Weird pre-inlet output

[mojtabaAmiraslanpour]

Hi,

I have been trying to use manual decomposition of the domain for both pre-inlet and the main domain while using the curvedflow_with_preinlet solver. I get weird results with different decomposition configurations:

And as I know I put the pre-inlet slice in the correct position.
Not sure what is happening. Any suggestions?

Best,
Moe

[mojtabaAmiraslanpour]

Just tested in two different clusters. One of them works fine, but the other produces these weird contours. I was thinking maybe the libraries are different, so I used the same conda environment on both clusters. Still happening and I don't have any clue what is going on.

[mojtabaAmiraslanpour]

Is there any chance that CPU type affects the results?

[gzavo]

Sorry I don't have an answer for this one. CPU vendor, etc. of course should not have effect on the functionality.
Since it happens on a boundary, maybe some problem around the MPI calls? I would first look at the MPI versions and implementation, and if the same is used for compilation and execution.