[nago]

hi,

i use OF 2.3.1.
i want to get average pressure of all cell in calculation region.
so i try to get it using "forAll(mesh.cells(),cid)" method .
At single core calcuation, i can get average pressure.
but in pallarel calc, forAll statement only return one-processor cell.

・for example, simplefoam pitzDaily model,
i only add following term in simplefoam.C (named my_simpleFoam.C).

---------------------------------------------
:
:
laminarTransport.correct();
turbulence->correct();

forAll(mesh.cells()){
Info << "cellID : " << cid <<endl;
}

runTime.write();
:
:
------------------------------------------

when calculate pitzDaily using my_simpleFoam in single core, output messege has 12225 cell IDs(All cells).
but in pallarel ( 2cores ), output has only 6113 cell IDs(half number of all cells !).

i thougtr forAll(mesh.cells) can look all cells.
what is this behaiver?
and how can i get all-cell averaged pressure?


regard.

[clapointe]

I think you can do it an easier way -- try adding this :

Info << "Average Pressure : " << gSum(p)/mesh.C().size() << endl;

gSum should take care of the parallel/ multi processor issue.

Caelan

[zhangyan]

Info<< "Mean pressure:" << p.weightedAverage(mesh.V()).value() << endl;

Ref: https://github.com/OpenFOAM/OpenFOAM...m/logSummary.H

[nago]

hi clapointe,
thank you for your quick reply.

i add gSum(p) in my_simpleFoam to check the behavior in pitzDaily.
---------------------------------------------------------------
turbulence->correct();
runTime.write();

forAll(mesh.cells(),ci)
{
const cell &c = mesh.cells()[ci];
p_total += p.internalField()[ci];
}
Info << "p_total : " << p_total << endl;
Info << "gsum Pressure : " << gSum(p) << endl;
Info << "Average Pressure : " << gSum(p)/mesh.C().size() << endl;

Info<< "ExecutionTime = " << runTime.elapsedCpuTime() << " s"
---------------------------------------------------------------

and I add following functionObject in controlDict.
------------------------------------------------------
press_sum
{
type cellSource;
functionObjectLibs ("libfieldFunctionObjects.so");
fields (p);
valueOutput false;
source all;
operation sum;
outputControl timeStep;
outputInterval 1;
}
press_aveg
{
type cellSource;
functionObjectLibs ("libfieldFunctionObjects.so");
fields (p);
valueOutput false;
source all;
operation average;
outputControl timeStep;
outputInterval 1;
}
------------------------------------------------------

after running 1step calculation,result is following.

(single)
p_total : 1.7186e+07
gsum Pressure : 1.7186e+07
Average Pressure : 1405.81
cellSource press_sum output:
sum(sampledSurface) for p = 1.7186e+07
cellSource press_aveg output:
average(sampledSurface) for p = 1405.81

(2 core)
p_total : 5.69918e+06
gsum Pressure : 1.72788e+07
Average Pressure : 2839.57
cellSource press_sum output:
sum(sampledSurface) for p = 1.72788e+07
cellSource press_aveg output:
average(sampledSurface) for p = 1413.4

in single calc , p_total and gsumPressure is same Value.
in 2 parallel, as expected,two value differs.

but in parallel calc, result looks like bad, especially in Average Pressure.
and i mind difference gsum Pressure between single and parallel.
in parallel calc, it looks "mesh.C().size()" picking up only one-processor cells.
why this happen?

[nago]

thanks zhangyan.

your advice is very useful for me.
i wanna use it.

BTW, for my interest, do you know why the problem in my previous post happen ?

[clapointe]

That's a good catch -- I'd tested it briefly and received something that looked reasonable -- sorry. Try this instead : gSum(p)/mesh.nCells(). This came directly from the aforementioned field average function object, so it should replicate it now.

Caelan

[nago]

hi clapointe,

i try "gSum(p)/mesh.nCells()" insted of "gSum(p)/mesh.C().size()".
but result does'nt change.


And now, i consulted OF information, i found "Reduce()" statment.
in parallel calculation, we have to use Reduce for all Regions.
I should have consulted more and found it but I could not ...

Thanks.

[clapointe]

Ok, with that hint lets try one more thing :

 label n = returnReduce(mesh.cells().size(), sumOp<label>());
 Info << "Average Pressure : " << gSum(p)/n << endl;

Caelan

[nago]

hi,

i try the code clapointe suggested.
finally i get AveragePressure, same as one from functionObject cellSource- average.

thank you for giving advice to my crude question.

Regard.