CFD Online Discussion Forums - The effect of write precision on CFD solution with codeStream

Hi,

Following my previous post, https://www.cfd-online.com/Forums/op...operators.html, I met another "funny" thing. I am new to OF and did not find this information anywhere so I decided to share it. I am using OF2106.

If one is using codeStream/setExprField, and I guess coded boundaries conditions. The writePrecision parameter in controlDict can have some significance!

I have the following codestream code for defining the internalField in p:

Code:



dimensions      [1 -1 -2 0 0 0 0];



internalField nonuniform #codeStream

{

    codeInclude

    #{

        #include "fvCFD.H"

    #};



    codeOptions

    #{

        -I$(LIB_SRC)/finiteVolume/lnInclude \

        -I$(LIB_SRC)/meshTools/lnInclude

    #};



    codeLibs

    #{

        -lmeshTools \

        -lfiniteVolume

    #};



    code

    #{

        const IOdictionary& d = static_cast<const IOdictionary&>(dict);

        const fvMesh& mesh = refCast<const fvMesh>(d.db());

        scalarField myInteriorField(mesh.nCells(), 0.0);

        const volVectorField& C = mesh.C();



        forAll(myInteriorField, cellI)

        {

            const scalar x = C[cellI].x();

            const scalar y = C[cellI].y();

            const scalar tmp0 = (5.0/4.0)*M_PI; 

            myInteriorField[cellI] = 1000.0*Foam::sin(tmp0*y) + 1000.0*Foam::cos(tmp0*x) + 100000.0; 

        }



        os.writeEntry("", myInteriorField);

    #};

};

Inside the code I am computing the difference between a dummy field and the pressure field. The dummy field has the same value as the pressureField.

Code:



const volVectorField& C = mesh.C(); 



volScalarField dummyField

(

    IOobject

    (

        "dummyField",

        runTime.timeName(),

        mesh,

        IOobject::NO_READ,

        IOobject::AUTO_WRITE

    ),

    mesh,

    dimensionedScalar("dummyScalar", dimless, 0.0)

);



forAll(dummyField,  cellI)          

    {                               

        const scalar x = C[cellI].x(); 

        const scalar y = C[cellI].y(); 

                             

        const scalar tmp0 = (5.0/4.0)*M_PI; 

        const scalar solution = 1000.0*Foam::sin(tmp0*y) + 1000.0*Foam::cos(tmp0*x) + 100000.0; 

                                                                 

        dummyField[cellI] = mag(solution - p[cellI]); 

    }

And finally computing the L2 error norm as:

Code:

const scalarField& V = mesh.V();  



Info << "Error is "    << Foam::sqrt( gSum(dummyField*dummyField*V)/gSum(V) )    << endl;

One would expect for the norm to be zero.. but the value differs depending on the value of writePrecision... The default value is 6, so if I proceed I will get a L2 value of 0.22. I only get a value of 0 with 17 decimal cases. I was under the impression that writePrecision would control the amount of decimal cases for output information of the field. But apparently it also has an effect on the input given.