Hi Lattice Boltzamnn and CFD experts,

I'm implementing MPI for a 3D lid-driven cavity flow using a basic setup with just two processes. I have successfully implemented it with a simple case where one side has a value of one and the other side has a value of zero. I have also added ghost cells to test the MPI communication, and it works correctly.

However, when I incorporate collision and streaming and place the MPI transfer between them, the results remain unchanged. I have placed the MPI transfer at the beginning. It produces some results, but it's strange that the order of operations matters. The MPI transfer is only for the ghost cells, while the other parts should undergo collision and streaming. It's puzzling why the results remain unchanged. On the other hand, the serial version of the code works fine.

Thank you in advance for your advice.