mirror of
https://github.com/OPM/opm-simulators.git
synced 2024-12-23 07:53:29 -06:00
Allow well operators with FlexibleSolver.
This commit is contained in:
parent
6d644da88e
commit
c94eec872f
@ -43,54 +43,56 @@ public:
|
||||
using MatrixType = MatrixTypeT;
|
||||
using VectorType = VectorTypeT;
|
||||
|
||||
/// Base class type of the operator passed to the solver.
|
||||
using AbstractOperatorType = Dune::AssembledLinearOperator<MatrixType, VectorType, VectorType>;
|
||||
/// Base class type of the contained preconditioner.
|
||||
using AbstractPrecondType = Dune::PreconditionerWithUpdate<VectorType, VectorType>;
|
||||
|
||||
/// Create a sequential solver.
|
||||
FlexibleSolver(const MatrixType& matrix,
|
||||
FlexibleSolver(AbstractOperatorType& op,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorTypeT()>& weightsCalculator = std::function<VectorTypeT()>());
|
||||
const std::function<VectorType()>& weightsCalculator = std::function<VectorType()>());
|
||||
|
||||
/// Create a parallel solver (if Comm is e.g. OwnerOverlapCommunication).
|
||||
template <class Comm>
|
||||
FlexibleSolver(const MatrixType& matrix,
|
||||
FlexibleSolver(AbstractOperatorType& op,
|
||||
const Comm& comm,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorTypeT()>& weightsCalculator = std::function<VectorTypeT()>());
|
||||
const std::function<VectorType()>& weightsCalculator = std::function<VectorType()>());
|
||||
|
||||
virtual void apply(VectorType& x, VectorType& rhs, Dune::InverseOperatorResult& res) override;
|
||||
|
||||
virtual void apply(VectorType& x, VectorType& rhs, double reduction, Dune::InverseOperatorResult& res) override;
|
||||
|
||||
/// Type of the contained preconditioner.
|
||||
using AbstractPrecondType = Dune::PreconditionerWithUpdate<VectorType, VectorType>;
|
||||
|
||||
/// Access the contained preconditioner.
|
||||
AbstractPrecondType& preconditioner();
|
||||
|
||||
virtual Dune::SolverCategory::Category category() const override;
|
||||
|
||||
private:
|
||||
using AbstractOperatorType = Dune::AssembledLinearOperator<MatrixType, VectorType, VectorType>;
|
||||
using AbstractScalarProductType = Dune::ScalarProduct<VectorType>;
|
||||
using AbstractSolverType = Dune::InverseOperator<VectorType, VectorType>;
|
||||
|
||||
// Machinery for making sequential or parallel operators/preconditioners/scalar products.
|
||||
template <class Comm>
|
||||
void initOpPrecSp(const MatrixType& matrix, const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorTypeT()> weightsCalculator, const Comm& comm);
|
||||
void initOpPrecSp(AbstractOperatorType& op, const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorType()> weightsCalculator, const Comm& comm);
|
||||
|
||||
void initOpPrecSp(const MatrixType& matrix, const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorTypeT()> weightsCalculator, const Dune::Amg::SequentialInformation&);
|
||||
void initOpPrecSp(AbstractOperatorType& op, const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorType()> weightsCalculator, const Dune::Amg::SequentialInformation&);
|
||||
|
||||
void initSolver(const boost::property_tree::ptree& prm, bool isMaster);
|
||||
void initSolver(const boost::property_tree::ptree& prm, const bool is_iorank);
|
||||
|
||||
// Main initialization routine.
|
||||
// Call with Comm == Dune::Amg::SequentialInformation to get a serial solver.
|
||||
template <class Comm>
|
||||
void init(const MatrixType& matrix,
|
||||
void init(AbstractOperatorType& op,
|
||||
const Comm& comm,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorTypeT()> weightsCalculator);
|
||||
const std::function<VectorType()> weightsCalculator);
|
||||
|
||||
std::shared_ptr<AbstractOperatorType> linearoperator_;
|
||||
AbstractOperatorType* linearoperator_for_solver_;
|
||||
std::shared_ptr<AbstractOperatorType> linearoperator_for_precond_;
|
||||
std::shared_ptr<AbstractPrecondType> preconditioner_;
|
||||
std::shared_ptr<AbstractScalarProductType> scalarproduct_;
|
||||
std::shared_ptr<AbstractSolverType> linsolver_;
|
||||
|
@ -39,23 +39,23 @@ namespace Dune
|
||||
/// Create a sequential solver.
|
||||
template <class MatrixType, class VectorType>
|
||||
FlexibleSolver<MatrixType, VectorType>::
|
||||
FlexibleSolver(const MatrixType& matrix,
|
||||
FlexibleSolver(AbstractOperatorType& op,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorType()>& weightsCalculator)
|
||||
{
|
||||
init(matrix, Dune::Amg::SequentialInformation(), prm, weightsCalculator);
|
||||
init(op, Dune::Amg::SequentialInformation(), prm, weightsCalculator);
|
||||
}
|
||||
|
||||
/// Create a parallel solver (if Comm is e.g. OwnerOverlapCommunication).
