New feature: DOLFIN Quad/Hex mesh support

43 views
Skip to first unread message

Jack Hale

unread,
Sep 5, 2017, 7:12:55 AM9/5/17
to fenics-support, fenic...@googlegroups.com
Dear all,

We are delighted to announce that DOLFIN master branch now contains support for solving PDEs on Quadrilateral and Hexahedral (quad/hex) cell meshes. This feature will be released in the upcoming 2017.2 version of the FEniCS Project. This was implemented by Ivan Yaschuk under the Google Summer of Code program through our fiscal sponsorship arrangement with NumFOCUS. You can read Ivan’s detailed technical notes here: https://nbviewer.jupyter.org/gist/IvanYashchuk/48bff62b155e31f85de975b24b23e4cb

We are looking for more in-depth testing and feedback from the community on this new feature. The easiest way to try the new feature is to use our pre-built Docker containers:


If you do find any issues, please file an issue on the appropriate Bitbucket bug tracker https://bitbucket.org/fenics-project. If in doubt, file under the DOLFIN tracker and we will move it to the appropriate place.

Summary of the new functionality:

- Support for FiniteElements of the Continuous and Discontinuous Lagrange family (no H(curl) or H(div) conforming elements, yet).
- Many existing user codes can be changed to use quadrilateral or hexahedral meshes simply by changing the mesh construction. The following documented demos now use quad/hex:


- The following DOLFIN features are currently unsupported when using quad/hex meshes: 
    - point evaluation (*) (requires changes to FIAT and FFC, see Ivan’s notes)
    - CellSize(mesh) UFL/DOLFIN primitive (*) (already in progress)
    - H(div) and H(curl) elements (-) (requires implementation of consistent mesh ordering (facets/faces/edges) in DOLFIN mesh class)
    - mesh refinement (-) (would require supporting meshes with hanging nodes)
    - collision detection (-) (requires new routines in dolfin/geometry)
    - VTKFile output (XDMF works, however)
    - matplotlib plotting (X3DOM works, however)

Any help implementing the above features would be greatly appreciated! Stop by #development on Slack and we will be more than happy to assist.

The features marked (*) will be fixed as soon as possible. The features marked (-) are long term goals. Unmarked features are unlikely to be implemented as their are sufficiently good alternatives in place.

Jack Hale
Research Scientist
University of Luxembourg 
Reply all
Reply to author
Forward
0 new messages