libMesh::RBConstruction Class Reference

This class is part of the rbOOmit framework. More...

#include <rb_construction.h>

Inheritance diagram for libMesh::RBConstruction:

Public Types
typedef RBConstruction	sys_type
	The type of system. More...
typedef RBConstructionBase< LinearImplicitSystem >	Parent
	The type of the parent. More...
typedef Number(*	ValueFunctionPointer) (const Point &p, const Parameters &Parameters, const std::string &sys_name, const std::string &unknown_name)
	Projects arbitrary functions onto the current solution. More...
typedef Gradient(*	GradientFunctionPointer) (const Point &p, const Parameters &parameters, const std::string &sys_name, const std::string &unknown_name)
typedef std::map< std::string, std::unique_ptr< NumericVector< Number > >, std::less<> >::iterator	vectors_iterator
	Vector iterator typedefs. More...
typedef std::map< std::string, std::unique_ptr< NumericVector< Number > >, std::less<> >::const_iterator	const_vectors_iterator
typedef std::map< std::string, std::unique_ptr< SparseMatrix< Number > >, std::less<> >::iterator	matrices_iterator
	Matrix iterator typedefs. More...
typedef std::map< std::string, std::unique_ptr< SparseMatrix< Number > >, std::less<> >::const_iterator	const_matrices_iterator

Public Member Functions
	RBConstruction (EquationSystems &es, const std::string &name, const unsigned int number)
	Constructor. More...
	RBConstruction (RBConstruction &&)=default
	Special functions. More...
	RBConstruction (const RBConstruction &)=delete
RBConstruction &	operator= (const RBConstruction &)=delete
RBConstruction &	operator= (RBConstruction &&)=delete
virtual	~RBConstruction ()
virtual void	solve_for_matrix_and_rhs (LinearSolver< Number > &input_solver, SparseMatrix< Number > &input_matrix, NumericVector< Number > &input_rhs)
	Assembles & solves the linear system A*x=b for the specified matrix input_matrix and right-hand side rhs. More...
void	set_rb_evaluation (RBEvaluation &rb_eval_in)
	Set the RBEvaluation object. More...
RBEvaluation &	get_rb_evaluation ()
	Get a reference to the RBEvaluation object. More...
const RBEvaluation &	get_rb_evaluation () const
bool	is_rb_eval_initialized () const
RBThetaExpansion &	get_rb_theta_expansion ()
	Get a reference to the RBThetaExpansion object that that belongs to rb_eval. More...
const RBThetaExpansion &	get_rb_theta_expansion () const
void	set_rb_assembly_expansion (RBAssemblyExpansion &rb_assembly_expansion_in)
	Set the rb_assembly_expansion object. More...
RBAssemblyExpansion &	get_rb_assembly_expansion ()
sys_type &	system ()
virtual void	clear () override
	Clear all the data structures associated with the system. More...
virtual std::string	system_type () const override
virtual Real	truth_solve (int plot_solution)
	Perform a "truth" solve, i.e. More...
virtual Real	train_reduced_basis (const bool resize_rb_eval_data=true)
	Train the reduced basis. More...
Real	train_reduced_basis_with_greedy (const bool resize_rb_eval_data)
	Train the reduced basis using the "Greedy algorithm.". More...
void	enrich_basis_from_rhs_terms (const bool resize_rb_eval_data=true)
	This function computes one basis function for each rhs term. More...
void	train_reduced_basis_with_POD ()
	Train the reduced basis using Proper Orthogonal Decomposition (POD). More...
virtual Real	compute_max_error_bound ()
	(i) Compute the a posteriori error bound for each set of parameters in the training set, (ii) set current_parameters to the parameters that maximize the error bound, and (iii) return the maximum error bound. More...
const RBParameters &	get_greedy_parameter (unsigned int i)
	Return the parameters chosen during the i^th step of the Greedy algorithm. More...
void	set_rel_training_tolerance (Real new_training_tolerance)
	Get/set the relative tolerance for the basis training. More...
Real	get_rel_training_tolerance () const
void	set_abs_training_tolerance (Real new_training_tolerance)
	Get/set the absolute tolerance for the basis training. More...
Real	get_abs_training_tolerance () const
void	set_normalize_rb_bound_in_greedy (bool normalize_rb_bound_in_greedy_in)
	Get/set the boolean to indicate if we normalize the RB error in the greedy. More...
bool	get_normalize_rb_bound_in_greedy () const
virtual bool	is_serial_training_type (const std::string &RB_training_type_in)
void	set_RB_training_type (const std::string &RB_training_type_in)
	Get/set the string that determines the training type. More...
const std::string &	get_RB_training_type () const
unsigned int	get_Nmax () const
	Get/set Nmax, the maximum number of RB functions we are willing to compute. More...
virtual void	set_Nmax (unsigned int Nmax)
virtual void	load_basis_function (unsigned int i)
	Load the i^th RB function into the RBConstruction solution vector. More...
virtual void	load_rb_solution ()
	Load the RB solution from the most recent solve with rb_eval into this system's solution vector. More...
Real	compute_residual_dual_norm_slow (const unsigned int N)
	The slow (but simple, non-error prone) way to compute the residual dual norm. More...
SparseMatrix< Number > *	get_inner_product_matrix ()
	Get a pointer to inner_product_matrix. More...
const SparseMatrix< Number > *	get_inner_product_matrix () const
SparseMatrix< Number > *	get_non_dirichlet_inner_product_matrix ()
	Get the non-Dirichlet (or more generally no-constraints) version of the inner-product matrix. More...
const SparseMatrix< Number > *	get_non_dirichlet_inner_product_matrix () const
SparseMatrix< Number > *	get_non_dirichlet_inner_product_matrix_if_avail ()
	Get the non-Dirichlet inner-product matrix if it's available, otherwise get the inner-product matrix with constraints. More...
const SparseMatrix< Number > *	get_non_dirichlet_inner_product_matrix_if_avail () const
SparseMatrix< Number > *	get_Aq (unsigned int q)
	Get a pointer to Aq. More...
SparseMatrix< Number > *	get_non_dirichlet_Aq (unsigned int q)
	Get a pointer to non_dirichlet_Aq. More...
SparseMatrix< Number > *	get_non_dirichlet_Aq_if_avail (unsigned int q)
	Get a pointer to non_dirichlet_Aq if it's available, otherwise get Aq. More...
virtual void	initialize_rb_construction (bool skip_matrix_assembly=false, bool skip_vector_assembly=false)
	Allocate all the data structures necessary for the construction stage of the RB method. More...
NumericVector< Number > *	get_Fq (unsigned int q)
	Get a pointer to Fq. More...
NumericVector< Number > *	get_non_dirichlet_Fq (unsigned int q)
	Get a pointer to non-Dirichlet Fq. More...
NumericVector< Number > *	get_non_dirichlet_Fq_if_avail (unsigned int q)
	Get a pointer to non_dirichlet_Fq if it's available, otherwise get Fq. More...
NumericVector< Number > *	get_output_vector (unsigned int n, unsigned int q_l)
	Get a pointer to the n^th output. More...
NumericVector< Number > *	get_non_dirichlet_output_vector (unsigned int n, unsigned int q_l)
	Get a pointer to non-Dirichlet output vector. More...
virtual void	get_all_matrices (std::map< std::string, SparseMatrix< Number > *> &all_matrices)
	Get a map that stores pointers to all of the matrices. More...
virtual void	get_all_vectors (std::map< std::string, NumericVector< Number > *> &all_vectors)
	Get a map that stores pointers to all of the vectors. More...
virtual void	get_output_vectors (std::map< std::string, NumericVector< Number > *> &all_vectors)
	Get a map that stores pointers to all of the vectors. More...
virtual void	assemble_affine_expansion (bool skip_matrix_assembly, bool skip_vector_assembly)
	Assemble the matrices and vectors for this system. More...
void	assemble_inner_product_matrix (SparseMatrix< Number > *input_matrix, bool apply_dof_constraints=true)
	Assemble the inner product matrix and store it in input_matrix. More...
void	assemble_Aq_matrix (unsigned int q, SparseMatrix< Number > *input_matrix, bool apply_dof_constraints=true)
	Assemble the q^th affine matrix and store it in input_matrix. More...
void	assemble_Fq_vector (unsigned int q, NumericVector< Number > *input_vector, bool apply_dof_constraints=true)
	Assemble the q^th affine vector and store it in input_matrix. More...
void	add_scaled_Aq (Number scalar, unsigned int q_a, SparseMatrix< Number > *input_matrix, bool symmetrize)
	Add the scaled q^th affine matrix to input_matrix. More...
virtual void	write_riesz_representors_to_files (const std::string &riesz_representors_dir, const bool write_binary_residual_representors)
	Write out all the Riesz representor data to files. More...
virtual void	read_riesz_representors_from_files (const std::string &riesz_representors_dir, const bool write_binary_residual_representors)
	Read in all the Riesz representor data from files. More...
virtual void	recompute_all_residual_terms (const bool compute_inner_products=true)
	This function computes all of the residual representors, can be useful when restarting a basis training computation. More...
virtual void	process_parameters_file (const std::string &parameters_filename)
	Read in from the file specified by parameters_filename and set the this system's member variables accordingly. More...
void	set_rb_construction_parameters (unsigned int n_training_samples_in, bool deterministic_training_in, int training_parameters_random_seed_in, bool quiet_mode_in, unsigned int Nmax_in, Real rel_training_tolerance_in, Real abs_training_tolerance_in, bool normalize_rb_error_bound_in_greedy_in, const std::string &RB_training_type_in, const RBParameters &mu_min_in, const RBParameters &mu_max_in, const std::map< std::string, std::vector< Real >> &discrete_parameter_values_in, const std::map< std::string, bool > &log_scaling, std::map< std::string, std::vector< RBParameter >> *training_sample_list=nullptr)
	Set the state of this RBConstruction object based on the arguments to this function. More...
virtual void	print_info () const
	Print out info that describes the current setup of this RBConstruction. More...
void	print_basis_function_orthogonality () const
	Print out a matrix that shows the orthogonality of the RB basis functions. More...
unsigned int	get_delta_N () const
	Get delta_N, the number of basis functions we add to the RB space per iteration of the greedy algorithm. More...
void	set_inner_product_assembly (ElemAssembly &inner_product_assembly_in)
	Set the rb_assembly_expansion object. More...
ElemAssembly &	get_inner_product_assembly ()
void	set_energy_inner_product (const std::vector< Number > &energy_inner_product_coeffs_in)
	Specify the coefficients of the A_q operators to be used in the energy inner-product. More...
void	zero_constrained_dofs_on_vector (NumericVector< Number > &vector) const
	It is sometimes useful to be able to zero vector entries that correspond to constrained dofs. More...
virtual bool	check_if_zero_truth_solve () const
void	set_convergence_assertion_flag (bool flag)
	Setter for the flag determining if convergence should be checked after each solve. More...
bool	get_preevaluate_thetas_flag () const
	Get/set flag to pre-evaluate the theta functions. More...
void	set_preevaluate_thetas_flag (bool flag)
void	set_quiet_mode (bool quiet_mode_in)
	Set the quiet_mode flag. More...
bool	is_quiet () const
	Is the system in quiet mode? More...
void	set_normalize_solution_snapshots (bool value)
	Set the boolean option that indicates if we normalization solution snapshots or not. More...
numeric_index_type	get_n_training_samples () const
	Get the number of global training samples. More...
numeric_index_type	get_local_n_training_samples () const
	Get the total number of training samples local to this processor. More...
numeric_index_type	get_first_local_training_index () const
	Get the first local index of the training parameters. More...
numeric_index_type	get_last_local_training_index () const
	Get the last local index of the training parameters. More...
virtual void	initialize_training_parameters (const RBParameters &mu_min, const RBParameters &mu_max, const unsigned int n_global_training_samples, const std::map< std::string, bool > &log_param_scale, const bool deterministic=true)
	Initialize the parameter ranges and indicate whether deterministic or random training parameters should be used and whether or not we want the parameters to be scaled logarithmically. More...
