search

inducer / meshmode

High-order unstructured mesh representation and discrete function spaces
https://documen.tician.de/meshmode/
25 stars 24 forks source link

Newest meshmode gmsh ingest chain has some problems #414

Open MTCam opened 1 month ago

MTCam commented 1 month ago

Thanks @inducer for the new updates to gmsh ingest - bunch of improvement there. There still seems to be a couple of issues with 2D that we should work out before nailing down 3D.

In short, we noticed that previously working mirgecom TPE cases began to fail on gmsh_interop latest update.

Specifically, these old cases all had negatively oriented (clockwise connectivities) elements because that was previously required to make gmsh meshes work in mirgecom. Those meshes can be ingested without error when sending the force_positive_orientation=True option, however they fail to work in mirgecom because this option appears to lose boundary information (see #413 ).

Interestingly, reverting gmsh_interop to its previous version (ed0bff5) allows the old negatively oriented meshes to be read without any errors at all and then they run fine in mirgecom.

I think this issue is effectively demonstrated/reproduced with very simple meshes in #413.

CC: @anderson2981

MTCam commented 1 month ago

One potential fix is propagating the face_vertex_indices_to_tags map through to the element flipper. See (https://github.com/illinois-ceesd/meshmode/pull/7)