[jhierroa]

I am interested in knowing whether there is any way to work with a mesh divided into different sets (for instance rings around a central cylinder with different transport coefficients) where the final computation of the equations is, moreover, divided into different processors (with divisions along the z axis).

In other words:

Let it be a cylinder with rings around it with different transport properties and each cylinder is modelled by a different set (cellSet is used in the main code to set the transport properties in each ring). Is there any chance to compute the resulting system in parallel?

NOTE: a standard use of decomposePar works fine with everything (faces, nodes, points, patches, ...) except for the definition of the different sets which are not decomposed among the different processors.

[mattijs]

decomposePar does decompose zones (cellZones, faceZones, pointZones). Run setsToZones to convert your sets to zones and use e.g. the 'zoneToCell' cellSet source to convert back.

[gschaider]

Hi Juan!

I have written a small modification to decomposePar that decomposes the sets as well. If you're interested, send me a mail.

[jhierroa]

Yes, I am interested in that small modification. I have already sent you a mail.

[lr103476]

Hi Bernhard,

This post is quite old, but I'm also interested in the modification of decomposePar to decompose cellSets......Could you post it here ?

Thanks and regards, Frank

[gschaider]

Hi Frank!

I've adapted it to compile on 1.4.1 (due to that it won't compile on 1.4 or lower) and put it on the svn on openfoam-extend. You can get it (and a test-case) with this command:

svn checkout https://openfoam-extend.svn.sourcefo...oseParWithSets

As an alternative I'll attach a patch here that contains all the necessary changes to the original decomposePar-sources and should work on 1.3 and 1.4 too.

I havn't used the modification in a long time, but it still seems to work

Bernhard


domainDecomposition.C.patch

[lr103476]

Thanks Bernhard!

It seems to work. I now have decomposed a case with two mesh subsets, but something real strange occurs:

The wing motion of the same case with 2 subsets is different in parallel compared to the serial case. Very strange, but at least I can decompose now.....thanks for that.

Frank

[cricke]

Hi
I've tried setsToZone but then get problems with memory allocation?!

Where do I compile the domainDecompositiion.C.patch file? Does it affect my original decomposePar or can I switch back if it doesnt work? What command to execute the new decomposePar?

RG
Christofer Ivarsson