virtual void	load_training_set (const std::map< std::string, std::vector< RBParameter >> &new_training_set)
	Overwrite the training parameters with new_training_set. More...
void	set_training_parameter_values (const std::string &param_name, const std::vector< RBParameter > &values)
	Overwrite the local training samples for param_name using values. More...
void	broadcast_parameters (const unsigned int proc_id)
	Broadcasts parameters from processor proc_id to all processors. More...
void	set_training_random_seed (int seed)
	Set the seed that is used to randomly generate training parameters. More...
void	set_deterministic_training_parameter_name (const std::string &name)
	In some cases we only want to allow discrete parameter values, instead of parameters that may take any value in a specified interval. More...
const std::string &	get_deterministic_training_parameter_name () const
	Get the name of the parameter that we will generate deterministic training parameters for. More...
virtual void	reinit () override
	Reinitializes the member data fields associated with the system, so that, e.g., assemble() may be used. More...
virtual void	assemble () override
	Prepares matrix and _dof_map for matrix assembly. More...
virtual void	restrict_solve_to (const SystemSubset *subset, const SubsetSolveMode subset_solve_mode=SUBSET_ZERO) override
	After calling this method, any solve will be limited to the given subset. More...
virtual void	solve () override
	Assembles & solves the linear system A*x=b. More...
virtual LinearSolver< Number > *	get_linear_solver () const override
virtual void	assembly (bool get_residual, bool get_jacobian, bool apply_heterogeneous_constraints=false, bool apply_no_constraints=false) override
	Assembles a residual in rhs and/or a jacobian in matrix, as requested. More...
unsigned int	n_linear_iterations () const
Real	final_linear_residual () const
void	attach_shell_matrix (ShellMatrix< Number > *shell_matrix)
	This function enables the user to provide a shell matrix, i.e. More...
void	detach_shell_matrix ()
	Detaches a shell matrix. More...
ShellMatrix< Number > *	get_shell_matrix ()
virtual void	create_static_condensation () override
	Request that static condensation be performed for this system. More...
virtual void	disable_cache () override
	Avoids use of any cached data that might affect any solve result. More...
virtual std::pair< unsigned int, Real >	get_linear_solve_parameters () const
virtual void	assemble_residual_derivatives (const ParameterVector &parameters) override
	Residual parameter derivative function. More...
virtual std::pair< unsigned int, Real >	sensitivity_solve (const ParameterVector &parameters) override
	Assembles & solves the linear system(s) (dR/du)*u_p = -dR/dp, for those parameters contained within parameters. More...
virtual std::pair< unsigned int, Real >	weighted_sensitivity_solve (const ParameterVector &parameters, const ParameterVector &weights) override
	Assembles & solves the linear system(s) (dR/du)*u_w = sum(w_p*-dR/dp), for those parameters p contained within parameters weighted by the values w_p found within weights. More...
virtual std::pair< unsigned int, Real >	adjoint_solve (const QoISet &qoi_indices=QoISet()) override
	Assembles & solves the linear system (dR/du)^T*z = dq/du, for those quantities of interest q specified by qoi_indices. More...
virtual std::pair< unsigned int, Real >	weighted_sensitivity_adjoint_solve (const ParameterVector &parameters, const ParameterVector &weights, const QoISet &qoi_indices=QoISet()) override
	Assembles & solves the linear system(s) (dR/du)^T*z_w = sum(w_p*(d^2q/dudp - d^2R/dudp*z)), for those parameters p contained within parameters, weighted by the values w_p found within weights. More...
virtual void	adjoint_qoi_parameter_sensitivity (const QoISet &qoi_indices, const ParameterVector &parameters, SensitivityData &sensitivities) override
	Solves for the derivative of each of the system's quantities of interest q in qoi[qoi_indices] with respect to each parameter in parameters, placing the result for qoi i and parameter j into sensitivities[i][j]. More...
virtual void	forward_qoi_parameter_sensitivity (const QoISet &qoi_indices, const ParameterVector &parameters, SensitivityData &sensitivities) override
	Solves for the derivative of each of the system's quantities of interest q in qoi[qoi_indices] with respect to each parameter in parameters, placing the result for qoi i and parameter j into sensitivities[i][j]. More...
virtual void	qoi_parameter_hessian (const QoISet &qoi_indices, const ParameterVector &parameters, SensitivityData &hessian) override
	For each of the system's quantities of interest q in qoi[qoi_indices], and for a vector of parameters p, the parameter sensitivity Hessian H_ij is defined as H_ij = (d^2 q)/(d p_i d p_j) This Hessian is the output of this method, where for each q_i, H_jk is stored in hessian.second_derivative(i,j,k). More...
virtual void	qoi_parameter_hessian_vector_product (const QoISet &qoi_indices, const ParameterVector &parameters, const ParameterVector &vector, SensitivityData &product) override
	For each of the system's quantities of interest q in qoi[qoi_indices], and for a vector of parameters p, the parameter sensitivity Hessian H_ij is defined as H_ij = (d^2 q)/(d p_i d p_j) The Hessian-vector product, for a vector v_k in parameter space, is S_j = H_jk v_k This product is the output of this method, where for each q_i, S_j is stored in sensitivities[i][j]. More...
const SparseMatrix< Number > &	get_system_matrix () const
SparseMatrix< Number > &	get_system_matrix ()
StaticCondensation &	get_static_condensation ()
virtual void	assemble_qoi (const QoISet &qoi_indices=QoISet()) override
	Prepares qoi for quantity of interest assembly, then calls user qoi function. More...
virtual void	assemble_qoi_derivative (const QoISet &qoi_indices=QoISet(), bool include_liftfunc=true, bool apply_constraints=true) override
	Prepares adjoint_rhs for quantity of interest derivative assembly, then calls user qoi derivative function. More...
void	init ()
	Initializes degrees of freedom on the current mesh. More...
virtual void	reinit_constraints ()
	Reinitializes the constraints for this system. More...
virtual void	reinit_mesh ()
	Reinitializes the system with a new mesh. More...
bool	is_initialized () const
virtual void	update ()
	Update the local values to reflect the solution on neighboring processors. More...