|
||||
template <class MatrixType, class VectorType>
|
||||
template <class Comm>
|
||||
FlexibleSolver<MatrixType, VectorType>::
|
||||
FlexibleSolver(const MatrixType& matrix,
|
||||
FlexibleSolver(AbstractOperatorType& op,
|
||||
const Comm& comm,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorType()>& weightsCalculator)
|
||||
{
|
||||
init(matrix, comm, prm, weightsCalculator);
|
||||
init(op, comm, prm, weightsCalculator);
|
||||
}
|
||||
|
||||
template <class MatrixType, class VectorType>
|
||||
@ -88,7 +88,7 @@ namespace Dune
|
||||
FlexibleSolver<MatrixType, VectorType>::
|
||||
category() const
|
||||
{
|
||||
return linearoperator_->category();
|
||||
return linearoperator_for_solver_->category();
|
||||
}
|
||||
|
||||
// Machinery for making sequential or parallel operators/preconditioners/scalar products.
|
||||
@ -96,56 +96,64 @@ namespace Dune
|
||||
template <class Comm>
|
||||
void
|
||||
FlexibleSolver<MatrixType, VectorType>::
|
||||
initOpPrecSp(const MatrixType& matrix, const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorType()> weightsCalculator, const Comm& comm)
|
||||
initOpPrecSp(AbstractOperatorType& op,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorType()> weightsCalculator,
|
||||
const Comm& comm)
|
||||
{
|
||||
// Parallel case.
|
||||
using ParOperatorType = Dune::OverlappingSchwarzOperator<MatrixType, VectorType, VectorType, Comm>;
|
||||
using pt = const boost::property_tree::ptree;
|
||||
auto linop = std::make_shared<ParOperatorType>(matrix, comm);
|
||||
linearoperator_ = linop;
|
||||
using ParOperatorType = Dune::OverlappingSchwarzOperator<MatrixType, VectorType, VectorType, Comm>;
|
||||
linearoperator_for_solver_ = &op;
|
||||
auto op_prec = std::make_shared<ParOperatorType>(op.getmat(), comm);
|
||||
auto child = prm.get_child_optional("preconditioner");
|
||||
preconditioner_
|
||||
= Opm::PreconditionerFactory<ParOperatorType, Comm>::create(*linop, child? *child : pt(),
|
||||
weightsCalculator, comm);
|
||||
scalarproduct_ = Dune::createScalarProduct<VectorType, Comm>(comm, linearoperator_->category());
|
||||
preconditioner_ = Opm::PreconditionerFactory<ParOperatorType, Comm>::create(*op_prec,
|
||||
child ? *child : pt(),
|
||||
weightsCalculator,
|
||||
comm);
|
||||
scalarproduct_ = Dune::createScalarProduct<VectorType, Comm>(comm, op.category());
|
||||
linearoperator_for_precond_ = op_prec;
|
||||
}
|
||||
|
||||
template <class MatrixType, class VectorType>
|
||||
void
|
||||
FlexibleSolver<MatrixType, VectorType>::
|
||||
initOpPrecSp(const MatrixType& matrix, const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorType()> weightsCalculator, const Dune::Amg::SequentialInformation&)
|
||||
initOpPrecSp(AbstractOperatorType& op,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorType()> weightsCalculator,
|
||||
const Dune::Amg::SequentialInformation&)
|
||||
{
|
||||
// Sequential case.
|
||||
using SeqOperatorType = Dune::MatrixAdapter<MatrixType, VectorType, VectorType>;
|
||||
using pt = const boost::property_tree::ptree;
|
||||
auto linop = std::make_shared<SeqOperatorType>(matrix);
|
||||
linearoperator_ = linop;
|
||||
using SeqOperatorType = Dune::MatrixAdapter<MatrixType, VectorType, VectorType>;
|
||||
linearoperator_for_solver_ = &op;
|
||||
auto op_prec = std::make_shared<SeqOperatorType>(op.getmat());
|
||||
auto child = prm.get_child_optional("preconditioner");
|
||||
preconditioner_ = Opm::PreconditionerFactory<SeqOperatorType>::create(*linop, child? *child : pt(),
|
||||
preconditioner_ = Opm::PreconditionerFactory<SeqOperatorType>::create(*op_prec,
|
||||
child ? *child : pt(),
|
||||
weightsCalculator);
|
||||
scalarproduct_ = std::make_shared<Dune::SeqScalarProduct<VectorType>>();
|
||||
linearoperator_for_precond_ = op_prec;
|
||||
}
|
||||
|
||||
template <class MatrixType, class VectorType>
|
||||
void
|
||||
FlexibleSolver<MatrixType, VectorType>::
|
||||
initSolver(const boost::property_tree::ptree& prm, bool isMaster)
|
||||
initSolver(const boost::property_tree::ptree& prm, const bool is_iorank)
|
||||
{
|
||||
const double tol = prm.get<double>("tol", 1e-2);
|
||||
const int maxiter = prm.get<int>("maxiter", 200);
|
||||
const int verbosity = isMaster? prm.get<int>("verbosity", 0) : 0;
|
||||
const int verbosity = is_iorank ? prm.get<int>("verbosity", 0) : 0;
|
||||
const std::string solver_type = prm.get<std::string>("solver", "bicgstab");
|
||||
if (solver_type == "bicgstab") {
|
||||
linsolver_.reset(new Dune::BiCGSTABSolver<VectorType>(*linearoperator_,
|
||||
linsolver_.reset(new Dune::BiCGSTABSolver<VectorType>(*linearoperator_for_solver_,
|
||||
*scalarproduct_,
|
||||
*preconditioner_,
|
||||
tol, // desired residual reduction factor
|
||||
maxiter, // maximum number of iterations
|
||||
verbosity));
|
||||
} else if (solver_type == "loopsolver") {
|
||||
linsolver_.reset(new Dune::LoopSolver<VectorType>(*linearoperator_,
|
||||
linsolver_.reset(new Dune::LoopSolver<VectorType>(*linearoperator_for_solver_,
|
||||
*scalarproduct_,
|
||||
*preconditioner_,
|
||||
tol, // desired residual reduction factor
|
||||
@ -153,7 +161,7 @@ namespace Dune
|
||||
verbosity));
|
||||
} else if (solver_type == "gmres") {
|
||||
int restart = prm.get<int>("restart", 15);
|
||||
linsolver_.reset(new Dune::RestartedGMResSolver<VectorType>(*linearoperator_,
|
||||
linsolver_.reset(new Dune::RestartedGMResSolver<VectorType>(*linearoperator_for_solver_,
|
||||
*scalarproduct_,
|
||||
*preconditioner_,
|
||||
tol,
|
||||
@ -163,7 +171,7 @@ namespace Dune
|
||||
#if HAVE_SUITESPARSE_UMFPACK
|
||||
} else if (solver_type == "umfpack") {
|
||||
bool dummy = false;
|
||||
linsolver_.reset(new Dune::UMFPack<MatrixType>(linearoperator_->getmat(), verbosity, dummy));
|
||||
linsolver_.reset(new Dune::UMFPack<MatrixType>(linearoperator_for_solver_->getmat(), verbosity, dummy));
|
||||
#endif
|
||||
} else {
|
||||
OPM_THROW(std::invalid_argument, "Properties: Solver " << solver_type << " not known.");
|
||||
@ -177,13 +185,13 @@ namespace Dune
|
||||
template <class Comm>
|
||||
void
|
||||
FlexibleSolver<MatrixType, VectorType>::
|
||||
init(const MatrixType& matrix,
|
||||
init(AbstractOperatorType& op,
|
||||
const Comm& comm,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const std::function<VectorType()> weightsCalculator)
|
||||
{
|
||||
initOpPrecSp(matrix, prm, weightsCalculator, comm);
|
||||
initSolver(prm, comm.communicator().rank()==0);
|
||||
initOpPrecSp(op, prm, weightsCalculator, comm);
|
||||
initSolver(prm, comm.communicator().rank() == 0);
|
||||
}
|
||||
|
||||
} // namespace Dune
|
||||
@ -198,28 +206,33 @@ using BM = Dune::BCRSMatrix<Dune::FieldMatrix<double, N, N>>;
|
||||
template <int N>
|
||||
using OBM = Dune::BCRSMatrix<Opm::MatrixBlock<double, N, N>>;
|
||||
|
||||
// INSTANTIATE_CONSTRUCTOR instantiates the constructor that is a template,
|
||||
// this is only needed in the MPI case, since otherwise the Comm type is
|
||||
// not a template argument but always SequentialInformation.
|
||||
#if HAVE_MPI
|
||||
|
||||
using Comm = Dune::OwnerOverlapCopyCommunication<int, int>;
|
||||
#define INSTANTIATE_FLEXIBLESOLVER_CONSTRUCTOR(n) \
|
||||
template Dune::FlexibleSolver<OBM<n>, BV<n>>::FlexibleSolver(const MatrixType& matrix, \
|
||||
|
||||
// Note: we must instantiate the constructor that is a template.
|
||||
// This is only needed in the parallel case, since otherwise the Comm type is
|
||||
// not a template argument but always SequentialInformation.
|
||||
|
||||
#define INSTANTIATE_FLEXIBLESOLVER(N) \
|
||||
template class Dune::FlexibleSolver<BM<N>, BV<N>>; \
|
||||
template class Dune::FlexibleSolver<OBM<N>, BV<N>>; \
|
||||
template Dune::FlexibleSolver<BM<N>, BV<N>>::FlexibleSolver(AbstractOperatorType& op, \
|
||||
const Comm& comm, \
|
||||
const boost::property_tree::ptree& prm, \
|
||||
const std::function<BV<N>()>& weightsCalculator); \
|
||||
template Dune::FlexibleSolver<OBM<N>, BV<N>>::FlexibleSolver(AbstractOperatorType& op, \
|
||||
const Comm& comm, \
|
||||
const boost::property_tree::ptree& prm, \
|
||||
const std::function<BV<n>()>& weightsCalculator);
|
||||
#else
|
||||
#define INSTANTIATE_FLEXIBLESOLVER_CONSTRUCTOR(n)
|
||||
#endif
|
||||
const std::function<BV<N>()>& weightsCalculator);
|
||||
|
||||
// INSTANTIATE instantiates the class including any templated constructors if necessary.
|
||||
#define INSTANTIATE_FLEXIBLESOLVER(n) \
|
||||
/* Variants using Dune::FieldMatrix blocks. */ \
|
||||
template class Dune::FlexibleSolver<BM<n>, BV<n>>; \
|
||||
/* Variants using Opm::MatrixBlock blocks. */ \
|
||||
template class Dune::FlexibleSolver<OBM<n>, BV<n>>; \
|
||||
INSTANTIATE_FLEXIBLESOLVER_CONSTRUCTOR(n)
|
||||
#else // HAVE_MPI
|
||||
|
||||
#define INSTANTIATE_FLEXIBLESOLVER(N) \
|
||||
template class Dune::FlexibleSolver<BM<N>, BV<N>>; \
|
||||
template class Dune::FlexibleSolver<OBM<N>, BV<N>>;
|
||||
|
||||
#endif // HAVE_MPI
|
||||
|
||||
|
||||
#endif // OPM_FLEXIBLE_SOLVER_IMPL_HEADER_INCLUDED
|
||||
|
@ -113,6 +113,8 @@ DenseMatrix transposeDenseMatrix(const DenseMatrix& M)
|
||||
typedef typename GridView::template Codim<0>::Entity Element;
|
||||
typedef typename GET_PROP_TYPE(TypeTag, ElementContext) ElementContext;
|
||||
using FlexibleSolverType = Dune::FlexibleSolver<Matrix, Vector>;
|
||||
using AbstractOperatorType = Dune::AssembledLinearOperator<Matrix, Vector, Vector>;
|
||||
using WellModelOperator = WellModelAsLinearOperator<WellModel, Vector, Vector>;
|
||||
// Due to miscibility oil <-> gas the water eqn is the one we can replace with a pressure equation.
|
||||
static const bool waterEnabled = Indices::waterEnabled;
|
||||
static const int pindex = (waterEnabled) ? BlackOilDefaultIndexTraits::waterCompIdx : BlackOilDefaultIndexTraits::oilCompIdx;
|
||||
@ -176,17 +178,10 @@ DenseMatrix transposeDenseMatrix(const DenseMatrix& M)
|
||||
#endif
|
||||
extractParallelGridInformationToISTL(simulator_.vanguard().grid(), parallelInformation_);
|
||||
useWellConn_ = EWOMS_GET_PARAM(TypeTag, bool, MatrixAddWellContributions);
|
||||
|
||||
if (!useWellConn_ && useFlexible_)
|
||||
{
|
||||
OPM_THROW(std::logic_error, "Flexible solvers and CPR need the well contribution in the matrix. Please run with"
|
||||
" --matrix-add-well-contributions=true");
|
||||
}
|
||||
|
||||
ownersFirst_ = EWOMS_GET_PARAM(TypeTag, bool, OwnerCellsFirst);
|
||||
interiorCellNum_ = detail::numMatrixRowsToUseInSolver(simulator_.vanguard().grid(), ownersFirst_);
|
||||
|
||||
if ( isParallel() && (!ownersFirst_ || parameters_.linear_solver_use_amg_ || useFlexible_ ) ) {
|
||||
if ( isParallel() && (!ownersFirst_ || parameters_.linear_solver_use_amg_) ) {
|
||||
detail::setWellConnections(gridForConn, simulator_.vanguard().schedule().getWellsatEnd(), useWellConn_, wellConnectionsGraph_);
|
||||
// For some reason simulator_.model().elementMapper() is not initialized at this stage
|
||||
// Hence const auto& elemMapper = simulator_.model().elementMapper(); does not work.
|
||||
@ -741,11 +736,26 @@ DenseMatrix transposeDenseMatrix(const DenseMatrix& M)
|
||||
if (recreate_solver || !flexibleSolver_) {
|
||||
if (isParallel()) {
|
||||
#if HAVE_MPI
|
||||
assert(noGhostMat_);
|
||||
flexibleSolver_.reset(new FlexibleSolverType(getMatrix(), *comm_, prm_, weightsCalculator));
|
||||
if (useWellConn_) {
|
||||
assert(noGhostMat_);
|
||||
using ParOperatorType = Dune::OverlappingSchwarzOperator<Matrix, Vector, Vector, Comm>;
|
||||
linearOperatorForFlexibleSolver_ = std::make_unique<ParOperatorType>(getMatrix(), *comm_);
|
||||
flexibleSolver_ = std::make_unique<FlexibleSolverType>(*linearOperatorForFlexibleSolver_, *comm_, prm_, weightsCalculator);
|
||||
} else {
|
||||
if (!ownersFirst_) {
|
||||
OPM_THROW(std::runtime_error, "In parallel, the flexible solver requires "
|
||||
"--owner-cells-first=true when --matrix-add-well-contributions=false is used.");
|
||||
}
|
||||
using ParOperatorType = WellModelGhostLastMatrixAdapter<Matrix, Vector, Vector, true>;
|
||||
wellOperator_ = std::make_unique<WellModelOperator>(simulator_.problem().wellModel());
|
||||
linearOperatorForFlexibleSolver_ = std::make_unique<ParOperatorType>(getMatrix(), *wellOperator_, interiorCellNum_);
|
||||
flexibleSolver_ = std::make_unique<FlexibleSolverType>(*linearOperatorForFlexibleSolver_, *comm_, prm_, weightsCalculator);
|
||||
}
|
||||
#endif
|
||||
} else {
|
||||
flexibleSolver_.reset(new FlexibleSolverType(getMatrix(), prm_, weightsCalculator));
|
||||
using SeqLinearOperator = Dune::MatrixAdapter<Matrix, Vector, Vector>;
|
||||
linearOperatorForFlexibleSolver_ = std::make_unique<SeqLinearOperator>(getMatrix());
|
||||
flexibleSolver_ = std::make_unique<FlexibleSolverType>(*linearOperatorForFlexibleSolver_, prm_, weightsCalculator);
|
||||
}
|
||||
}
|
||||
else
|
||||
@ -1004,6 +1014,8 @@ DenseMatrix transposeDenseMatrix(const DenseMatrix& M)
|
||||
Vector *rhs_;
|
||||
|
||||
std::unique_ptr<FlexibleSolverType> flexibleSolver_;
|
||||
std::unique_ptr<AbstractOperatorType> linearOperatorForFlexibleSolver_;
|
||||
std::unique_ptr<WellModelAsLinearOperator<WellModel, Vector, Vector>> wellOperator_;
|
||||
std::vector<int> overlapRows_;
|
||||
std::vector<int> interiorRows_;
|
||||
std::vector<std::set<int>> wellConnectionsGraph_;
|
||||
|
@ -61,18 +61,19 @@ class ISTLSolverEbosFlexible
|
||||
using Simulator = typename GET_PROP_TYPE(TypeTag, Simulator);
|
||||
using Scalar = typename GET_PROP_TYPE(TypeTag, Scalar);
|
||||
using MatrixType = typename SparseMatrixAdapter::IstlMatrix;
|
||||
using WellModel = typename GET_PROP_TYPE(TypeTag, EclWellModel);
|
||||
#if HAVE_MPI
|
||||
using Communication = Dune::OwnerOverlapCopyCommunication<int, int>;
|
||||
#else
|
||||
using Communication = int; // Dummy type.
|
||||
#endif
|
||||
using AbstractOperatorType = Dune::AssembledLinearOperator<MatrixType, VectorType, VectorType>;
|
||||
using WellModelOpType = WellModelAsLinearOperator<WellModel, VectorType, VectorType>;
|
||||
using SolverType = Dune::FlexibleSolver<MatrixType, VectorType>;
|
||||
|
||||
// for quasiImpesWeights
|
||||
typedef typename GET_PROP_TYPE(TypeTag, GlobalEqVector) Vector;
|
||||
typedef typename GET_PROP_TYPE(TypeTag, Indices) Indices;
|
||||
//typedef typename GET_PROP_TYPE(TypeTag, EclWellModel) WellModel;
|
||||
//typedef typename GET_PROP_TYPE(TypeTag, Simulator) Simulator;
|
||||
typedef typename SparseMatrixAdapter::IstlMatrix Matrix;
|
||||
typedef typename SparseMatrixAdapter::MatrixBlock MatrixBlockType;
|
||||
typedef typename Vector::block_type BlockVector;
|
||||
@ -90,6 +91,9 @@ public:
|
||||
|
||||
explicit ISTLSolverEbosFlexible(const Simulator& simulator)
|
||||
: simulator_(simulator)
|
||||
, ownersFirst_(EWOMS_GET_PARAM(TypeTag, bool, OwnerCellsFirst))
|
||||
, matrixAddWellContributions_(EWOMS_GET_PARAM(TypeTag, bool, MatrixAddWellContributions))
|
||||
, interiorCellNum_(detail::numMatrixRowsToUseInSolver(simulator_.vanguard().grid(), ownersFirst_))
|
||||
{
|
||||
parameters_.template init<TypeTag>();
|
||||
prm_ = setupPropertyTree<TypeTag>(parameters_);
|
||||
@ -134,76 +138,53 @@ public:
|
||||
parinfo->copyValuesTo(comm_->indexSet(), comm_->remoteIndices(), size, 1);
|
||||
firstcall = false;
|
||||
}
|
||||
makeOverlapRowsInvalid(mat.istlMatrix());
|
||||
if (isParallel() && matrixAddWellContributions_) {
|
||||
makeOverlapRowsInvalid(mat.istlMatrix());
|
||||
}
|
||||
#endif
|
||||
// Decide if we should recreate the solver or just do
|
||||
// a minimal preconditioner update.
|
||||
const int newton_iteration = this->simulator_.model().newtonMethod().numIterations();
|
||||
bool recreate_solver = false;
|
||||
if (this->parameters_.cpr_reuse_setup_ == 0) {
|
||||
// Always recreate solver.
|
||||
recreate_solver = true;
|
||||
} else if (this->parameters_.cpr_reuse_setup_ == 1) {
|
||||
// Recreate solver on the first iteration of every timestep.
|
||||
if (newton_iteration == 0) {
|
||||
recreate_solver = true;
|
||||
}
|
||||
} else if (this->parameters_.cpr_reuse_setup_ == 2) {
|
||||
// Recreate solver if the last solve used more than 10 iterations.
|
||||
if (this->iterations() > 10) {
|
||||
recreate_solver = true;
|
||||
}
|
||||
} else {
|
||||
assert(this->parameters_.cpr_reuse_setup_ == 3);
|
||||
assert(recreate_solver == false);
|
||||
// Never recreate solver.
|
||||
}
|
||||
|
||||
std::function<VectorType()> weightsCalculator;
|
||||
matrix_ = &mat.istlMatrix(); // Store pointer for output if needed.
|
||||
std::function<VectorType()> weightsCalculator = getWeightsCalculator(mat.istlMatrix(), b);
|
||||
|
||||
auto preconditionerType = prm_.get("preconditioner.type", "cpr");
|
||||
if( preconditionerType == "cpr" ||
|
||||
preconditionerType == "cprt"
|
||||
)
|
||||
{
|
||||
bool transpose = false;
|
||||
if(preconditionerType == "cprt"){
|
||||
transpose = true;
|
||||
}
|
||||
|
||||
auto weightsType = prm_.get("preconditioner.weight_type", "quasiimpes");
|
||||
auto pressureIndex = this->prm_.get("preconditioner.pressure_var_index", 1);
|
||||
if(weightsType == "quasiimpes") {
|
||||
// weighs will be created as default in the solver
|
||||
weightsCalculator =
|
||||
[&mat, transpose, pressureIndex](){
|
||||
return Opm::Amg::getQuasiImpesWeights<MatrixType,
|
||||
VectorType>(
|
||||
mat.istlMatrix(),
|
||||
pressureIndex,
|
||||
transpose);
|
||||
};
|
||||
|
||||
}else if(weightsType == "trueimpes" ){
|
||||
weightsCalculator =
|
||||
[this, &b, pressureIndex](){
|
||||
return this->getTrueImpesWeights(b, pressureIndex);
|
||||
};
|
||||
}else{
|
||||
OPM_THROW(std::invalid_argument, "Weights type " << weightsType << "not implemented for cpr."
|
||||
<< " Please use quasiimpes or trueimpes.");
|
||||
}
|
||||
}
|
||||
|
||||
if (recreate_solver || !solver_) {
|
||||
if (shouldCreateSolver()) {
|
||||
if (isParallel()) {
|
||||
#if HAVE_MPI
|
||||
matrix_ = &mat.istlMatrix();
|
||||
solver_.reset(new SolverType(mat.istlMatrix(), *comm_, prm_, weightsCalculator));
|
||||
#endif
|
||||
if (matrixAddWellContributions_) {
|
||||
using ParOperatorType = Dune::OverlappingSchwarzOperator<MatrixType, VectorType, VectorType, Communication>;
|
||||
auto op = std::make_unique<ParOperatorType>(mat.istlMatrix(), *comm_);
|
||||
auto sol = std::make_unique<SolverType>(*op, *comm_, prm_, weightsCalculator);
|
||||
solver_ = std::move(sol);
|
||||
linear_operator_ = std::move(op);
|
||||
} else {
|
||||
if (!ownersFirst_) {
|
||||
OPM_THROW(std::runtime_error, "In parallel, the flexible solver requires "
|
||||
"--owner-cells-first=true when --matrix-add-well-contributions=false is used.");
|
||||
}
|
||||
using ParOperatorType = WellModelGhostLastMatrixAdapter<MatrixType, VectorType, VectorType, true>;
|
||||
auto well_op = std::make_unique<WellModelOpType>(simulator_.problem().wellModel());
|
||||
auto op = std::make_unique<ParOperatorType>(mat.istlMatrix(), *well_op, interiorCellNum_);
|
||||
auto sol = std::make_unique<SolverType>(*op, *comm_, prm_, weightsCalculator);
|
||||
solver_ = std::move(sol);
|
||||
linear_operator_ = std::move(op);
|
||||
well_operator_ = std::move(well_op);
|
||||
}
|
||||
#endif // HAVE_MPI
|
||||
} else {
|
||||
matrix_ = &mat.istlMatrix();
|
||||
solver_.reset(new SolverType(mat.istlMatrix(), prm_, weightsCalculator));
|
||||
if (matrixAddWellContributions_) {
|
||||
using SeqOperatorType = Dune::MatrixAdapter<MatrixType, VectorType, VectorType>;
|
||||
auto op = std::make_unique<SeqOperatorType>(mat.istlMatrix());
|
||||
auto sol = std::make_unique<SolverType>(*op, prm_, weightsCalculator);
|
||||
solver_ = std::move(sol);
|
||||
linear_operator_ = std::move(op);
|
||||
} else {
|
||||
using SeqOperatorType = WellModelMatrixAdapter<MatrixType, VectorType, VectorType, false>;
|
||||
auto well_op = std::make_unique<WellModelOpType>(simulator_.problem().wellModel());
|
||||
auto op = std::make_unique<SeqOperatorType>(mat.istlMatrix(), *well_op);
|
||||
auto sol = std::make_unique<SolverType>(*op, prm_, weightsCalculator);
|
||||
solver_ = std::move(sol);
|
||||
linear_operator_ = std::move(op);
|
||||
well_operator_ = std::move(well_op);
|
||||
}
|
||||
}
|
||||
rhs_ = b;
|
||||
} else {
|
||||
@ -245,6 +226,63 @@ public:
|
||||
|
||||
protected:
|
||||
|
||||
bool shouldCreateSolver() const
|
||||
{
|
||||
// Decide if we should recreate the solver or just do
|
||||
// a minimal preconditioner update.
|
||||
if (!solver_) {
|
||||
return true;
|
||||
}
|
||||
const int newton_iteration = this->simulator_.model().newtonMethod().numIterations();
|
||||
bool recreate_solver = false;
|
||||
if (this->parameters_.cpr_reuse_setup_ == 0) {
|
||||
// Always recreate solver.
|
||||
recreate_solver = true;
|
||||
} else if (this->parameters_.cpr_reuse_setup_ == 1) {
|
||||
// Recreate solver on the first iteration of every timestep.
|
||||
if (newton_iteration == 0) {
|
||||
recreate_solver = true;
|
||||
}
|
||||
} else if (this->parameters_.cpr_reuse_setup_ == 2) {
|
||||
// Recreate solver if the last solve used more than 10 iterations.
|
||||
if (this->iterations() > 10) {
|
||||
recreate_solver = true;
|
||||
}
|
||||
} else {
|
||||
assert(this->parameters_.cpr_reuse_setup_ == 3);
|
||||
assert(recreate_solver == false);
|
||||
// Never recreate solver.
|
||||
}
|
||||
return recreate_solver;
|
||||
}
|
||||
|
||||
std::function<VectorType()> getWeightsCalculator(const MatrixType& mat, const VectorType& b) const
|
||||
{
|
||||
std::function<VectorType()> weightsCalculator;
|
||||
|
||||
auto preconditionerType = prm_.get("preconditioner.type", "cpr");
|
||||
if (preconditionerType == "cpr" || preconditionerType == "cprt") {
|
||||
const bool transpose = preconditionerType == "cprt";
|
||||
const auto weightsType = prm_.get("preconditioner.weight_type", "quasiimpes");
|
||||
const auto pressureIndex = this->prm_.get("preconditioner.pressure_var_index", 1);
|
||||
if (weightsType == "quasiimpes") {
|
||||
// weighs will be created as default in the solver
|
||||
weightsCalculator = [&mat, transpose, pressureIndex]() {
|
||||
return Opm::Amg::getQuasiImpesWeights<MatrixType, VectorType>(mat, pressureIndex, transpose);
|
||||
};
|
||||
} else if (weightsType == "trueimpes") {
|
||||
weightsCalculator = [this, &b, pressureIndex]() {
|
||||
return this->getTrueImpesWeights(b, pressureIndex);
|
||||
};
|
||||
} else {
|
||||
OPM_THROW(std::invalid_argument,
|
||||
"Weights type " << weightsType << "not implemented for cpr."
|
||||
<< " Please use quasiimpes or trueimpes.");
|
||||
}
|
||||
}
|
||||
return weightsCalculator;
|
||||
}
|
||||
|
||||
/// Zero out off-diagonal blocks on rows corresponding to overlap cells
|
||||
/// Diagonal blocks on ovelap rows are set to diag(1.0).
|
||||
void makeOverlapRowsInvalid(MatrixType& matrix) const
|
||||
@ -267,7 +305,7 @@ protected:
|
||||
}
|
||||
}
|
||||
|
||||
VectorType getTrueImpesWeights(const VectorType& b,const int pressureVarIndex)
|
||||
VectorType getTrueImpesWeights(const VectorType& b, const int pressureVarIndex) const
|
||||
{
|
||||
VectorType weights(b.size());
|
||||
ElementContext elemCtx(simulator_);
|
||||
@ -289,14 +327,20 @@ protected:
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
const Simulator& simulator_;
|
||||
MatrixType* matrix_;
|
||||
std::unique_ptr<WellModelOpType> well_operator_;
|
||||
std::unique_ptr<AbstractOperatorType> linear_operator_;
|
||||
std::unique_ptr<SolverType> solver_;
|
||||
FlowLinearSolverParameters parameters_;
|
||||
boost::property_tree::ptree prm_;
|
||||
VectorType rhs_;
|
||||
Dune::InverseOperatorResult res_;
|
||||
std::any parallelInformation_;
|
||||
bool ownersFirst_;
|
||||
bool matrixAddWellContributions_;
|
||||
int interiorCellNum_;
|
||||
std::unique_ptr<Communication> comm_;
|
||||
std::vector<int> overlapRows_;
|
||||
std::vector<int> interiorRows_;
|
||||
|
@ -25,6 +25,7 @@
|
||||
#include <boost/property_tree/ptree.hpp>
|
||||
|
||||
#include <dune/istl/solver.hh>
|
||||
#include <dune/istl/owneroverlapcopy.hh>
|
||||
|
||||
namespace Dune
|
||||
{
|
||||
@ -55,19 +56,24 @@ namespace Amg
|
||||
* The operator will use one step of AMG to approximately solve
|
||||
* the coarse level system.
|
||||
*/
|
||||
struct PressureInverseOperator : public Dune::InverseOperator<X, X> {
|
||||
template <class Comm>
|
||||
PressureInverseOperator(Operator& op, const boost::property_tree::ptree& prm, const Comm& comm)
|
||||
struct PressureInverseOperator : public Dune::InverseOperator<X, X>
|
||||
{
|
||||
template <typename GlobalIndex, typename LocalIndex>
|
||||
PressureInverseOperator(Operator& op,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const Dune::OwnerOverlapCopyCommunication<GlobalIndex, LocalIndex>& comm)
|
||||
: linsolver_()
|
||||
{
|
||||
assert(op.category() == Dune::SolverCategory::overlapping);
|
||||
linsolver_ = std::make_unique<Solver>(op.getmat(), comm, prm, std::function<X()>());
|
||||
linsolver_ = std::make_unique<Solver>(op, comm, prm, std::function<X()>());
|
||||
}
|
||||
PressureInverseOperator(Operator& op, const boost::property_tree::ptree& prm, const SequentialInformation&)
|
||||
PressureInverseOperator(Operator& op,
|
||||
const boost::property_tree::ptree& prm,
|
||||
const SequentialInformation&)
|
||||
: linsolver_()
|
||||
{
|
||||
assert(op.category() != Dune::SolverCategory::overlapping);
|
||||
linsolver_ = std::make_unique<Solver>(op.getmat(), prm, std::function<X()>());
|
||||
linsolver_ = std::make_unique<Solver>(op, prm, std::function<X()>());
|
||||
}
|
||||
|
||||
|
||||
|
@ -75,7 +75,9 @@ testSolver(const boost::property_tree::ptree& prm, const std::string& matrix_fil
|
||||
prm.get<int>("preconditioner.pressure_var_index"),
|
||||
transpose);
|
||||
};
|
||||
Dune::FlexibleSolver<Matrix, Vector> solver(matrix, prm, wc);
|
||||
using SeqOperatorType = Dune::MatrixAdapter<Matrix, Vector, Vector>;
|
||||
SeqOperatorType op(matrix);
|
||||
Dune::FlexibleSolver<Matrix, Vector> solver(op, prm, wc);
|
||||
Vector x(rhs.size());
|
||||
Dune::InverseOperatorResult res;
|
||||
solver.apply(x, rhs, res);
|
||||
|
Loading…
Reference in New Issue
Block a user