look the constructors

polyMesh&

so no overhead

I see, if it would help I can always pass the mesh in the second method the mesh as a reference and then loop over it. I build the mesh with blockMesh so it gurantees the indexing.

> I see, if it would help I can always pass the mesh in the second method the mesh as a reference and then loop over it. I build the mesh with blockMesh so it gurantees the indexing. @User I'm not following you

sorry

When running the solver the mesh is also passed as a reference on each processor right? Do the two methods should be equivalent then?

Depends on how you pass your mesh to your subroutine. The mesh object is present in the objectRegistry db()

you can always retrieve a reference of the mesh using the lookUp method

Alright I'll look into that, this is helpful. Anyhow I need to assess what's the load when computing with the averages and without. Thanks

Will hit this chat if I face something.

arithmetic operations are almost inconsequential

I have been just summing over and then use reduce at the end of the loop then average at each z index.

If I compute the grad(U) once every time step and then use this sum operation for each z I believe it should be identical unless I'm missing something crucial.

Why you reduce? I don't see reason for that

You need to collect the sum of say Uavg on each processor right? And then get U(z)_global =U(z)_proc1+U(z)_proc2..... So on

That explains why you are forced to split the mesh as you do 🙂

Exactly! I don't see it as a major limitation since the mesh will be equally distributed on each processor.

Anyway, good luck with the performance analysis

Thanks

This was informative ,thank you.

Hi, I am doing LES on a locally refined mesh, so I need to smooth delta. I have found that the ratio of deltas between two cells is no larger than a specified amount, typically 1.15, as it says in "smoothDelta.H". But I don't know how to set the parameters in "turbulenceProperties". What is the "maxDeltaRatio" of the smoothCoeffs? Do I need to set this value to 1.15? Thanks

Some questions: (1) explicit/implicit LES? (2) By locally refined, What do you mean exactly?

the 1.15 you refer to is related to the refinement as in clustering (Geometrical) of cells normal to a wall. I assume you got that value from the work of Komen et al.

Hi @User , thanks for your reply. (1) I think openfoam uses implicit LES (2) the value 1.15 is the maximum ratio of delta between two cells, I read it from another paper, and the description in "smoothDelta.H" also said that. I just wonder what does maxDeltaRatio mean? do I need to set it to 1.15? I cannot understand the code...

Answers: (1) OpenFOAM is a library for FVM, the choice of turbulence model, is yours. If you refer to the assumption in which the mean value theorem used to formulate FVM acts as a median filter for EXPLICIT LES MODELS, at the cell level, then yes, OpenFOAM, being thought for low-order FMV, "implements" that hypothesis; (2) The description for smoothDelta is sufficiently clear: the value of delta = (dx*dy*dz)**(1/3) in a cell does not vary more than maxDeltaRatio w.r.t its neighbouring cells

@User Thank you so much! That is clear!

You didnt answer any of my questions, though...

Sorry, I did't see your questions, I supposed you were answering mine...