-
Notifications
You must be signed in to change notification settings - Fork 10
Open
Description
While mesh generation doesn't care about order of the topology, this is very relevant for the associated data once you start looking at chunking. I.e. if you do spatial subsetting, you'd like the data to be "close together". A space filling curve like a Hilbert curve gives a reasonable estimate of this closeness.
I think it makes sense to do a sort, either on vertices or on faces of the generated mesh, perhaps enabled by default (controlled by some enumerator). This should smooth over workflows, since the alternative is:
- An arbitrary order by triangle or gmsh
- Associate data with this order
- Find out it has terrible locality
- Reorder the topology, but now with ALL the data that's already been associated, generating enormous churn
We can nip this in the bud by providing a reasonable ordering by default.
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels