[kgkara]

Hello foamers!
I would like to ask if preserving cyclicAMI boundaries when decomposing a case is "mandatory" as a prerequisite for an "accurate" simulation.

Let me present some details:
I am using OpenFOAM-v2012 for an unsteady simulation. I have 2 runs with buoyantPimpleFOAM and time-step of the same order of magnitude; one simulation in which the couples of the cyclicAMI patches are preserved in the same processor and a second one in which they aren't. The results are different (average pressure on one of the cyclic patches is 2% different between the 2 simulations). By changing the reynolds number, this difference become bigger.

In some websites, it is suggested to preserve the cyclicAMI patches when decomposing the mesh. My question is if this is confirmed as a bug/problem or just a best practice for numerical reasons. Also, does someone have a rigid explanation about why this caused diffrerences before starting to dig into the code?

Thanks in advance.