bool	is_adjoint_already_solved () const
	Accessor for the adjoint_already_solved boolean. More...
void	set_adjoint_already_solved (bool setting)
	Setter for the adjoint_already_solved boolean. More...
virtual void	qoi_parameter_sensitivity (const QoISet &qoi_indices, const ParameterVector &parameters, SensitivityData &sensitivities)
	Solves for the derivative of each of the system's quantities of interest q in qoi[qoi_indices] with respect to each parameter in parameters, placing the result for qoi i and parameter j into sensitivities[i][j]. More...
virtual bool	compare (const System &other_system, const Real threshold, const bool verbose) const
const std::string &	name () const
void	project_solution (FunctionBase< Number > *f, FunctionBase< Gradient > *g=nullptr, std::optional< ConstElemRange > active_local_range=std::nullopt, std::optional< std::vector< unsigned int >> variable_numbers=std::nullopt) const
	Projects arbitrary functions onto the current solution. More...
void	project_solution (FEMFunctionBase< Number > *f, FEMFunctionBase< Gradient > *g=nullptr, std::optional< ConstElemRange > active_local_range=std::nullopt, std::optional< std::vector< unsigned int >> variable_numbers=std::nullopt) const
	Projects arbitrary functions onto the current solution. More...
void	project_solution (ValueFunctionPointer fptr, GradientFunctionPointer gptr, const Parameters &parameters, std::optional< ConstElemRange > active_local_range=std::nullopt, std::optional< std::vector< unsigned int >> variable_numbers=std::nullopt) const
	This method projects an arbitrary function onto the solution via L2 projections and nodal interpolations on each element. More...
void	project_vector (NumericVector< Number > &new_vector, FunctionBase< Number > *f, FunctionBase< Gradient > *g=nullptr, int is_adjoint=-1, std::optional< ConstElemRange > active_local_range=std::nullopt, std::optional< std::vector< unsigned int >> variable_numbers=std::nullopt) const
	Projects arbitrary functions onto a vector of degree of freedom values for the current system. More...
void	project_vector (NumericVector< Number > &new_vector, FEMFunctionBase< Number > *f, FEMFunctionBase< Gradient > *g=nullptr, int is_adjoint=-1, std::optional< ConstElemRange > active_local_range=std::nullopt, std::optional< std::vector< unsigned int >> variable_numbers=std::nullopt) const
	Projects arbitrary functions onto a vector of degree of freedom values for the current system. More...
void	project_vector (ValueFunctionPointer fptr, GradientFunctionPointer gptr, const Parameters &parameters, NumericVector< Number > &new_vector, int is_adjoint=-1, std::optional< ConstElemRange > active_local_range=std::nullopt, std::optional< std::vector< unsigned int >> variable_numbers=std::nullopt) const
	Projects arbitrary functions onto a vector of degree of freedom values for the current system. More...
void	boundary_project_solution (const std::set< boundary_id_type > &b, const std::vector< unsigned int > &variables, FunctionBase< Number > *f, FunctionBase< Gradient > *g=nullptr, std::optional< ConstElemRange > active_local_range=std::nullopt)
	Projects arbitrary boundary functions onto a vector of degree of freedom values for the current system. More...
void	boundary_project_solution (const std::set< boundary_id_type > &b, const std::vector< unsigned int > &variables, ValueFunctionPointer fptr, GradientFunctionPointer gptr, const Parameters &parameters, std::optional< ConstElemRange > active_local_range=std::nullopt)
	Projects arbitrary boundary functions onto a vector of degree of freedom values for the current system. More...
void	boundary_project_vector (const std::set< boundary_id_type > &b, const std::vector< unsigned int > &variables, NumericVector< Number > &new_vector, FunctionBase< Number > *f, FunctionBase< Gradient > *g=nullptr, int is_adjoint=-1, std::optional< ConstElemRange > active_local_range=std::nullopt) const
	Projects arbitrary boundary functions onto a vector of degree of freedom values for the current system. More...
void	boundary_project_vector (const std::set< boundary_id_type > &b, const std::vector< unsigned int > &variables, ValueFunctionPointer fptr, GradientFunctionPointer gptr, const Parameters &parameters, NumericVector< Number > &new_vector, int is_adjoint=-1, std::optional< ConstElemRange > active_local_range=std::nullopt) const
	Projects arbitrary boundary functions onto a vector of degree of freedom values for the current system. More...
unsigned int	number () const
void	update_global_solution (std::vector< Number > &global_soln) const
	Fill the input vector global_soln so that it contains the global solution on all processors. More...
void	update_global_solution (std::vector< Number > &global_soln, const processor_id_type dest_proc) const
	Fill the input vector global_soln so that it contains the global solution on processor dest_proc. More...
const MeshBase &	get_mesh () const
MeshBase &	get_mesh ()
const DofMap &	get_dof_map () const
DofMap &	get_dof_map ()
const EquationSystems &	get_equation_systems () const
EquationSystems &	get_equation_systems ()
bool	active () const
void	activate ()
	Activates the system. More...
void	deactivate ()
	Deactivates the system. More...
void	set_basic_system_only ()
	Sets the system to be "basic only": i.e. More...
vectors_iterator	vectors_begin ()
	Beginning of vectors container. More...
const_vectors_iterator	vectors_begin () const
	Beginning of vectors container. More...
vectors_iterator	vectors_end ()
	End of vectors container. More...
const_vectors_iterator	vectors_end () const
	End of vectors container. More...
matrices_iterator	matrices_begin ()
	Beginning of matrices container. More...
const_matrices_iterator	matrices_begin () const
	Beginning of matrices container. More...
matrices_iterator	matrices_end ()
	End of matrices container. More...
const_matrices_iterator	matrices_end () const
	End of matrices container. More...
NumericVector< Number > &	add_vector (std::string_view vec_name, const bool projections=true, const ParallelType type=PARALLEL)
	Adds the additional vector vec_name to this system. More...
void	remove_vector (std::string_view vec_name)
	Removes the additional vector vec_name from this system. More...
bool &	project_solution_on_reinit (void)
	Tells the System whether or not to project the solution vector onto new grids when the system is reinitialized. More...
bool	have_vector (std::string_view vec_name) const
const NumericVector< Number > *	request_vector (std::string_view vec_name) const
NumericVector< Number > *	request_vector (std::string_view vec_name)
const NumericVector< Number > *	request_vector (const unsigned int vec_num) const
NumericVector< Number > *	request_vector (const unsigned int vec_num)
const NumericVector< Number > &	get_vector (std::string_view vec_name) const
NumericVector< Number > &	get_vector (std::string_view vec_name)
const NumericVector< Number > &	get_vector (const unsigned int vec_num) const
NumericVector< Number > &	get_vector (const unsigned int vec_num)
const std::string &	vector_name (const unsigned int vec_num) const
const std::string &	vector_name (const NumericVector< Number > &vec_reference) const
void	set_vector_as_adjoint (const std::string &vec_name, int qoi_num)
	Allows one to set the QoI index controlling whether the vector identified by vec_name represents a solution from the adjoint (qoi_num >= 0) or primal (qoi_num == -1) space. More...
int	vector_is_adjoint (std::string_view vec_name) const
void	set_vector_preservation (const std::string &vec_name, bool preserve)
	Allows one to set the boolean controlling whether the vector identified by vec_name should be "preserved": projected to new meshes, saved, etc. More...
bool	vector_preservation (std::string_view vec_name) const
NumericVector< Number > &	add_adjoint_solution (unsigned int i=0)
NumericVector< Number > &	get_adjoint_solution (unsigned int i=0)
const NumericVector< Number > &	get_adjoint_solution (unsigned int i=0) const
NumericVector< Number > &	add_sensitivity_solution (unsigned int i=0)
NumericVector< Number > &	get_sensitivity_solution (unsigned int i=0)
const NumericVector< Number > &	get_sensitivity_solution (unsigned int i=0) const
NumericVector< Number > &	add_weighted_sensitivity_adjoint_solution (unsigned int i=0)
NumericVector< Number > &	get_weighted_sensitivity_adjoint_solution (unsigned int i=0)
const NumericVector< Number > &	get_weighted_sensitivity_adjoint_solution (unsigned int i=0) const
NumericVector< Number > &	add_weighted_sensitivity_solution ()
NumericVector< Number > &	get_weighted_sensitivity_solution ()
const NumericVector< Number > &	get_weighted_sensitivity_solution () const
NumericVector< Number > &	add_adjoint_rhs (unsigned int i=0)
NumericVector< Number > &	get_adjoint_rhs (unsigned int i=0)
const NumericVector< Number > &	get_adjoint_rhs (unsigned int i=0) const
NumericVector< Number > &	add_sensitivity_rhs (unsigned int i=0)
NumericVector< Number > &	get_sensitivity_rhs (unsigned int i=0)
const NumericVector< Number > &	get_sensitivity_rhs (unsigned int i=0) const
unsigned int	n_vectors () const
unsigned int	n_matrices () const
unsigned int	n_vars () const
unsigned int	n_variable_groups () const
unsigned int	n_components () const
dof_id_type	n_dofs () const
dof_id_type	n_active_dofs () const
dof_id_type	n_constrained_dofs () const
dof_id_type	n_local_constrained_dofs () const
dof_id_type	n_local_dofs () const
unsigned int	add_variable (std::string_view var, const FEType &type, const std::set< subdomain_id_type > *const active_subdomains=nullptr)
	Adds the variable var to the list of variables for this system. More...
unsigned int	add_variable (std::string_view var, const Order order=FIRST, const FEFamily=LAGRANGE, const std::set< subdomain_id_type > *const active_subdomains=nullptr, const bool p_refinement=true)
	Adds the variable var to the list of variables for this system. More...
unsigned int	add_variables (const std::vector< std::string > &vars, const FEType &type, const std::set< subdomain_id_type > *const active_subdomains=nullptr)
	Adds the variables vars to the list of variables for this system. More...
unsigned int	add_variables (const std::vector< std::string > &vars, const Order order=FIRST, const FEFamily=LAGRANGE, const std::set< subdomain_id_type > *const active_subdomains=nullptr, const bool p_refinement=true)
	Adds the variable var to the list of variables for this system. More...
unsigned int	add_variable_array (const std::vector< std::string > &vars, const FEType &type, const std::set< subdomain_id_type > *const active_subdomains=nullptr)
	Adds variables vars to the list of variables for this system. More...
const Variable &	variable (unsigned int var) const
	Return a constant reference to Variable var. More...
const VariableGroup &	variable_group (unsigned int vg) const
	Return a constant reference to VariableGroup vg. More...
bool	has_variable (std::string_view var) const
const std::string &	variable_name (const unsigned int i) const
unsigned int	variable_number (std::string_view var) const
void	get_all_variable_numbers (std::vector< unsigned int > &all_variable_numbers) const
	Fills all_variable_numbers with all the variable numbers for the variables that have been added to this system. More...
unsigned int	variable_scalar_number (std::string_view var, unsigned int component) const
unsigned int	variable_scalar_number (unsigned int var_num, unsigned int component) const
const FEType &	variable_type (const unsigned int i) const
const FEType &	variable_type (std::string_view var) const
bool	identify_variable_groups () const
void	identify_variable_groups (const bool)
	Toggle automatic VariableGroup identification. More...
Real	calculate_norm (const NumericVector< Number > &v, unsigned int var, FEMNormType norm_type, std::set< unsigned int > *skip_dimensions=nullptr) const
Real	calculate_norm (const NumericVector< Number > &v, const SystemNorm &norm, std::set< unsigned int > *skip_dimensions=nullptr) const
void	read_header (Xdr &io, std::string_view version, const bool read_header=true, const bool read_additional_data=true, const bool read_legacy_format=false)
	Reads the basic data header for this System. More...
template<typename ValType >
void	read_serialized_data (Xdr &io, const bool read_additional_data=true)
	Reads additional data, namely vectors, for this System. More...
void	read_serialized_data (Xdr &io, const bool read_additional_data=true)
	Non-templated version for backward compatibility. More...
template<typename InValType >
std::size_t	read_serialized_vectors (Xdr &io, const std::vector< NumericVector< Number > *> &vectors) const
	Read a number of identically distributed vectors. More...
std::size_t	read_serialized_vectors (Xdr &io, const std::vector< NumericVector< Number > *> &vectors) const
	Non-templated version for backward compatibility. More...
template<typename InValType >
void	read_parallel_data (Xdr &io, const bool read_additional_data)
	Reads additional data, namely vectors, for this System. More...
void	read_parallel_data (Xdr &io, const bool read_additional_data)
	Non-templated version for backward compatibility. More...
void	write_header (Xdr &io, std::string_view version, const bool write_additional_data) const
	Writes the basic data header for this System. More...
void	write_serialized_data (Xdr &io, const bool write_additional_data=true) const
	Writes additional data, namely vectors, for this System. More...
std::size_t	write_serialized_vectors (Xdr &io, const std::vector< const NumericVector< Number > *> &vectors) const
	Serialize & write a number of identically distributed vectors. More...
void	write_parallel_data (Xdr &io, const bool write_additional_data) const
	Writes additional data, namely vectors, for this System. More...
std::string	get_info () const
void	attach_init_function (void fptr(EquationSystems &es, const std::string &name))
	Register a user function to use in initializing the system. More...
void	attach_init_object (Initialization &init)
	Register a user class to use to initialize the system. More...
void	attach_assemble_function (void fptr(EquationSystems &es, const std::string &name))
	Register a user function to use in assembling the system matrix and RHS. More...
void	attach_assemble_object (Assembly &assemble)
	Register a user object to use in assembling the system matrix and RHS. More...
void	attach_constraint_function (void fptr(EquationSystems &es, const std::string &name))
	Register a user function for imposing constraints. More...
void	attach_constraint_object (Constraint &constrain)
	Register a user object for imposing constraints. More...
bool	has_constraint_object () const
Constraint &	get_constraint_object ()
	Return the user object for imposing constraints. More...
void	attach_QOI_function (void fptr(EquationSystems &es, const std::string &name, const QoISet &qoi_indices))
	Register a user function for evaluating the quantities of interest, whose values should be placed in System::qoi. More...
void	attach_QOI_object (QOI &qoi)
	Register a user object for evaluating the quantities of interest, whose values should be placed in System::qoi. More...
void	attach_QOI_derivative (void fptr(EquationSystems &es, const std::string &name, const QoISet &qoi_indices, bool include_liftfunc, bool apply_constraints))
	Register a user function for evaluating derivatives of a quantity of interest with respect to test functions, whose values should be placed in System::rhs. More...
void	attach_QOI_derivative_object (QOIDerivative &qoi_derivative)
	Register a user object for evaluating derivatives of a quantity of interest with respect to test functions, whose values should be placed in System::rhs. More...
virtual void	user_initialization ()
	Calls user's attached initialization function, or is overridden by the user in derived classes. More...
virtual void	user_assembly ()
	Calls user's attached assembly function, or is overridden by the user in derived classes. More...
virtual void	user_constrain ()
	Calls user's attached constraint function, or is overridden by the user in derived classes. More...
virtual void	user_QOI (const QoISet &qoi_indices)
	Calls user's attached quantity of interest function, or is overridden by the user in derived classes. More...
virtual void	user_QOI_derivative (const QoISet &qoi_indices=QoISet(), bool include_liftfunc=true, bool apply_constraints=true)
	Calls user's attached quantity of interest derivative function, or is overridden by the user in derived classes. More...
virtual void	re_update ()
	Re-update the local values when the mesh has changed. More...
virtual void	restrict_vectors ()
	Restrict vectors after the mesh has coarsened. More...
virtual void	prolong_vectors ()
	Prolong vectors after the mesh has refined. More...
Number	current_solution (const dof_id_type global_dof_number) const
unsigned int	n_qois () const
	Number of currently active quantities of interest. More...
void	init_qois (unsigned int n_qois)
	Accessors for qoi and qoi_error_estimates vectors. More...
void	set_qoi (unsigned int qoi_index, Number qoi_value)
void	set_qoi (std::vector< Number > new_qoi)
Number	get_qoi_value (unsigned int qoi_index) const
std::vector< Number >	get_qoi_values () const
	Returns a copy of qoi, not a reference. More...
void	set_qoi_error_estimate (unsigned int qoi_index, Number qoi_error_estimate)
Number	get_qoi_error_estimate_value (unsigned int qoi_index) const
Number	point_value (unsigned int var, const Point &p, const bool insist_on_success=true, const NumericVector< Number > *sol=nullptr) const
Number	point_value (unsigned int var, const Point &p, const Elem &e, const NumericVector< Number > *sol=nullptr) const
Number	point_value (unsigned int var, const Point &p, const Elem *e) const
	Calls the version of point_value() which takes a reference. More...
Number	point_value (unsigned int var, const Point &p, const NumericVector< Number > *sol) const
	Calls the parallel version of point_value(). More...
Gradient	point_gradient (unsigned int var, const Point &p, const bool insist_on_success=true, const NumericVector< Number > *sol=nullptr) const
Gradient	point_gradient (unsigned int var, const Point &p, const Elem &e, const NumericVector< Number > *sol=nullptr) const
Gradient	point_gradient (unsigned int var, const Point &p, const Elem *e) const
	Calls the version of point_gradient() which takes a reference. More...
Gradient	point_gradient (unsigned int var, const Point &p, const NumericVector< Number > *sol) const
	Calls the parallel version of point_gradient(). More...
Tensor	point_hessian (unsigned int var, const Point &p, const bool insist_on_success=true, const NumericVector< Number > *sol=nullptr) const
Tensor	point_hessian (unsigned int var, const Point &p, const Elem &e, const NumericVector< Number > *sol=nullptr) const
Tensor	point_hessian (unsigned int var, const Point &p, const Elem *e) const
	Calls the version of point_hessian() which takes a reference. More...
