A complete nonlinear elliptic solver stack. Use to compose an executable. More...
#include <Solver.hpp>
Classes | |
| struct | RandomizeInitialGuess |
Public Types | |
| using | system = System |
| using | background_tag = elliptic::Tags::Background< elliptic::analytic_data::Background > |
| using | initial_guess_tag = elliptic::Tags::InitialGuess< elliptic::analytic_data::InitialGuess > |
| using | fields_tag = ::Tags::Variables< typename system::primal_fields > |
| These are the fields we solve for. | |
| using | fluxes_tag = ::Tags::Variables< typename system::primal_fluxes > |
| These are the fluxes corresponding to the fields, i.e. essentially their first derivatives. These are background fields for the linearized sources. | |
| using | fixed_sources_tag = db::add_tag_prefix<::Tags::FixedSource, fields_tag > |
| These are the fixed sources, i.e. the RHS of the equations. | |
| using | operator_applied_to_fields_tag = implementation defined |
| using | amr_iteration_id = Convergence::Tags::IterationId<::amr::OptionTags::AmrGroup > |
| using | nonlinear_solver = NonlinearSolver::newton_raphson::NewtonRaphson< Metavariables, fields_tag, OptionTags::NewtonRaphsonGroup, fixed_sources_tag, LinearSolver::multigrid::Tags::IsFinestGrid > |
| The nonlinear solver algorithm. | |
| using | nonlinear_solver_iteration_id = Convergence::Tags::IterationId< typename nonlinear_solver::options_group > |
| using | linear_solver = LinearSolver::gmres::Gmres< Metavariables, tmpl::conditional_t< is_linear, fields_tag, typename nonlinear_solver::linear_solver_fields_tag >, OptionTags::GmresGroup, true, tmpl::conditional_t< is_linear, fixed_sources_tag, typename nonlinear_solver::linear_solver_source_tag >, LinearSolver::multigrid::Tags::IsFinestGrid > |
| The linear solver algorithm. We use GMRES since the operator is not necessarily symmetric. Using CG here would be an optimization for symmetric problems. | |
| using | linear_solver_iteration_id = Convergence::Tags::IterationId< typename linear_solver::options_group > |
| using | multigrid = LinearSolver::multigrid::Multigrid< Metavariables, volume_dim, typename linear_solver::operand_tag, OptionTags::MultigridGroup, elliptic::dg::Tags::Massive, typename linear_solver::preconditioner_source_tag > |
| Precondition each linear solver iteration with a multigrid V-cycle. | |
| using | subdomain_operator = elliptic::dg::subdomain_operator::SubdomainOperator< system, OptionTags::SchwarzSmootherGroup > |
| Smooth each multigrid level with a number of Schwarz smoothing steps. | |
| using | subdomain_preconditioners = implementation defined |
| using | schwarz_smoother = LinearSolver::Schwarz::Schwarz< typename multigrid::smooth_fields_tag, OptionTags::SchwarzSmootherGroup, subdomain_operator, subdomain_preconditioners, typename multigrid::smooth_source_tag, LinearSolver::multigrid::Tags::MultigridLevel > |
| using | vars_tag = typename linear_solver::operand_tag |
| For the GMRES linear solver we need to apply the DG operator to its internal "operand" in every iteration of the algorithm. | |
| using | operator_applied_to_vars_tag = db::add_tag_prefix< LinearSolver::Tags::OperatorAppliedTo, vars_tag > |
| using | fluxes_vars_tag = db::add_tag_prefix< NonlinearSolver::Tags::Correction, fluxes_tag > |
| The correction fluxes can be stored in an arbitrary tag. We don't need to access them anywhere, they're just a memory buffer for the linearized operator. | |
| using | observe_fields = implementation defined |
| Fields that may be observed to monitor the state of the solver. | |
| using | observed_reduction_data_tags = observers::collect_reduction_data_tags< tmpl::flatten< tmpl::list< tmpl::conditional_t< is_linear, tmpl::list<>, nonlinear_solver >, linear_solver, multigrid, schwarz_smoother > > > |
| Collect all reduction tags for observers. | |
| template<bool Linearized> | |
| using | dg_operator = elliptic::dg::Actions::DgOperator< system, Linearized, tmpl::conditional_t< Linearized, linear_solver_iteration_id, nonlinear_solver_iteration_id >, tmpl::conditional_t< Linearized, vars_tag, fields_tag >, tmpl::conditional_t< Linearized, fluxes_vars_tag, fluxes_tag >, tmpl::conditional_t< Linearized, operator_applied_to_vars_tag, operator_applied_to_fields_tag > > |
| using | build_matrix = LinearSolver::Actions::BuildMatrix< typename dg_operator< true >::temporal_id_tag, fields_tag, fixed_sources_tag, vars_tag, operator_applied_to_vars_tag, domain::Tags::Coordinates< volume_dim, Frame::Inertial >, LinearSolver::multigrid::Tags::IsFinestGrid > |
| using | build_matrix_actions = typename build_matrix::template actions< typename dg_operator< true >::apply_actions > |
| using | init_analytic_solution_action = elliptic::Actions::InitializeOptionalAnalyticSolution< volume_dim, background_tag, tmpl::append< typename system::primal_fields, typename system::primal_fluxes >, elliptic::analytic_data::AnalyticSolution > |
| using | initialization_actions = implementation defined |
| using | register_actions = implementation defined |
| template<typename Label > | |
| using | smooth_actions = typename schwarz_smoother::template solve< typename dg_operator< true >::apply_actions, Label > |
| using | subdomain_init_tags = implementation defined |
| using | communicated_overlap_tags = implementation defined |
| This data needs to be communicated on subdomain overlap regions. | |
| using | init_subdomain_action = elliptic::dg::subdomain_operator::Actions::InitializeSubdomain< system, background_tag, typename schwarz_smoother::options_group, false > |
| template<typename StepActions > | |
| using | linear_solve_actions = typename linear_solver::template solve< tmpl::list< typename multigrid::template solve< typename dg_operator< true >::apply_actions, smooth_actions< LinearSolver::multigrid::VcycleDownLabel >, smooth_actions< LinearSolver::multigrid::VcycleUpLabel > >, ::LinearSolver::Actions::make_identity_if_skipped< multigrid, typename dg_operator< true >::apply_actions > >, StepActions > |
| template<typename StepActions > | |
| using | nonlinear_solve_actions = typename nonlinear_solver::template solve< typename dg_operator< false >::apply_actions, tmpl::list< LinearSolver::multigrid::Actions::ReceiveFieldsFromFinerGrid< volume_dim, tmpl::list< fields_tag, fluxes_tag >, typename multigrid::options_group >, LinearSolver::multigrid::Actions::SendFieldsToCoarserGrid< tmpl::list< fields_tag, fluxes_tag >, typename multigrid::options_group, void >, LinearSolver::Schwarz::Actions::SendOverlapFields< communicated_overlap_tags, typename schwarz_smoother::options_group, false, nonlinear_solver_iteration_id >, LinearSolver::Schwarz::Actions::ReceiveOverlapFields< volume_dim, communicated_overlap_tags, typename schwarz_smoother::options_group, false, nonlinear_solver_iteration_id >, LinearSolver::Schwarz::Actions::ResetSubdomainSolver< typename schwarz_smoother::options_group >, linear_solve_actions< tmpl::list<> > >, StepActions > |
| template<typename StepActions > | |
| using | solve_actions = implementation defined |
| template<typename Tag > | |
| using | overlaps_tag = LinearSolver::Schwarz::Tags::Overlaps< Tag, volume_dim, OptionTags::SchwarzSmootherGroup > |
| using | amr_projectors = implementation defined |
| using | component_list = implementation defined |
Static Public Attributes | |
| static constexpr size_t | volume_dim = Dim |
| static constexpr bool | is_linear |
A complete nonlinear elliptic solver stack. Use to compose an executable.
The elliptic solver stack is described in detail in [201].
Uses Metavariables to instantiate parallel components.
|
staticconstexpr |