Triangulation distributed without p4est

[Wayne901]

Hi dealii community,
I'm new to dealii and attracted by its comprehensive tutorials, but I have one question. Thats is can mesh be distributed in parallel computation without p4est framework? I came to this question when reading step-40 and it says previous mpi tutotrial like step-17 based on METIS partition keep a global triangulation in every process, which is unwanted for large scale multi-process projects. However I'm not sure if orthogonal meshes suit problems with complex geometries.
More specifically, I plan to try matrix-free method with multigrid method for unorthogonal quad/hex mesh (as in step1), and of course with parallelization. It's this possible in dealii? Thank you.

[drwells]

It sounds like parallel::fullydistributed::Triangulation might be the data structure you want.

However I'm not sure if orthogonal meshes suit problems with complex geometries.

In general, yes, but fundamentally p4est works well with any quad or hex mesh with a coarse grid of a few thousand cells which has been adaptively or uniformly refined - whether it comes from a cube or a mesh of some complex object isn't really relevant to p4est since it just manages topology.

If you still have questions, could you help us by posting them to the mailing list instead? There's a lot more people who subscribe to the mailing list than issue tracker and it's where we typically hold this kind of discussion. Thanks!

[Wayne901]

Hi Wells.
Great to here p4est works for arbitrary quad/hex mesh. I'll post questions to mail the next time.
Thank you!