[AmrShaalan]

Hello,

I have 2 cases with one using dynamicRefineFvMesh and another not using dynamicRefineFvMesh. The case without refinement is at about 6 million cells and takes an average 3 seconds per tiemstep. I coarsened the mesh and started the dynamicRefineFvMesh case so it starts at a lower cell count (3 million) and rely on automatic refinement instead.

The automatic refinement case starts really fast (as expected). But even with the first refinement cycle that adds about 500,000 cells, I see that the speed goes dramatically down (starts taking 5 seconds per timestep at a much lower cell count than the static mesh). I understand that adaptive mesh refinement adds a time penalty, but it should be for the time taken to refine but the timestep itself should be comparable to the static mesh. I think the culprit is that the refined mesh is not optimized by the "renumberMesh" so I was wondering if I it is possible to do automatic renumbering after each refinement attempt?

Note: I am of course comparing the two cases with the same hardware (same computing nodes and number of cores).

Would really appreciate any help or if someone points out something wrong in what I said.

Thanks!