[jianfeng]

Dear all,
I'm working on OpenFOAM2012. I'm using a interFoam with dynamicRefineFvMesh to compute a gas/fluid interface. The mesh is decomposed at t=0 and the problem is solved for some time, like t=t0. To achieve a good load balance, I want to make a mesh decomposition again. The function 'redistributePar' has been tested in my study, but I failed to achieve a new mesh decomposition with right cells and interpolated solution on them. Will anyone tell me the right steps to work with 'redistributePar'? Thanks very much.

[HPE]

Hi,

Any chance to elaborate on "The function 'redistributePar' has been tested in my study, but I failed to achieve a new mesh decomposition with right cells and interpolated solution on them" with a minimal working example, if possible? Thanks!

[clapointe]

What you're looking for is commonly referred to as load balancing. This has been published e.g. by Rettenmaier et al for interFoam (https://github.com/ElsevierSoftwareX/SOFTX_2018_143). I think some derivative of their code may have even been integrated into v2012 -- check out dynamicRefineBalancedMesh.

Caelan

[guin]

Originally Posted by jianfeng

Dear all,
I'm working on OpenFOAM2012. I'm using a interFoam with dynamicRefineFvMesh to compute a gas/fluid interface. The mesh is decomposed at t=0 and the problem is solved for some time, like t=t0. To achieve a good load balance, I want to make a mesh decomposition again. The function 'redistributePar' has been tested in my study, but I failed to achieve a new mesh decomposition with right cells and interpolated solution on them. Will anyone tell me the right steps to work with 'redistributePar'? Thanks very much.

I suggest you to take a look at following thread: Run-Time Parallel Load Balancing

[HPE]

- No code has been implemented from Rettenmaier et al.'s work into any of OpenFOAM software, I'm afraid.
- Maybe this package can be useful: interIsoFoam + multiDimAMR.

[clapointe]

I must have misremembered and tried it myself, thanks for checking. If you check out the info provided with the other repo, though, you'll see they reference Rettenmaier et al as well and it looks like they have it available for e.g. v2012 (Henning Scheufler's code does appear to be available as of very recently : commit). Regardless of where it comes from, the load balancing code is available for recent versions of interFoam.

Caelan

[HPE]

multiDimAMR externally uses Rettenmaier et al.'s work which is not available in v20*. danke.