Fix checkpointing for FEniCS FSI tutorials

[BenjaminRodenberg]

Only u_n is considered by the checkpointing. But we should also properly checkpoint v_n and a_n. There is already a prototype implementation on this branch (don't merge this branch, it has a different purpose!)

The most important part is

tutorials/FSI/cylinderFlap/OpenFOAM-FEniCS/Solid/cyl-flap.py

Lines 238 to 243 in b3b4985

	solve(A, u_np1.vector(), b_forces)
	a_np1 = project(update_acceleration(u_np1, u_n, v_n, a_n, ufl=False), V)
	v_np1 = project(update_velocity(a_np1, u_n, v_n, a_n, ufl=False), V)
	t, n, precice_timestep_complete, precice_dt, Forces_x, Forces_y = precice.advance(u_np1, (u_np1, v_np1, a_np1), (u_n, v_n, a_n), t, float(dt), n)

[BenjaminRodenberg]

Be aware that this bug does not matter, if we do not use subcycling!

[BenjaminRodenberg]

We should first merge precice/fenics-adapter#34, since this makes handling the solver state much easier.

[BenjaminRodenberg]

This issue is still up-to-date.

tutorials/perpendicular-flap/solid-fenics/perp-flap.py

Lines 185 to 186 in 0c3407a

	if precice.is_action_required(precice.action_write_iteration_checkpoint()): # write checkpoint
	precice.store_checkpoint(u_n, t, n)

should be transferred to something like

 if precice.is_action_required(precice.action_write_iteration_checkpoint()):  # write checkpoint 
    state = (u_n,v_n,a_n) 
    precice.store_checkpoint(state, t, n) 

Same for precice.retreive_checkpoint:

tutorials/perpendicular-flap/solid-fenics/perp-flap.py

Lines 215 to 216 in 0c3407a

	if precice.is_action_required(precice.action_read_iteration_checkpoint()): # roll back to checkpoint
	u_cp, t_cp, n_cp = precice.retrieve_checkpoint()

The commit from the branch mentioned above is heavily outdated, but might still be helpful: b3b4985#diff-e897be07ec33cea9a79328933528b3996749c98b04f78d3dfe69515b489a8780R240-R243

[precice-bot]

This issue has been mentioned on preCICE Forum on Discourse. There might be relevant details there:

https://precice.discourse.group/t/fenics-precice-implicit-coupling-with-multiple-fields-in-checkpoint/1806/1

[BenjaminRodenberg]

I performed some experiments with the perpendicular flap case and subcycling for my thesis. This generally works for the current state of the tutorials and the FEniCS adapter. There is one PR that we still need to close precice/fenics-adapter#172. I think we can then also close this issue.

[BenjaminRodenberg]

Since this issue has been lying around for quite a while, here is a quick update:

• I did several experiments with subcycling and checkpointing for my thesis. From my perspective the state provided via Revert #554 and require fenicsadapter 2.2.0 to fix subcycling #578 works.
• Important: We have to use a recent version of the FEniCS adapter. I'm currently preparing a release in Release v2.2.0 fenics-adapter#184. The main change that is relevant in the context of checkpointing is Use copy(deepcopy=true) for checkpointing fenics-adapter#172.

We can close this issue as soon as these PRs have been merged.