![]() |
Lethe Open-Source Library version 1.2.0
Developer Guide
|
#include <inexact_newton_non_linear_solver_strategy.h>
Public Member Functions | |
| InexactNewtonNonLinearSolverStrategy (PhysicsSolver< VectorType > *physics_solver, const Parameters::NonLinearSolver ¶m) | |
| void | solve () override |
| Public Member Functions inherited from PhysicsSolverStrategy< VectorType > | |
| PhysicsSolverStrategy (PhysicsSolver< VectorType > *physics_solver, const Parameters::NonLinearSolver ¶ms) | |
| PhysicsSolverStrategy (PhysicsSolver< VectorType > *physics_solver) | |
| virtual | ~PhysicsSolverStrategy ()=default |
| unsigned int | get_current_newton_iteration () const |
Private Attributes | |
| bool | matrix_requires_assembly |
Additional Inherited Members | |
| Protected Member Functions inherited from PhysicsSolverStrategy< VectorType > | |
| bool | skip_linear_solve_if_fresh_rhs_is_already_converged (const bool rhs_was_assembled_this_iteration, const double rescale_metric, double ¤t_res, double &last_res, double &global_res) |
| Protected Attributes inherited from PhysicsSolverStrategy< VectorType > | |
| PhysicsSolver< VectorType > * | physics_solver |
| Parameters::NonLinearSolver | params |
| unsigned int | outer_iteration |
Non-linear solver for non-linear systems of equations which uses an inexact Newton method with \(\alpha\) relaxation to ensure that the residual is monotonically decreasing. It allows to reuse the matrix and avoid its reassembly for every Newton step.
Definition at line 19 of file inexact_newton_non_linear_solver_strategy.h.
| InexactNewtonNonLinearSolverStrategy< VectorType >::InexactNewtonNonLinearSolverStrategy | ( | PhysicsSolver< VectorType > * | physics_solver, |
| const Parameters::NonLinearSolver & | param ) |
Constructor.
| [in] | physics_solver | A pointer to the physics solver to which the non-linear solver is attached. |
| [in] | param | Non-linear solver parameters as specified in the simulation parameter file. |
Definition at line 50 of file inexact_newton_non_linear_solver_strategy.h.
References matrix_requires_assembly, PhysicsSolverStrategy< VectorType >::params, PhysicsSolverStrategy< VectorType >::physics_solver, and PhysicsSolverStrategy< VectorType >::PhysicsSolverStrategy().
|
overridevirtual |
Solve the non-linear system of equations.
Implements PhysicsSolverStrategy< VectorType >.
Definition at line 60 of file inexact_newton_non_linear_solver_strategy.h.
References PhysicsSolver< VectorType >::apply_constraints(), PhysicsSolver< VectorType >::assemble_system_matrix(), PhysicsSolver< VectorType >::assemble_system_rhs(), PhysicsSolver< VectorType >::get_current_residual(), PhysicsSolver< VectorType >::get_evaluation_point(), PhysicsSolver< VectorType >::get_local_evaluation_point(), PhysicsSolver< VectorType >::get_newton_update(), PhysicsSolver< VectorType >::get_present_solution(), PhysicsSolver< VectorType >::get_residual_rescale_metric(), PhysicsSolver< VectorType >::get_system_rhs(), matrix_requires_assembly, PhysicsSolverStrategy< VectorType >::outer_iteration, PhysicsSolver< VectorType >::output_newton_update_norms(), PhysicsSolverStrategy< VectorType >::params, PhysicsSolver< VectorType >::pcout, PhysicsSolverStrategy< VectorType >::physics_solver, Parameters::quiet, PhysicsSolver< VectorType >::setup_preconditioner(), PhysicsSolverStrategy< VectorType >::skip_linear_solve_if_fresh_rhs_is_already_converged(), and PhysicsSolver< VectorType >::solve_linear_system().
|
private |
Definition at line 46 of file inexact_newton_non_linear_solver_strategy.h.