Pruning TDycore

[jeff-cohere]

In the interest of focusing on what we need to be relevant to a project like E3SM, I think it would be good for us to start a conversation about what we can get rid of, now that we know a bit more what we're doing.

Here are some starter questions (at varying levels of provocation):

• What algorithms are we no longer including in the TDycore library?
• What demos are no longer relevant?
• What are the combinations of modes/discretizations/solvers that we wish to support explicitly?
• Do we still need 2D capability (e.g. for transect calculations), or can we get everything we need from 3D?

@bishtgautam , if it's appropriate, maybe we can talk about this at our next joint meeting.

[bishtgautam]

Fate of demos

1. SPE10/steady.c
   • No test
2. mpfao/mpfao.c
   • Steady state MPFA-O tests (mostly 2D, one 3D test, one test using an exodus file 64xy_3z_wedge.exo)
   • Delete the demo and move the 64xy_3z_wedge.exo test in richards/richards_driver.c
3. qa_test/qa_test.c
   • For QAQC framework
   • Keep it for now
4. richards/richards_driver.c
   • MPFA-O tests for SNES/TS, serial/parallel, reading permeability datasets
   • DEFINITELY keep this demo
5. steady/steady.c
   • MFE tests: WY, TPF and BDM
6. steadyblock/steadyblock.c
   • No test
   • Delete the demo
7. steadyf90/steadyf90.F90
   • This is a FORTRAN version of steady/steady.c
   • Delete the test
8. th/th_driver.c
   • Serial and parallel tests for TH
   • DEFINITELY keep this demo
9. transient/transient.c
   • WY serial and parallel tests
10. transient/transient_mpfao.c
    • MPFA-O tests for various time integrators: TS, TS+DAE, TS_TRANSIENTVAR
    • Delete the demo and move the tests in richards/richards_driver.c
11. transient/transient_mpfaof90.F90
    • MPFA-O tests using TS time-integrators for different
      1. `discretization methods (MPFA-O/2 point flux)
      2. `different BCs
    • Keep the demo
12. transient/transient_snes_mpfaof90.F90
    • MPFA-O tests using SNES time-integrators for multiple configurations
    • DEFINITELY keep this demo
13. transient/transient_th_mpfao.c
    • Tests transient TH
    • Delete it as TH is tested by th/th_driver.

[jeff-cohere]

Thanks, Gautam!

[nocollier]

This is a helpful directory of what we have. The SPE10 problem is really just setting up a flow via a imposed pressure gradient diagonal to the mesh using the permeability fields from that problem. We used it as a stress test for solvers--it doesn't really solve anything real. On the one hand, it does not really need a test and we could remove. On the other, why relocate it to another place making it harder to find when we need it again? I would leave it, but don't feel strongly about it.

[jeff-cohere]

I think it's fine to keep "historical" things around, as long as we document their purpose. A stress test for solvers sounds like a nice tool/benchmark to keep around.

[bishtgautam]

• Get rid of 2D MPFA-O code. MFE is mesh independent
• MPFA-O: RICHARDS & TH mode for SNES and TS+TransientVar (BDF2 is more relevant for this project)
• MPFA-O: RICHARDS TS and TS+DAE should be kept for now
• Keep SPE10 problem

[jeff-cohere]

I'm going to remove the 2D MPFA-O code in a pull request, and then I'll turn my attention to the demos in general.

[jeff-cohere]

Seems like we can close this issue now. Any objections, @bishtgautam ?

[bishtgautam]

The is still a bunch of 2D MPFA-O code that can be deleted. e.g. https://github.com/TDycores-Project/TDycore/blob/master/src/mpfao/tdycoremesh.c#L1362. If no one else is working on it, I can work next week on removing the remaining MPFA-O 2D code.