Solid cell/nodes in Amrex without using Embedded Boundary functions

[LordOfBunnies]

I'm currently writing a Lattice Boltzmann Method flow solver in Fortran 95. The code is node based, making it a little more fun.

I'm wondering about a way to specify solid geometry without using the Embedded Boundary system already in Amrex. My code is able to determine whether a node is within a solid. It takes .stl files or .cgns files with tessellated surfaces. However, Amrex does not know if the node is a fluid or solid. I figure this may end up causing problems when using things like average_down, fill_patch, etc. when it's near the wall.

I've been reading through the embedded boundary system, but it doesn't look like it can handle generic geometry. I could certainly be wrong.

Is there a way to inform Amrex which nodes are solid/fluid/etc.? Can the embedded boundary system be used with generic geometry? Is there a way to work around the system and directly tell the EB system which ones are which?

[WeiqunZhang]

What EB information do you need? If it's just whether a node is inside or outside a solid body, you can save that information in an iMultiFab.

[LordOfBunnies]

I'm already saving the information in an iMultiFab, but I didn't see a way to let Amrex know that information. It may be because I'm using Fortran and the EB stuff is C++. I'm worried about using things like fill_patch and average_down as they might pick up bad information from inside the solid.

Right now, I'm designating things as fluid or solid, and that determines how the equations interact. I don't think Amrex knows that those nodes are fluid/solid/etc. Is it using EBCellFlagFab? Can that work for nodes as well? I've already got the system doing interpolative bounce-back, so I don't need cut cell information, just whether it's a valid node or not.

[WeiqunZhang]

Right. AMReX won't know about the EB information. But you might be able handle it yourself.