Tomás Chor

> We should make the changes needed for both nonhydrostatic and hydrostatic models. @glwagner the reason why I hadn't removed masking from the hydrostatic model is that I'm really not...

> We have to keep masking for velocities in the hydrostatic model. > > In the nonhydrostatic model, the equivalent masking procedure acts on the predictor velocity field prior to...

@simone-silvestri @glwagner I think tests should be passing now. Ready for reviews

Also, I don't understand why we're calling masking events at update for every `NonhydrostaticModel` regardless of it having an `ImmersedGrid` or not: https://github.com/CliMA/Oceananigans.jl/blob/d4a45adf21918f95ca4d23ec7167e720c44c501b/src/Models/NonhydrostaticModels/update_nonhydrostatic_model_state.jl#L14-L19 The same goes for the hydrostatic model...

Not sure if I should open another issue for this, but I was trying to set-up a 2D topology like `(Bounded, Flat, Bounded)` but couldn't. Apparently, however, `(Flat, Bounded, Bounded)`...

> most of the time we are just using ordinary numbers, fields, etc as "computed dependencies" and in that case it's silly to call them "computed dependencies". My impression is...

> What if we get rid of `parameters` and `computed_dependencies`, and call `compute!` on all the arguments? That'd work for me. Would that have the same performance or would it...

If it's the same, then I'm okay calling compute on everything and don't separate between `parameters` and `computed_dependencies` anymore. I think we gotta make that clear on the docstring though!

Good catch! I think we should test that, but indeed a slimmer version of that test is preferred.

> Why though it takes 1.5hrs? Any ideas? Not sure. It runs two models for over 3000 time steps, but the GPU should make that pretty fast. The only "unusual"...