Lucas:
I suspect this is a simple task, but I can't find anything which does it
outright. I'd like to find the local minima associated with some quantity
derived from a finite element field. The features that I'm looking at are
reasonably well-behaved, in that they're smooth except exactly at the minimum
(where they're continuous but not differentiable), and of predictable size.
For a uniform mesh I think a reasonable (naive) approach would be to calculate
the derived quantity at the quadrature points of each cell, then go through
each cell and check using the CellAccessor::neighbor() function (possibly in a
recursive manner, until we reach some distance \delta x from the original
cell) whether that cell contains the minimum of all its neighbors (or
neighbors-of-neighbors or whatever). However, for an adaptively refined mesh,
the cell returned by neighbor() might be further refined, in which case some
of the child-cells border the original cell, and some do not. One related
question is: how do I access just those child cells which border the original
cell? What if those child cells also have children? And are there other
complications of this sort which might arise from trying to iterate through
neighbors of different refinement?
cell->neighbor_child_on_subface() is what you're looking for.
As an additional complication, I'd like to be able to do this process on
distributed meshes, and I suspect that the features may be cut up by the mesh
partition given that there is a lot of refinement happening in those regions.
Is there any way I can easily do some kind of point-to-point communications to
check whether the local minimum has been split along a mesh partition?
This seems like the kind of thing that is common enough to have already been
implemented, probably in a more sophisticated way, but I'm not sure where to
look. Any help is appreciated!
Unless computing the derived quantity is exceptionally expensive, I will
venture the guess that
* computing the derived quantity at all quadrature points of all locally owned
cells
* storing the quadrature point with the currently least value on the current
process
* comparing for the least value among processes
is fast enough. It's a relatively cheap O(N) operation, it should be vastly
faster than assembly or linear solver. It might not be worth optimizing it any
further.
Best
W.
--
------------------------------------------------------------------------
Wolfgang Bangerth email: [email protected]
www: http://www.math.colostate.edu/~bangerth/
--
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see
https://groups.google.com/d/forum/dealii?hl=en
---
You received this message because you are subscribed to the Google Groups "deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/dealii/02d1f0cc-e401-cb43-7219-e77c07da52a5%40colostate.edu.
