[FLASH-USERS] boundary conditions (was Re: Typo in unsplit_rad Hydro_computeDT)

[Klaus Weide]

On Thu, 3 Mar 2016, Mark Richardson wrote:

>   1) Why would changing only the hydro boundary conditions impact gpot (and
> by ~20 orders of magnitude).

Not much use speculating as long as we are not even sure which Poisson 
solver was used to compute gpot (see my previous message).

However, as a general remark:

The boundary conditions set by [xyz][lr]_boundary_type should not be 
understood as just "hydro" boundary conditions; they actually affect the 
topology of the Grid. Think of them as "Grid boundary conditions". In 
particulary, if a boundary is "periodic", this changes the neighbor 
relationships between blocks; effectively a rectangular domain becomes a 
torus when all 2D boundaries are periodic, etc.

Not just hydro is affected by this. Combining a "periodic" Grid boundary
condition (in a given direction) with a non-periodic boundary condition
for a Poisson or diffusion solver is not expected to work, and may give
nonsensical results. Same for combining non-periodic Grid boundary 
conditions with periodic Poisson or diffusion boundaries.

Klaus