Tensor	point_hessian (unsigned int var, const Point &p, const NumericVector< Number > *sol) const
	Calls the parallel version of point_hessian(). More...
void	local_dof_indices (const unsigned int var, std::set< dof_id_type > &var_indices) const
	Fills the std::set with the degrees of freedom on the local processor corresponding the the variable number passed in. More...
void	zero_variable (NumericVector< Number > &v, unsigned int var_num) const
	Zeroes all dofs in v that correspond to variable number var_num. More...
bool	get_project_with_constraints ()
	Setter and getter functions for project_with_constraints boolean. More...
void	set_project_with_constraints (bool _project_with_constraints)
bool &	hide_output ()
void	projection_matrix (SparseMatrix< Number > &proj_mat) const
	This method creates a projection matrix which corresponds to the operation of project_vector between old and new solution spaces. More...
SparseMatrix< Number > &	add_matrix (std::string_view mat_name, ParallelType type=PARALLEL, MatrixBuildType mat_build_type=MatrixBuildType::AUTOMATIC)
	Adds the additional matrix mat_name to this system. More...
template<template< typename > class>
SparseMatrix< Number > &	add_matrix (std::string_view mat_name, ParallelType=PARALLEL)
	Adds the additional matrix mat_name to this system. More...
SparseMatrix< Number > &	add_matrix (std::string_view mat_name, std::unique_ptr< SparseMatrix< Number >> matrix, ParallelType type=PARALLEL)
	Adds the additional matrix mat_name to this system. More...
void	remove_matrix (std::string_view mat_name)
	Removes the additional matrix mat_name from this system. More...
bool	have_matrix (std::string_view mat_name) const
const SparseMatrix< Number > *	request_matrix (std::string_view mat_name) const
SparseMatrix< Number > *	request_matrix (std::string_view mat_name)
const SparseMatrix< Number > &	get_matrix (std::string_view mat_name) const
SparseMatrix< Number > &	get_matrix (std::string_view mat_name)
void	prefer_hash_table_matrix_assembly (bool preference)
	Sets whether to use hash table matrix assembly if the matrix sub-classes support it. More...
void	prefix_with_name (bool value)
	Instructs this system to prefix solve options with its name for solvers that leverage prefixes. More...
bool	prefix_with_name () const
std::string	prefix () const
bool	has_static_condensation () const
void	solve_for_unconstrained_dofs (NumericVector< Number > &, int is_adjoint=-1) const
const Parallel::Communicator &	comm () const
processor_id_type	n_processors () const
processor_id_type	processor_id () const
void	initialize_parameters (const RBParameters &mu_min_in, const RBParameters &mu_max_in, const std::map< std::string, std::vector< Real >> &discrete_parameter_values)
	Initialize the parameter ranges and set current_parameters. More...
void	initialize_parameters (const RBParametrized &rb_parametrized)
	Initialize the parameter ranges and set current_parameters. More...
unsigned int	get_n_params () const
	Get the number of parameters. More...
unsigned int	get_n_continuous_params () const
	Get the number of continuous parameters. More...
unsigned int	get_n_discrete_params () const
	Get the number of discrete parameters. More...
const RBParameters &	get_parameters () const
	Get the current parameters. More...
bool	set_parameters (const RBParameters &params)
	Set the current parameters to params The parameters are checked for validity; an error is thrown if the number of parameters or samples is different than expected. More...
const RBParameters &	get_parameters_min () const
	Get an RBParameters object that specifies the minimum allowable value for each parameter. More...
const RBParameters &	get_parameters_max () const
	Get an RBParameters object that specifies the maximum allowable value for each parameter. More...
Real	get_parameter_min (const std::string &param_name) const
	Get minimum allowable value of parameter param_name. More...
Real	get_parameter_max (const std::string &param_name) const
	Get maximum allowable value of parameter param_name. More...
void	print_parameters () const
	Print the current parameters. More...
void	write_parameter_data_to_files (const std::string &continuous_param_file_name, const std::string &discrete_param_file_name, const bool write_binary_data)
	Write out the parameter ranges to files. More...
void	read_parameter_data_from_files (const std::string &continuous_param_file_name, const std::string &discrete_param_file_name, const bool read_binary_data)
	Read in the parameter ranges from files. More...
bool	is_discrete_parameter (const std::string &mu_name) const
	Is parameter mu_name discrete? More...
const std::map< std::string, std::vector< Real > > &	get_discrete_parameter_values () const
	Get a const reference to the discrete parameter values. More...
void	print_discrete_parameter_values () const
	Print out all the discrete parameter values. More...

Static Public Member Functions
static std::unique_ptr< DirichletBoundary >	build_zero_dirichlet_boundary_object ()
	It's helpful to be able to generate a DirichletBoundary that stores a ZeroFunction in order to impose Dirichlet boundary conditions. More...
static std::pair< std::size_t, std::size_t >	generate_training_parameters_random (const Parallel::Communicator &communicator, const std::map< std::string, bool > &log_param_scale, std::map< std::string, std::vector< RBParameter >> &local_training_parameters_in, const unsigned int n_global_training_samples_in, const RBParameters &min_parameters, const RBParameters &max_parameters, const int training_parameters_random_seed=-1, const bool serial_training_set=false)
	Static helper function for generating a randomized set of parameters. More...
static std::pair< std::size_t, std::size_t >	generate_training_parameters_deterministic (const Parallel::Communicator &communicator, const std::map< std::string, bool > &log_param_scale, std::map< std::string, std::vector< RBParameter >> &local_training_parameters_in, const unsigned int n_global_training_samples_in, const RBParameters &min_parameters, const RBParameters &max_parameters, const bool serial_training_set=false)
	Static helper function for generating a deterministic set of parameters. More...
static std::string	get_info ()
	Gets a string containing the reference information. More...
static std::string	get_info ()
	Gets a string containing the reference information. More...
static void	print_info (std::ostream &out_stream=libMesh::out)
	Prints the reference information, by default to libMesh::out. More...
static void	print_info (std::ostream &out_stream=libMesh::out)
	Prints the reference information, by default to libMesh::out. More...
static unsigned int	n_objects ()
	Prints the number of outstanding (created, but not yet destroyed) objects. More...
static unsigned int	n_objects ()
	Prints the number of outstanding (created, but not yet destroyed) objects. More...
static void	enable_print_counter_info ()
	Methods to enable/disable the reference counter output from print_info() More...
static void	enable_print_counter_info ()
	Methods to enable/disable the reference counter output from print_info() More...
static void	disable_print_counter_info ()
static void	disable_print_counter_info ()
static Real	get_closest_value (Real value, const std::vector< Real > &list_of_values)

Public Attributes
std::vector< Real >	training_error_bounds
	Vector storing the values of the error bound for each parameter in the training set — the parameter giving the largest error bound is chosen for the next snapshot in the Greedy basis training. More...
std::unique_ptr< LinearSolver< Number > >	inner_product_solver
	We store an extra linear solver object which we can optionally use for solving all systems in which the system matrix is set to inner_product_matrix. More...
