Physical Boundary Conditions - Fortran

[pbrady]

I cannot find any example fortran code that uses the amrex_physbc fortran interface to supply boundary conditions. The current examples in Tuturials/*_CF use periodic bcs. I did notice that in some of the _C tutorials, an f77 file was present which provided a filcc routine. To impose boundary conditions, should I customize the filcc routines or is there a better way utilizing amrex_physbc?

[drummerdoc]

filcc is an app-generic routine to handle reflections, extrapolations, etc, i.e., grow cells that don't require user intervention to fill. Typically, one calls filcc first to fill grow cells on such boundaries, and then follows in the bc routine with loops over each face of the box to be filled, checking for Dirichlet conditions that weren't handled by filcc (i.e., bc(dir,lohi,comp) == EXT_DIR, where dir=1,2,3 and lohi=1,2) and if the box to be filled sticks out past the boundary on that face. If such cases are found, the user-supplied code will need to set the Dirichlet values explicitly. Note that the values set in these grow cells will be enforced on the boundary FACE. We should add such an example to the tutorials.

[WeiqunZhang]

I will update amrex/Tutorials/AMR_Adv_CF/Source/fillpatch_mod.F90's subroutine fill_physbc to show an example of setting physical boundary conditions.

[zingale]

here's an example from Castro:

https://github.com/BoxLib-Codes/Castro/blob/development/Exec/hydro_tests/double_bubble/bc_fill_2d.F90

[pbrady]

Thanks for the example. What's the use case of the amrex_physbc_* routines in AMRex_physbc_mod.F90?

[WeiqunZhang]

The Castro example was built with C++. As I understand, you are trying to use the Fortran interfaces Src/F_Interfaces.

[WeiqunZhang]

Please check out the latest development branch. Tutorials/AMR_Adv_CF/Exec/SingleVortex has been modified to use a simple extrapolation for physical boundaries (https://github.com/AMReX-Codes/amrex/tree/development/Tutorials/AMR_Adv_CF). In inputs.physbc, we have geometry.is_periodic = 0 0 0 to set the boundaries to non-periodic. In https://github.com/AMReX-Codes/amrex/blob/development/Tutorials/AMR_Adv_CF/Source/my_amr_mod.F90#L74, we set the non-periodic bc to first-order extrapolation. The fillpatch module (https://github.com/AMReX-Codes/amrex/blob/development/Tutorials/AMR_Adv_CF/Source/fillpatch_mod.F90#L62) is modified accordingly.

If there are any problems, let us know.

[pbrady]

Thanks for the updated example. It's very helpful.

So, if my problem specific boundary conditions are not handled by amrex_fillcc I would modify the fill_physbc subroutine to appropriately account for this (i.e. time dependent dirichlet bc). In that case, I wouldn't need to change to the fillpatch/fillcoarsepatch wrappers in the AMR_Adv_CF example, just the fill_physbc subroutine which is passed to the amrex_fillpatch routines.

However, If my boundaries warranted a unique discretization stencil, then I would have to take this explicitly into account in whatever problem specific "advance" routine I write. And based on @drummerdoc's comments, a custom discretization may be required depending on if the boundary conditions are interpreted as face of cell data.

Does this sound right?

[drummerdoc]

Yes, that is correct. The linear solvers, for example, do this explicitly with polynomial extrapolation/interpolation.

[WeiqunZhang]

First, I forgot to commit a new file. So you need to pull the latest version.

What Marc (@drummerdoc) said about face is true in some cases where filcc is used. It is however not always true. For example, in the hyperbolic solver of Castro, the ghost cells values are always treated as cell centered, not on faces. So it depends on how the application code views it.

[pbrady]

I pulled the latest version and everything seems clear. This issue can be closed unless you think it would be better left open for a while as guide

[WeiqunZhang]

The tutorial mentioned above has been moved to https://github.com/AMReX-Codes/amrex/tree/master/Tutorials/Amr/Advection_F/