Source term for additional scalars

[bnkeeton]

Hi all, I'm extending the lowMach example to add sources to additional scalars. To do so, I changed the relevant section in the udf to:

void userq(nrs_t *nrs, dfloat time, occa::memory o_S, occa::memory o_FS)
{
  cds_t *cds = nrs->cds;
  mesh_t *mesh = cds->mesh[0];
  fillQ(mesh->Nelements, nrs->cds->fieldOffset[0], o_FS, o_S);
}

The corresponding oudf file becomes:

@kernel void fillQ(const dlong Nelements, const dlong sOffset, @ restrict dfloat *QVOL, @ restrict const dfloat *TEMP)
{
  for (dlong e = 0; e < Nelements; ++e; @outer(0)) {
    for (int n = 0; n < p_Np; ++n; @inner(0)) {
      const int id = e * p_Np + n;
      const dfloat temp1  = TEMP[id];
      const dfloat zz     = TEMP[id + 1 * sOffset];
      QVOL[id + 0 * sOffset] = temp1*zz;
      QVOL[id + 1 * sOffset] = 1 / temp1;
    }
  }
}

To check results, I output the sources in o_FS for temperature and scalar01 by adding the following lines to UDF_ExecuteStep:

void UDF_ExecuteStep(nrs_t *nrs, dfloat time, int tstep)
{
  cds_t *cds = nrs->cds;
  if (nrs->isOutputStep) {
    nek::ocopyToNek(time, tstep);
    nek::userchk();
    const auto fieldOffsetBytes = nrs->fieldOffset * sizeof(dfloat);
    auto o_field1 = nrs->cds->o_FS.slice(0 * fieldOffsetBytes, fieldOffsetBytes);
    auto o_field2 = nrs->cds->o_FS.slice(1 * fieldOffsetBytes, fieldOffsetBytes);
    bool coords = true;
    bool FP64 = true;
    int Nscalar = nrs->Nscalar;
    occa::memory o_null;
    writeFld("fs1", time, tstep, coords, FP64, &o_null, &o_field1, &o_null, Nscalar);
    writeFld("fs2", time, tstep, coords, FP64, &o_null, &o_field2, &o_null, Nscalar);
  }
}

I have the corresponding nek5000 simulation as a check, and can confirm that this method does not add the correct source terms to the scalars. Does anyone notice anything wrong with the approach? I just switched over to nekrs so perhaps I have missed something.

Thanks for the help,
Ben Keeton

[stgeke]

What you see as cds->o_FS in UDF_ExecuteStep will not contain just the source term. Additional things like the convection term will be added to cds->o_FS. Moreover it's weighted by Jw.

[bnkeeton]

Thanks, Stefan. I output the source term to a new variable with copyFrom and got the correct result.

Ben