[OUTDATED][WIP]: Make Taal and Taam produce identical (to floating point accuracy) results

[sloede]

Closes #252.

Elixirs to check for results to match their TOML counterpart:

• examples/1d/elixir_advection_amr.jl
• examples/1d/elixir_advection_amr_nonperiodic.jl
• examples/1d/elixir_advection_basic.jl
• examples/1d/elixir_euler_blast_wave_shockcapturing.jl
• examples/1d/elixir_euler_ec.jl
• examples/1d/elixir_euler_nonperiodic.jl
• examples/1d/elixir_euler_source_terms.jl
• examples/2d/elixir_advection_amr.jl
• examples/2d/elixir_advection_basic.jl
• examples/2d/elixir_advection_mortar.jl
• examples/2d/elixir_euler_blast_wave_shockcapturing.jl
• examples/2d/elixir_euler_blast_wave_shockcapturing_amr.jl
• examples/2d/elixir_euler_blob_mortar_shockcapturing.jl
• examples/2d/elixir_euler_blob_shockcapturing_amr.jl
• examples/2d/elixir_euler_density_wave.jl
• examples/2d/elixir_euler_ec.jl
• examples/2d/elixir_euler_khi_shockcapturing.jl
• examples/2d/elixir_euler_khi_shockcapturing_amr.jl not yet tested
• examples/2d/elixir_euler_nonperiodic.jl
• examples/2d/elixir_euler_sedov_blast_wave_shockcapturing_amr.jl More adaptations for Taal/Taam #268
• examples/2d/elixir_euler_source_terms.jl
• examples/2d/elixir_euler_vortex.jl
• examples/2d/elixir_euler_vortex_mortar.jl
• examples/2d/elixir_euler_vortex_mortar_shockcapturing.jl
• examples/2d/elixir_euler_vortex_mortar_split.jl
• examples/2d/elixir_euler_vortex_shockcapturing.jl
• examples/2d/elixir_euler_weak_blast_wave_shockcapturing.jl
• examples/2d/elixir_eulergravity_jeans_instability.jl
• examples/2d/elixir_eulergravity_sedov_blast_wave_shockcapturing_amr.jl
• examples/2d/elixir_hyp_diff_harmonic_nonperiodic.jl
• examples/2d/elixir_hyp_diff_llf.jl
• examples/2d/elixir_hyp_diff_nonperiodic.jl
• examples/2d/elixir_hyp_diff_upwind.jl
• examples/2d/elixir_mhd_alfven_wave.jl
• examples/2d/elixir_mhd_alfven_wave_mortar.jl
• examples/2d/elixir_mhd_blast_wave_shockcapturing_amr.jl cannot be tested -> crashes
• examples/2d/elixir_mhd_ec.jl
• examples/2d/elixir_mhd_orszag_tang_shockcapturing_amr.jl
• examples/2d/elixir_mhd_rotor_shockcapturing_amr.jl up to t_end=0.02 the differences to Taam are FP zero, while for 0.04 they are already O(1e-8)

Relevant only after merging dev:

• examples/2d/elixir_euler_vortex_amr.jl
• examples/3d/elixir_advection_amr.jl
• examples/3d/elixir_advection_basic.jl
• examples/3d/elixir_euler_source_terms.jl
• examples/3d/elixir_hyp_diff_llf.jl
• examples/3d/elixir_mhd_ec.jl
• Remaining 3D examples (tracked in Taal: Porting parameters_*.toml to elixir_*.jl #219)
• Remaining examples ported in Convert further TOML examples to elixirs #245

[ranocha]

The CFL computation in Taam was still different from Taal; I've adapted it for Euler 2D in b7ae3a1. Now, 2d/elixir_euler_ec.jl and parameters_euler_ec.toml both result in the same number of time steps when running till t_end = 1.0. I'll leave the remaining adaptations to you, @sloede

[andrewwinters5000]

Locally I merged the shock capturing with nonconservative terms with this branch. The Orszag-Tang vortex is not reproducible there with Taam vs Taal

[sloede]

Locally I merged the shock capturing with nonconservative terms with this branch. The Orszag-Tang vortex is not reproducible there with Taam vs Taal

Hm, that's weird. But - wait, did you implement everything in Taal and Taam exactly identical in your branch? Cause if not, than that's bound to happen. If you did - well, then it is apparently not identical somewhere....

[sloede]

My plan for this branch is as follows: Finish the examples that are currently present here. Then update the tests such that all tests pass again. That way I know what breaks when merging in new stuff from dev. It's more work upfront, but I hope that this way no bugs introduced upstream (not that I expect you to add anything, just in case) do not slip through.

[andrewwinters5000]

Hm, that's weird. But - wait, did you implement everything in Taal and Taam exactly identical in your branch? Cause if not, than that's bound to happen. If you did - well, then it is apparently not identical somewhere....

I am not sure. But I did check that if I deactivate AMR then the orszag-tang with the new shock capturing matches. So I think it might be that I set up the AMR resolutions differently or something

[sloede]

I think this can be closed. What do you think, @ranocha?

[ranocha]

Yes (if you made sure that we've copied every significant and important change).

[sloede]

Alas, you're right. When closing this PR we should also delete the corresponding branch, but that's still useful for e.g., the fixes to the MHD fluxes. I'll leave this open then until we've merged #267.

[ranocha]

Can we close this now, @sloede?

[sloede]

Yes. Done.

[sloede]

Is it OK to also delete the corresponding branch? I think yes, but it's irreversible, so feel free to do it if you also think it's ok.

[ranocha]

I think so, too. To be safe, I pushed the branch also to my fork.