[XJ_Wang]

Hi foamers:

I developed an inlet boundary condition for velocity, but can not run it in parallel.

In 0/U, I defined my inlet as:

inlet
    {
         type            DSRFGVelocity;
         M              600;
         N              100;
        uref            22.5091;
        alpha           0.1538;
        zref            1.3860;
        inflowPar       pkqomega;
        value           uniform (0 0 0);

    }

Then with a good mesh, I run:

decomposePar
mpirun -np xx renumberMesh -overwrite -parallel

It shows the error message:

[1] --> FOAM FATAL IO ERROR:
[1] compound has already been transferred from token
    on line 33 the empty compound of type List<vector>
[1]
[1] file: /ddnA/work/xxxx/DSRFGinflow/processor1/0/U.boundaryField.inlet.value at line 33.
[1]
[1]     From function Foam::token::compound& Foam::token::transferCompoundToken(const Foam::Istream&)
[1]     in file lnInclude/token.C at line 99.
[1]
FOAM parallel run aborting

Then I go to the file in processor1/0/U,

It is:

 inlet
    {
        type            DSRFGVelocity;
        M               600;
        N               100;
        zref            1.386;
        alpha           0.1538;
        uref            22.5091;
        inflowPar       pkqomega;
        value           nonuniform List<vector>   // this is line 33
1086
(
(29.289282 3.8500723 -1.4503862)
(29.288161 3.8519444 -1.450854)
(29.287024 3.8538043 -1.4513087)

My updateCoeffs() function in my source code looks like:

void Foam::DSRFGVelocityFvPatchVectorField::updateCoeffs()
{
    if (this->updated())
    {
        return;
    }
    if (curTimeIndex_ != this->db().time().timeIndex())
    {


    xxxxxxxxxx some calculation procedure

    this->operator==(Utotal);
    curTimeIndex_ = this->db().time().timeIndex();
    }
    fixedValueFvPatchVectorField::updateCoeffs();
}

Can anyone help me with this?

[serles]

Something is wrong with the I/O. During reading of fields from the disk, OpenFOAM constructs a dictionary from the file (processor1/0/U in your case), where the Lists are stored as compoundToken (since dictionary is able to work only with tokens). In the next step, this dictionary is supplied to the GeometricField constructor, whereby the List is transferred (not read from disk because it's already in memory) to the GeometricField instance.

I assume you didn't change the source code of renumberMesh. So the problem should be somewhere within the reading operations of your boundary condition. Can you provide the related members like "read" or "operator>>" from you BC?

[XJ_Wang]

Hi Sergey ,

Thanks for your reply. To be honest, I did not see any "read" members in my source code. Below is my updateCoeffs() member function:

void Foam::DSRFGVelocityFvPatchVectorField::updateCoeffs()
{
    if (this->updated())
    {
        return;
    }
    if (curTimeIndex_ != this->db().time().timeIndex())
{
    scalar time = this->db().time().value();
	vectorField& U = *this;
    vectorField Uavg(patch().size(),vector::zero);
    vectorField Uprim(patch().size(),vector::zero);
    //vectorField Utotal(patch().size(),vector::zero);
    scalar kx=0.;
        scalar vx=0.;
        scalar vy = 0.;
        scalar vz = 0.;
    forAll(patch().Cf(),faceI)
    {
        scalar zcell=patch().Cf()[faceI][2];
        Uavg[faceI][0]=uref_*pow(zcell/zref_,alpha_);
    }
    


    forAll(patch().Cf(),faceI)
    {
        kx=0.;
        vx=0.;
        vy=0.;
        vz=0.;

        
        forAll(allK_,ki)
        {
            kx=patch().Cf()[faceI][0]*allK_[ki][0]/Ls_[ki][0]+patch().Cf()[faceI][1]*allK_[ki][1]/Ls_[ki][1]+patch().Cf()[faceI][2]*allK_[ki][2]/Ls_[ki][2]+omega_[ki]*time;
            vx+=allP_[ki][0]*sin(kx)+allQ_[ki][0]*cos(kx);
            vy+=allP_[ki][1]*sin(kx)+allQ_[ki][1]*cos(kx);
            vz+=allP_[ki][2]*sin(kx)+allQ_[ki][2]*cos(kx);
        }

        Uprim[faceI][0]=vx;
        Uprim[faceI][1]=vy;
        Uprim[faceI][2]=vz;
            



    }
U=Uavg+Uprim;

curTimeIndex_ = this->db().time().timeIndex();
}
    fixedValueFvPatchVectorField::updateCoeffs();
}

I am also not sure of the autoMap and rmap member function. I just claimed the function but don't know how to use them. My code can run in single processor.

void Foam::DSRFGVelocityFvPatchVectorField::autoMap
(
    const fvPatchFieldMapper& m
)
{
    fixedValueFvPatchVectorField::autoMap(m);
   
    
}


void Foam::DSRFGVelocityFvPatchVectorField::rmap
(
    const fvPatchField& ptf,
    const labelList& addr
)
{
    fixedValueFvPatchVectorField::rmap(ptf, addr);

    const DSRFGVelocityFvPatchVectorField& tiptf =
        refCast<const DSRFGVelocityFvPatchVectorField>(ptf);
      
}

Thank you very much!!!

Originally Posted by serles

Something is wrong with the I/O. During reading of fields from the disk, OpenFOAM constructs a dictionary from the file (processor1/0/U in your case), where the Lists are stored as compoundToken (since dictionary is able to work only with tokens). In the next step, this dictionary is supplied to the GeometricField constructor, whereby the List is transferred (not read from disk because it's already in memory) to the GeometricField instance.

I assume you didn't change the source code of renumberMesh. So the problem should be somewhere within the reading operations of your boundary condition. Can you provide the related members like "read" or "operator>>" from you BC?

[serles]

To be honest, I did not see any "read" members in my source code.

OK, then I assume you construct your BC with a dictionary, don't you? Anyway, can you provide the constructor you use?


What class do you inherit from?

[XJ_Wang]

Thank you very much Sergey!!!

You are correct! There is something wrong with my constructor. My BC inherits from fixedValue class and I used 'foamNewBC -f -v' to generate the templates of my BC dictionary. I am using Openfoam V6, but I modified my code according to a code from Openfoam v2112 so I changed my constructor. There is nothing wrong when I compiled my library but it just can not work well. Now I changed my constructor and it works!


A personal experience: I did not find too many similar problems in this forum, which means my problem is unusual and I might make something wrong at the first step.

[serles]

You're welcome