-
Notifications
You must be signed in to change notification settings - Fork 32
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Example for volOctree with refinement and MPI #424
Comments
You can have a look inside the test directory: "test/integration_tests/voloctree". The files with "parallel" in their name are tests that partition the mesh among the processes. |
Here a simple example on how the code needed to partition a VolOctree patch:
|
Dear Andrea Thank you very much for the example! I completely missed the parallel integration tests. I've added a pseudo-code snippet below that describes the current algorithm. Thank you for your help!
|
Usually, this type of error means that the input STL has some inconsistencies. For example it's not a closed solid, the normals are not pointing in the same "direction" (some are pointing inwards and some are pointing outwards) or there are some degenerate triangles (e.g., triangles with all three vertices along the same line). Would it be possible for you to perform some checks on the STL? |
Also, if you don't need the levelset sign outside the narrow band, you can disable sign propagation using |
Dear Andrea Thank you for your reply. The STL itself is definitely ok, because it works fine when not using MPI. My code currently doesn't decompose the STL data to different partitions, so every MPI rank should have the full STL data available. As for the sign propagation, unfortunately I need the sign to check for inside/outside. [edit] The STL I'm trying is a box, consisting of 12 triangles. |
The levelset should be updated also after the patch has been partitioned, try adding a |
I noticed that there were some other updates missing:
|
Ah that makes sense. Thank you very much! I will have to think about my algorithm a bit to account for all the required updates in parallel. Thank you very much for your help! |
Hi
Would it be possible to supply a small working example of volOctree with full MPI support (partitioning and rebalancing after refinement/coarsening)?
Or does voloctree_adaptation_example_00001 already provide that? It seems to me that the current version would cause every partition to generate the grid independently.
Thanks
Dave
The text was updated successfully, but these errors were encountered: