Dsv/swi diffusion

[dasvyat]

Dirichlet BC for transport diffusion. The default option currently is not enforcing Dirchlet BC. To enforce dirichlet BC it is required to add parameter in diffusion operator parameter list:

  <ParameterList name="diffusion">
    <Parameter name="discretization primary" type="string" value="fv: default"/>
    <ParameterList name="inverse">
      <Parameter name="preconditioning method" type="string" value="boomer amg"/>
      <Parameter name="iterative method" type="string" value="pcg"/>
      <ParameterList name="boomer amg parameters">
        <!-- <Parameter name="verbosity" type="int" value="0" /> -->
        <Parameter name="tolerance" type="double" value="0.0"/>
        <Parameter name="smoother sweeps" type="int" value="3"/>
        <Parameter name="cycle applications" type="int" value="5"/>
        <Parameter name="strong threshold" type="double" value="0.5"/>
      </ParameterList>
      <ParameterList name="pcg parameters">
        <Parameter name="maximum number of iterations" type="int" value="400"/>
        <Parameter name="error tolerance" type="double" value="1e-20"/>
        <Parameter name="convergence criteria" type="Array(string)" value="{relative rhs, make one iteration}"/>
        <Parameter name="size of Krylov space" type="int" value="5"/>
      </ParameterList>
    </ParameterList>
    <Parameter name="enforce boundary conditions" type="bool" value="true"/>
  </ParameterList>

[saubhagya-gatech]

Hi @dasvyat ,

Thanks for this PR. There is one more issue with salinity EOS. When you add additional species, like some tracer (does not matter whether you add it through alquimia or directly through ATS), we get an error

libc++abi: terminating due to uncaught exception of type Errors::Message: Requested components are not consistent with previous request.

This is because the EOS is expecting single scalar and not a composite vector? In the salinity EOS, if we have more than one species, we should specify, which DOF is salt.

if it helps, I add some print statements for debug in eos_evaluator.cc as and ran a simulation with salt and a tracer:

` if (!UnionAndConsistent_(names,
locations,
num_dofs,
mastermaps,
ghostmaps,
names_,
locations_,
num_dofs_,
mastermaps_,
ghostmaps_)) {

    // ============================================================
    // DEBUG OUTPUT
    // ============================================================
    std::cout << "\n[DEBUG] CompositeVectorSpace inconsistency detected!" << std::endl;

    std::cout << "  Existing components: ";
    for (auto& n : names_) std::cout << n << " ";
    std::cout << std::endl;

    std::cout << "  Requested components: ";
    for (auto& n : names) std::cout << n << " ";
    std::cout << std::endl;

    // entity kinds
    std::cout << "  Existing entity kinds: ";
    for (auto& ek : locations_) std::cout << ek << " ";
    std::cout << std::endl;

    std::cout << "  Requested entity kinds: ";
    for (auto& ek : locations) std::cout << ek << " ";
    std::cout << std::endl;

    // DOFs
    std::cout << "  Existing DOFs: ";
    for (auto& d : num_dofs_) std::cout << d << " ";
    std::cout << std::endl;

    std::cout << "  Requested DOFs: ";
    for (auto& d : num_dofs) std::cout << d << " ";
    std::cout << std::endl;

    // map addresses (to see if maps differ)
    std::cout << "  Existing master map ptrs: ";
    for (auto& p : mastermaps_) std::cout << p.second.get() << " ";
    std::cout << std::endl;

    std::cout << "  Requested master map ptrs: ";
    for (auto& p : mastermaps) std::cout << p.second.get() << " ";
    std::cout << std::endl;

    std::cout << "--------------------------------------------------------------" << std::endl;
    // ============================================================

Errors::Message message("Requested components are not consistent with previous request.");
Exceptions::amanzi_throw(message);
}
}

InitIndexMap_();

return this;
};`

and got this

[DEBUG] CompositeVectorSpace inconsistency detected! Existing components: cell Requested components: cell Existing entity kinds: 4 Requested entity kinds: 4 Existing DOFs: 2 Requested DOFs: 1 Existing master map ptrs: 0x6000093922e0 Requested master map ptrs: 0x6000093922e0

[dasvyat]

Hi @saubhagya-gatech ,

thank you for testing this PR. I'll take a look at this issue shortly.

D

[ecoon]

@dasvyat I'm not sure where we left this today -- is this still WIP then? Can you change the title of the PR to WIP until you are ready with the fixes to ensure compatibility?

[ecoon]

@dasvyat

#20 554.3 /home/amanzi_user/amanzi/src/physics/ats/src/constitutive_relations/eos/eos_evaluator.cc:360:13: error: 'class Amanzi::CompositeVectorSpace' has no member named 'UpdateSameNumDofs'
#20 554.3 360 | dep_fac.UpdateSameNumDofs(fac);
#20 554.3 | ^~~~~~~~~~~~~~~~~
#20 555.0 make[2]: ***

Is there a branch on Amanzi as well? Or do you need to push something on Amanzi? Make sure the branch names match if you need CI to pick up Amanzi's branch.

[dasvyat]

@ecoon There is corresponding branch in amanzi dsv/swi_diffusion. I've double checked that UpdateSameNumDofs exists there, but CI takes master and this is causing this error.

[ecoon]

Is this needed? I don't see if this is used in the web API, but suspect it is old/dead/debugging code.