search

underworldcode / underworld3

https://underworldcode.github.io/underworld3/
Other
21 stars 10 forks source link

Generic solver where F0 and F1 can be set by the user #217

Open bknight1 opened 3 months ago

bknight1 commented 3 months ago

Is your feature request related to a problem? Please describe. Allow F0 and F1 to be set in a generic solver. This would allow users to create their own solvers for various problems.

Describe the solution you'd like Allow F0 and F1 to be set up by the user. Maybe in the generic solver, happy to hear other ideas

Describe alternatives you've considered Currently have to copy the constitutive model, modify the flux term for F1 and then insert terms in the f variable for F0. This works currently but might not in the future. Would be better to have F0 and F1 terms settable in a generic solver.

This would help with the benchmark scripts for the lithostatic pressure solver, the diffusion-only models I'm working on and various other equations people are implementing.

Additional context Happy to implement

julesghub commented 3 months ago

An advanced user feature but worth it says I

bknight1 commented 3 months ago

We already have the SNES_Scalar and SNES_Vector classes in the cython solvers file. Shall the generic solver inherit from these classes and just have F0 and F1 setters? Or is there a better approach?

The SNES_Scalar and SNES_Vector code also looks very similar. Is it worth re-writing a generic class that can handle both?

julesghub commented 3 months ago

Starting work on this now.... will change the setters on the generic class for F0 & F1.

It would be nice to combine SNES_Scalar & SNES_Vector once the setters rewrite is done.

lmoresi commented 2 months ago

@julesghub - are we there yet ? @bknight1 - is he there yet ?

bknight1 commented 2 months ago

Still a WIP, Julian has cleaned up the code in preparation but hasn't written the generic class yet