[M.Davoodi]

Hi Guys,

I have been trying to add a small function at the end of controlDict to calculate \int(alpha.dv) above and below the line y=0 in my geometry.
So I have put an "if" condition to separate y=0 and then the integral will be simply sum of alpha*Cellvolume.
The codes works fine, but it produces different results when I am running it in a "single core" or when it is run in "parallel". REALLY ODD!

I have used other functions both in parallel and in single core with the same solver and none of them have this issue. Am I missing something here?

Does anyone have any idea what I am doing wrong?


the code is provided bellow:





functions
{

Wout
{
functionObjectLibs ("libutilityFunctionObjects.so");
type coded;
redirectType calCd;
writeControl timeStep;
writeInterval 10;
enabled yes;

codeWrite
#{

// Lookup/create variables

const volScalarField& alpha1 = mesh().lookupObject<volScalarField>("alpha");


const volVectorField& C = mesh().C();
const scalarField& volume = mesh().V();

scalar totalVolumeUp(0.0);
scalar totalVolumeDown(0.0);


forAll(C,idx)
{
if (C[idx].component(1)>0)
{


totalVolumeUp += volume[idx]*alpha1[idx];


}
if (C[idx].component(1)<0)
{

totalVolumeDown += volume[idx]*alpha1[idx];

}

}

scalarList list;
list.append(mesh().time().value()); // Time (col 0)
list.append(totalVolumeUp); // vol up (col 1)
list.append(totalVolumeDown); // vol down (col 2)
list.append(totalVolumeDown+totalVolumeUp); // sum vol (col 3)
list.append((totalVolumeDown-totalVolumeUp)/(totalVolumeDown+totalVolumeUp)); //the lifting parameter (col 4)
// Write data

string comsh;
string filename("LiftingParameter.txt");
std::stringstream doub2str; doub2str.precision(18);

comsh = "./writeData " + filename;
forAll(list, id)
{
doub2str.str(std::string());
doub2str << list[id];
comsh += " " + doub2str.str();
}

if (Pstream::master())
{
system(comsh);
}

#};
}

}

[Tobermory]

totalVolumeUp += volume[idx]*alpha1[idx];

Be careful if you are calculating sums across all the cells in a parallel run ... each processor only knows about its own slice of the domain. To get the complete sum, you need to use functions like gSum, to sum across all processors. Try googling for more info.

[M.Davoodi]

Originally Posted by Tobermory

Be careful if you are calculating sums across all the cells in a parallel run ... each processor only knows about its own slice of the domain. To get the complete sum, you need to use functions like gSum, to sum across all processors. Try googling for more info.

Hi Tobermory,
Thanks for your reply. This is exactly what I was thinking and was worries about.
I agree gsum would be an excellent choice but I am not sure if there is a way that I can use the "if" condition to separate the line y=0 in combination with gsum. Are you aware of a way for doing this?

Cheers

[gionni]

I'm not sure if it's relevant, but you could try to implement the function in the solver itself.
I had a similar issue, where I needed to evaluate the maximum of a field in a BC: doing it in the BC itself would lead to the maximum being evaluated in the partial domain only, while doing it in the solver solved the problem.
I know it doesn't answer your question, but might be an alternative way to achieve what you need.

[M.Davoodi]

Originally Posted by gionni

I'm not sure if it's relevant, but you could try to implement the function in the solver itself.
I had a similar issue, where I needed to evaluate the maximum of a field in a BC: doing it in the BC itself would lead to the maximum being evaluated in the partial domain only, while doing it in the solver solved the problem.
I know it doesn't answer your question, but might be an alternative way to achieve what you need.

Hi Gionni,
Thanks for the response.
I actually managed to sort the issue out by adding this line at the end of the forAll loop condition:

reduce(totalVolumeDown, sumOp<scalar>());// Sync sum across processors

You might find it helpful too