jwscook]

This is waiting on a meeting with Dave on Monday

[jwscook]

Dave suggested that:

• he will get the nektar-diffusion miniapp working with DG
• boundary conditions for DG in general should be easy to implement ... but we don't have knowledge of his timeline.

[jwscook]

Docker images are being pulled from Chris' personal dockerhub. We will need dockerhub / ghcr https://github.blog/2020-09-01-introducing-github-container-registry/ in the long run.

[jwscook]

Needs Neumann BCs for DG in the core - work should be done for 3D HW equations by DM. Dave has started working on this.

DM needs to put the DG capable anisotropic diffusion into the Nektar++ release / branch.

[jwscook]

DM ran some code but this gives @cmacmackin NaNs. Perhaps we don't have the right nektar++ commit to build against? Extras:

• a forcing term for MMS
• an implicit time-stepper and/or a steady state solver

[jwscook]

DM has given some code that runs albeit with a low timestep for an explicit stepper.

We still need a forcing term for MMS.

[jwscook]

DoOdeRhs could have a ForcingTerm from SolverUtils. There's an example in SimpleSOL: `NESO/examples/SimpleSol/

[jwscook]

The non-conformal mesh exhibits instability at the top and bottom edges (i.e. away from the source BCs) after a while, where the edge closer to the signal goes first. Try with Dirichlet rather than Neumann.

[jwscook]

Friday 27th Sept, DM was made aware of some IC / BC problems...

[oparry-ukaea]

Some progress with Dave Moxey in collaboration kickoff meeting; Dave will get in touch with @cmacmackin .

[jwscook]

On going with DM

[jwscook]

@ethrelfall showed that NUMMODES=2 is very different to NUMMODES=4 suggesting that the integrals of numerical flux along the nonconformal faces could be responsible.

[jwscook]

@JamesEdgeley to help @cmacmackin

[jwscook]

Working on other things until next week.

[jwscook]

Will need to use NUMMODES>=3 - errors occur with low numbers.