Rectangular domains without anisotropic cells

Open fractalsbyx opened this issue 7 months ago • 3 comments

Currently, the only way to do this is by increasing the refinement in one direction, but it doesn't work well with AMR. Might need to also be an integer-only type of thing. I was thinking connected cubes/squares. Is it possible?

$fractalsbyx avatar$ Jun 02 '25 13:06 fractalsbyx

Don't think there's much of a way around this with octree grid refinement.

At most, we can use 1 less subdivision with anisotropic grid refinement. Don't think this is supported well with the global coarsening multigrid.

Jun 03 '25 18:06 landinjm

Not planned for 3.0, but I'll leave the issue open for the future.

Jun 03 '25 18:06 landinjm

One potential way is to manually move the points on the coarsest mesh. However, this would be a bit involved when dealing with AMR.

Not planned for 3.0, but maybe in the future.

Jun 03 '25 19:06 landinjm