Karapiperis/updates on time integration business

We might need to think about the time step we use for psi in cosmo runs.

added MHD label

requested review from @fastasyszyn

assigned to @matthieu

mentioned in merge request !1896 (closed)

approved this merge request

Looks good to me.

Interested to know whether we need the same for the VP scheme.

We should ideally use

$\delta t_{therm}$

- rather than

$\delta t_{\alpha}$

- in the time integration of

$\psi$

, right ? This would require a couple of extra modifications to the code.

I think they are the same.

It is mentioned in one of the code comments that

$\delta t_{\alpha}$

is non-cosmological ?

delta t drift and delta t therm are the same.

Ah but I see you mean delta alpha. Yes, that needs fixing.

You need to multiply delta_alpha by a^2 to get a proper delta t in the function.

Can you multiply by a^2 here?

We can do the proper passing of delta t alpha later.

added 1 commit

• 42fcf649 - Small fixes for consistency / clarity

Compare with previous version

I also added a few minor changes (e.g. making

$\mu_0$

a float everywhere) I would personally like to be implemented. Lmk what you think and I can include them to FDI and VP if you agree.

Let's not touch VF or FDI for now.

Can you run the formatting script please?

@fastasyszyn any thoughts?

So, to understand: we keep the values for full and predicted values. Use the correct one in the accelerations. It has sense to me. Not sure about the calculation of the errors, but if we agree in zeroing instead of caping, no problem. Let's port it to the other schemes

You mean the disappearance of the additive 1e-18 ?

yes

Likely does not make a difference in practice but I prefer this version as the 1e-18 is introducing a scale that is not related to unit choices

I can add the VP and FDI changes... in this branch or just merge this and I open another?

For VP and FDI, let's make another branch.

mentioned in commit 9470847b

merged