Belos solver manager BLOCK GMRES with standardized name BLOCK GMRES has not been registered.

[zhengshuang2005]

Dear Developers,

I want to ask a question related to Belos solver factory.

I used the Belos solver package and followed the Belos solver package tutorial as shown on the Doxygen page "https://trilinos.org/docs/r12.12/packages/belos/doc/html/classBelos_1_1SolverFactory.html". It used to work with the old Trilinos version I was using. However, recently I upgraded my Trilinos version to the latest one (12.12) and when I ran the solver again, it gives the following error:

terminate called after throwing an instance of 'std::invalid_argument'
  what():  /home/shuang/foam/foam-extend-3.2/ThirdParty/packages/trilinos/include/BelosSolverFactory.hpp:375:

Throw number = 1

Throw test that evaluated to true: it == get_solverManagers().end()

Belos solver manager BLOCK GMRES with standardized name BLOCK GMRES has not been registered."

I don't know what has happened and could you please point out where I did wrong? My function using the Belis solver is as follows. I have also attached my Trilinos cmake script for your reference.

void TpetraSolver::solve()
{
    // A matrix fill complete
    A_->fillComplete();

    // B vector fill complete
    local_ordinal_type nodeNumElements = map_->getNodeNumElements();
    for (local_ordinal_type i = 0; i < nodeNumElements; i++)
    {
        B_->get1dViewNonConst()[i] = b_[elementList_[i]];
    }

    // Preconditioner
    RCP<op_type> M = createPreconditioner(A_);

    // Solver
    RCP<ParameterList> solverParams = parameterList();
    solverParams->set("Num Blocks", 40);
    solverParams->set("Maximum Iterations", 1000);
    solverParams->set("Convergence Tolerance", 1.0e-6);
    Belos::SolverFactory<scalar_type, vec_type, op_type> factory;
    RCP<Belos::SolverManager<scalar_type, vec_type, op_type> > solver =
        factory.create("Block GMRES", solverParams);

    // Linaer problem
    RCP<LinearProblem> problem = rcp(new LinearProblem(A_, X_, B_));

    // Solve
    problem->setRightPrec(M);
    problem->setProblem();
    solver->setProblem(problem);
    Belos::ReturnType result = solver->solve();
    const int numIters = solver->getNumIters();

    // Redistribute
    redistributeResult();

    if (result == Belos::Converged)
    {
        cout << "The Belos solve took " << numIters
            << " iteration(s) to reach a relative residual tolerance of "
            << 1.0e-6 << "."
            << endl;
    }
    else
    {
        cout << "The Belos solve took " << numIters
            << " iteration(s), but did not reach a relative "
               "residual tolerance of " << 1.0e-6 << "."
            << endl;
    }
}

