Reimplement and parallelize adaptive mesh refinement
Adaptive mesh refinement needs a complete overhaul/rewrite:
- Both bisection and regular cut refinement are broken and may give hanging nodes (see separate bug report)
- Regular cut refinement only works in 2D
- Bisection refinement is slow
- No refinement algorithm works in parallel
Blueprint information
- Status:
- Not started
- Approver:
- Anders Logg
- Priority:
- High
- Drafter:
- Andy R Terrel
- Direction:
- Needs approval
- Assignee:
- Andy R Terrel
- Definition:
- Approved
- Series goal:
- None
- Implementation:
-
Unknown
- Milestone target:
- None
- Started by
- Completed by
Whiteboard
ART: I will take over this blue print. The idea is to add this feature from a well-maintained open source package rather than reimplement our own. There are two tracks find a suitable package for simplicies and one for quads and hexs.
ART: There are several moving parts to this blue print. Most especially the actual implementation base. Since we are adding quad/hexes at the same time, I am postponing this work until that is done. Currently the plan is to use three different approaches all of which have good uses.
pragmatic: https:/
dolfin-hpc: (port back to mainline dolfin) https:/
p4est: http://
Work Items
Work items:
Evaluate options for integration: INPROGRESS
Implement simplices: TODO
Implement quads/hexes: TODO
Dependency tree
* Blueprints in grey have been implemented.
