Re: Questions on parallel::distributed::GridRefinement::refine_and_coarsen_fixed_number in deal.II 9.0.1 released

[Wolfgang Bangerth]

On 11/5/19 3:55 AM, kwan...@163.com wrote:
>
> Thanks for your kind suggestions. If I set the top fraction to 1.0 and the
> bottom fraction 0.0, my grid is almost uniformly coarsed by one level (a few
> cells are even coarsed by two levels). This is the case when I use a fine grid
> (8 cells per periodicity) before the refinement. If much coarser initial grid
> is used instead, it works fine. Maybe the grid refinement scheme is not
> applied for my case. Right?

I really can't tell without the code, but I do know that essentially every
deal.II-based code uses these functions and it seems to work for everyone else
-- in other words, it seems unlikely that the functions you mention are
fundamentally flawed. In any case, each call to
triangulation.execute_coarsening_and_refinement() only refines by one level,
so when you say "a few cells are even coarsened by two levels" makes me think
that something is not quite the way you describe it: Maybe you only look at
the mesh you get after *several* refinement/coarsening cycles instead of one?

Best
WB

--
------------------------------------------------------------------------
Wolfgang Bangerth email: bang...@colostate.edu
www: http://www.math.colostate.edu/~bangerth/