[thomas]

I encountered a problem when doing mesh motion in parallel.

When I run the tutorial case movingCone, using the icoDyMFoam solver, and do a domain decomposition having 2 domains in y-direction, the number of iterations for solving motionUx/Uy/Uz increases by a factor of 10 or more compared to solving in seriell. Also, the motion solver now solves for motion in z-direction, although it's a 2-dimensional problem and 2d-motion is set to yes.
And, in some time steps the maximum number of iterations is reached.

Solving for the fields U and p, there's not much difference between seriell and parallel runs.

Using 4 domains, the behavior is similar.

I am doing the simple method and use the developer version 280307 from the fsb server.

I know that using 2 or more processors for 2000 elements does not really make sense, but I do not understand this behavior and would like to know where it comes from.

Thank you very much!

[thomas]

Does no one have an idea? Should I give more details?

Thank's very much!

[fra76]

I used parallel mesh motion with 3D cases, both with OF 1.3 and OF 1.4, and I didn't notice this kind of problems...

Which version are you using? There are a lot of changes in the mesh motion libraries between 1.3 and 1.4 ...

[thomas]

I tried the official OF 1.3 and the 2 developer versions available from the FSB-web site. I did not try OF 1.4

I also did mesh motion in parallel on large problems with 100.000 to 500.000 cells, here it was no problem, but in the tutorial case with only 2000 cells the behavior is strange.

I am using the tetDecompositionMotionSolver, which is based on a node-mesh (FE), as I think.