Tomás Chor

> What are the differences? I haven't plotted/analyzed the solutions for the regression tests carefully yet (tbh I'm not familiar yet with how the regression tests work and how the...

> Don't the regression tests output text indicating how many grid points are different, and what the maximum differences are? Yes! Sorry, is that what you were asking? I apologize,...

> > and a scalar-index issue on GPUs. > > What's the scalar indexing issue? It's already fixed. It was a test that was failing because we were comparing point-wise...

In order to build confidence and move forward with this PR, here's a comparison between the same simulations on `main` and on this branch. I'm using a channel set-up (`Bounded,...

> @tomchor the description of this issue is a little vague. The dynamics weren't "weird" --- the problem was a bug in reconstructing buoyancy in the momentum equations, right? It'd...

> The issue was that buoyancy was not reconstructed properly in the vertical. (There was also a problem with reconstructing buoyancy in the horizontal, but this only affects tilted domains.)...

> What do you mean by cleaner code? You mean `update_state!`? We need all of these functions still for the hydrostatic model so I don't think on the whole there's...

> I think it makes sense to merge #3188 before this PR --- if we can show that the new pressure solver in #3188 is unaffected by this PR for...

I'll leave this here for the record. I'm currently experiencing the first significant dynamical difference I've seen so far between the model with and without the pressure separation. In a...

I'm not sure what the state of this PR is, but last I checked it was working, but was about 5 times slower than the FFT-based solver, no? Coming from...