How to setup a periodic simulation?

[dineshadepu]

Hi all,

For validation I need to do some periodic cases, such as, Taylor-Green vortex, and Poiseuille flow problems. Unfortunately, I couldn't set one up with the current documentation and example cases. Can someone help regarding this?

Many thanks.

[sslattery]

Hello

This example sets up a periodic mesh: https://github.com/ECP-copa/Cabana/blob/master/example/cajita_tutorial/10_fft_heffte/heffte_fast_fourier_transform_example.cpp

[dineshadepu]

Thanks for the example case @sslattery . Currently I am only dealing with numerical methods which are only related to particles and no mesh is involved. The code I have developed so far, I have only used Cabana with no Cajita's algorithms or mesh related functions.

I have checked this example, but I do not know if we can make the particles periodic as well. Can you please clarify on this?

Many thanks.

[streeve]

In that case it's much more closely tied to your algorithm/physics. Before you get into MPI you can just shift the particles by the box size if it crosses a boundary.

Otherwise start with https://github.com/ECP-copa/Cabana/wiki/Core-MPI-Communication. We have current work to add new features in Cabana to make this easier, but at the moment you will need to build the list of particles to be communicated (whether it's a periodic boundary or internal MPI boundary).

We have a few examples which do this, https://github.com/ORNL/CabanaPD only uses particle halos (particles never migrate), but shows how to use newer options for performant particle gathers. https://github.com/ECP-copa/CabanaMD is older and more complicated (uses a multi-pass approach), but has both particle gather/scatter and migration. Note these both use Cajita grids for convenience even though they're mesh-free

[dineshadepu]

Thanks for the detailed answer @streeve . I don't think I can currently implement this. I will get back to this once I start to use MPI.