Dear all,
I have a question, that is can we control the domain decomposition by ourselves?
For example, for a fluid-structure interaction(large fluid and small solid), if I have ten processors, I want to allocate one for the whole solid domain and the else 9 for fluid part. Is it possible?
Yes, look for UserPartitioning in the examples/hpddm folder.
Many thanks for your reply. I have another question. I learned from your expert video on Youtube that ‘movemesh‘ cannot be directly used in parallel for distributed meshes because there could be incoherent on the overlapping mesh. Is there a good method to movemesh now? Actually, I have a idea, that is I firstly do a partition of unit to make the displacement function coherent, and then movemesh.
No, you need to centralize your solution, movemesh only on a single process, and re-decompose.
Got it, many thanks! Wow, it could be a challenge to move a fine 3D mesh. It may need enough memory, I think…….