CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM (http://www.cfd-online.com/Forums/openfoam/)
-   -   error directMappedPatch LES and parallel (http://www.cfd-online.com/Forums/openfoam/74835-error-directmappedpatch-les-parallel.html)

Thomas Baumann April 9, 2010 12:32

error directMappedPatch LES and parallel
 
Hi all,

I'm simulating a channel flow with LES. I have built a the channel in blockMesh, the inlet is a directMappedPatch. Starting this simulation not parallel works good. Starting the simulation parallel everything works at the beginning, too. But using "reconstructPar" I get following message:Create mesh for time = 0

Time = 0.0002

Reconstructing FV fields

Reconstructing volScalarFields

p
k
--> FOAM Warning :
From function average(const UList<Type>&)
in file /home/dm2/henry/OpenFOAM/OpenFOAM-1.6/src/OpenFOAM/lnInclude/FieldFunctions.C at line 458
empty field, returning zero


Did not find sample (1 0.00166459 -0.0496671) on any processor of regionregion0

From function directMappedPatchBase::findSamples(const pointField&, labelList&, labelList&, pointField&)
in file directMapped/directMappedPolyPatch/directMappedPatchBase.C at line 337.

FOAM exiting


I have tested many things, but nothing helps. Has anybody an idea? I'm using OpenFOAM 1.6.

Thanks, Thomas

Thomas Baumann April 12, 2010 13:12

Hi all,

the problem is the same as cyclic boundary conditions with parallel running jobs. So you have to make a decomposeParDict, where the directMapped bc's are at one processor (inlet and outlet) dictionary. That means you can split the inlet in p.e. four processor domains, but each one must have the according outlet-part, too.

For example the simulation of a channel with mapped bc's for a fully developed flow in x1 direction can be solved parallel with these conditions in the decomposeParDict for 4 processors:

numberOfSubdomains 4;

method hierarchical;
//method metis;
//method parMetis;

simpleCoeffs
{
n (1 1 1);
delta 0.001;
}


hierarchicalCoeffs
{
n (1 4 1); //here 4 processors in x2 direction
delta 0.001;
order xyz;
}


You can see in x1-direction is only one processor. So inlet- and outlet-parts you need for the directMappedPatch are saved at one processor.

So it works now. For further informations search the forum with the catchwords "cylic" "parallel".

Regards Thomas


All times are GMT -4. The time now is 14:10.