~/foam/foam-extend-3.2/ThirdParty/rpmBuild/BUILD/cmake-3.11.0-Linux-x86_64/bin/cmake -D CMAKE_INSTALL_PREFIX:PATH="~/foam/foam-extend-3.2/ThirdParty/packages/trilinos" \
-D TPL_ENABLE_MPI:BOOL=ON \
-D MPI_BASE_DIR:PATH="~/foam/foam-extend-3.2/ThirdParty/packages/openmpi-1.6.5/platforms/linux64GccDPOpt" \
-D CMAKE_CXX_FLAGS:STRING="-O2 -std=c++11 -pedantic -ftrapv -Wall -Wno-long-long" \
-D CMAKE_BUILD_TYPE:STRING=RELEASE \
-D Trilinos_ENABLE_Fortran:BOOL=OFF \
-D BUILD_SHARED_LIBS:BOOL=ON \
-D Trilinos_WARNINGS_AS_ERRORS_FLAGS:STRING="" \
-D Trilinos_ENABLE_ALL_PACKAGES:BOOL=OFF \
-D Trilinos_ENABLE_Teuchos:BOOL=ON \
-D Trilinos_ENABLE_Epetra:BOOL=ON \
-D Trilinos_ENABLE_EpetraExt:BOOL=ON \
-D Trilinos_ENABLE_Ifpack:BOOL=ON \
-D Trilinos_ENABLE_Ifpack2:BOOL=ON \
-D Trilinos_ENABLE_AztecOO:BOOL=ON \
-D Trilinos_ENABLE_Amesos:BOOL=ON \
-D Trilinos_ENABLE_MueLu:BOOL=OFF \
-D Trilinos_ENABLE_Anasazi:BOOL=ON \
-D Trilinos_ENABLE_Belos:BOOL=ON \
-D Trilinos_ENABLE_Tpetra:BOOL=ON \
-D Trilinos_ENABLE_OpenMP:BOOL=OFF \
-D Trilinos_ENABLE_Xpetra:BOOL=OFF \
-D Trilinos_ENABLE_Isorropia:BOOL=ON \
-D Trilinos_ENABLE_TESTS:BOOL=ON \
-D TPL_ENABLE_BLAS:BOOL=ON \
-D TPL_BLAS_LIBRARIES:PATH="~/foam/foam-extend-3.2/ThirdParty/rpmBuild/BUILD/BLAS-3.8.0/libblas.so" \
-D TPL_ENABLE_LAPACK:BOOL=ON \
-D LAPACK_LIBRARY_DIRS:PATH="~/foam/foam-extend-3.2/ThirdParty/packages/LAPACK/lib" \
-D TPL_ENABLE_Boost:BOOL=ON \
-D Boost_INCLUDE_DIRS:PATH="~/foam/foam-extend-3.2/ThirdParty/rpmBuild/BUILD/boost_1_66_0" \
-D Boost_LIBRARY_DIRS:PATH="~/foam/foam-extend-3.2/ThirdParty/rpmBuild/BUILD/boost_1_66_0/stage/lib" \
-D CMAKE_VERBOSE_MAKEFILE:BOOL=OFF \
-D Trilinos_VERBOSE_CONFIGURE:BOOL=OFF \
-D Amesos_ENABLE_SuperLUDist:BOOL=OFF \
-D TPL_ENABLE_ParMETIS:BOOL=ON \
-D ParMETIS_INCLUDE_DIRS:PATH="~/foam/foam-extend-3.2/ThirdParty/packages/parmetis-4.0.3/platforms/linux64GccDPOpt/include" \
-D ParMETIS_LIBRARY_DIRS:PATH="~/foam/foam-extend-3.2/ThirdParty/packages/parmetis-4.0.3/platforms/linux64GccDPOpt/lib" \
-D TPL_ENABLE_SuperLUDist:BOOL=OFF \
..

Thank you and looking forward to your reply!

Best regards, Shuang

[cgcgcg]

@trilinos/belos

[zhengshuang2005]

Problem resolved. It happened because long as global indices Is disabled in the Trilinos compilation.

[mhoemmen]

@zhengshuang2005 You should only ever use global index types that are enabled at configure time (e.g., with a CMake option).

[abodado]

Where is this "https://trilinos.org/docs/r12.12/packages/belos/doc/html/classBelos_1_1SolverFactory.html". file that @zhengshuang2005 mentioned? It looks like a good example, but I don't see the full file anywhere. Looks like that link is no longer valid.

[cgcgcg]

Check here: https://docs.trilinos.org/dev/packages/belos/doc/html/classBelos_1_1BelosSolverFactory.html

[abodado]

@cgcgcg, thank you. However, I do not see a "tutorial" on that page as was referenced by @zhengshuang2005. He stated "I used the Belos solver package and followed the Belos solver package tutorial as shown on the Doxygen page". Perhaps he made his own code based on the documentation.

[cgcgcg]

As far as I know there is no such tutorial, but lemme tag @hkthorn.

[hkthorn]

@abodado There are not adequate examples for using the Belos linear solver factory at this time. The examples we have only illustrate constructing a specific solver object directly. That is an omission I'll look into correcting ASAP.

[abodado]

Thank you @cgcgcg and @hkthorn for looking into this.