LinearSolver< Number > *	extra_linear_solver
	Also, we store a pointer to an extra linear solver. More...
std::unique_ptr< SparseMatrix< Number > >	inner_product_matrix
	The inner product matrix. More...
std::vector< Number >	truth_outputs
	Vector storing the truth output values from the most recent truth solve. More...
std::vector< std::vector< Number > >	output_dual_innerprods
	The vector storing the dual norm inner product terms for each output. More...
std::vector< std::unique_ptr< NumericVector< Number > > >	Fq_representor
	Vector storing the residual representors associated with the right-hand side. More...
std::vector< Number >	Fq_representor_innerprods
	Vectors storing the residual representor inner products to be used in computing the residuals online. More...
bool	skip_residual_in_train_reduced_basis
	Boolean flag to indicate if we skip residual calculations in train_reduced_basis. More...
bool	exit_on_repeated_greedy_parameters
	Boolean flag to indicate whether we exit the greedy if we select the same parameters twice in a row. More...
bool	impose_internal_fluxes
	Boolean flag to indicate whether we impose "fluxes" (i.e. More...
bool	skip_degenerate_sides
	In some cases meshes are intentionally created with degenerate sides as a way to represent, say, triangles using a hex-only mesh. More...
bool	compute_RB_inner_product
	Boolean flag to indicate whether we compute the RB_inner_product_matrix. More...
bool	store_dirichlet_operators
	Boolean flag to indicate whether we store affine operator matrices and vectors with constraints enforced. More...
bool	store_non_dirichlet_operators
	Boolean flag to indicate whether we store a second copy of each affine operator and vector which does not have Dirichlet bcs enforced. More...
bool	store_untransformed_basis
	Boolean flag to indicate whether we store a second copy of the basis without constraints or dof transformations applied to it. More...
bool	use_empty_rb_solve_in_greedy
	A boolean flag to indicate whether or not we initialize the Greedy algorithm by performing rb_solves on the training set with an "empty" (i.e. More...
bool	Fq_representor_innerprods_computed
	A boolean flag to indicate whether or not the Fq representor norms have already been computed — used to make sure that we don't recompute them unnecessarily. More...
SparseMatrix< Number > *	matrix
	The system matrix. More...
bool	zero_out_matrix_and_rhs
	By default, the system will zero out the matrix and the right hand side. More...
std::unique_ptr< LinearSolver< Number > >	linear_solver
	This class handles all the details of interfacing with various linear algebra packages like PETSc or LASPACK. More...
NumericVector< Number > *	rhs
	The system matrix. More...
Parameters	parameters
	Parameters for the system. If a parameter is not provided, it should be retrieved from the EquationSystems. More...
bool	assemble_before_solve
	Flag which tells the system to whether or not to call the user assembly function during each call to solve(). More...
bool	use_fixed_solution
	A boolean to be set to true by systems using elem_fixed_solution, for optional use by e.g. More...
int	extra_quadrature_order
	A member int that can be employed to indicate increased or reduced quadrature order. More...
std::unique_ptr< NumericVector< Number > >	solution
	Data structure to hold solution values. More...
std::unique_ptr< NumericVector< Number > >	current_local_solution
	All the values I need to compute my contribution to the simulation at hand. More...
Real	time
	For time-dependent problems, this is the time t at the beginning of the current timestep. More...
bool	verbose_mode
	Public boolean to toggle verbose mode. More...

Protected Types
typedef std::map< std::string, std::pair< unsigned int, unsigned int > >	Counts
	Data structure to log the information. More...
typedef std::map< std::string, std::pair< unsigned int, unsigned int > >	Counts
	Data structure to log the information. More...

Protected Member Functions
virtual void	allocate_data_structures ()
	Helper function that actually allocates all the data structures required by this class. More...
virtual void	truth_assembly ()
	Assemble the truth matrix and right-hand side for current_parameters. More...
virtual std::unique_ptr< DGFEMContext >	build_context ()
	Builds a DGFEMContext object with enough information to do evaluations on each element. More...
virtual SparseMatrix< Number > &	get_matrix_for_output_dual_solves ()
	Return the matrix for the output residual dual norm solves. More...
virtual bool	greedy_termination_test (Real abs_greedy_error, Real initial_greedy_error, int count)
	Function that indicates when to terminate the Greedy basis training. More...
void	update_greedy_param_list ()
	Update the list of Greedily chosen parameters with current_parameters. More...
void	add_scaled_matrix_and_vector (Number scalar, ElemAssembly *elem_assembly, SparseMatrix< Number > *input_matrix, NumericVector< Number > *input_vector, bool symmetrize=false, bool apply_dof_constraints=true)
	This function loops over the mesh and applies the specified interior and/or boundary assembly routines, then adds the scaled result to input_matrix and/or input_vector. More...
virtual void	post_process_elem_matrix_and_vector (DGFEMContext &)
	This function is called from add_scaled_matrix_and_vector() before each element matrix and vector are assembled into their global counterparts. More...
virtual void	post_process_truth_solution ()
	Similarly, provide an opportunity to post-process the truth solution after the solve is complete. More...
virtual void	set_context_solution_vec (NumericVector< Number > &vec)
	Set current_local_solution = vec so that we can access vec from FEMContext during assembly. More...
virtual void	assemble_misc_matrices ()
	Assemble and store all the inner-product matrix, the constraint matrix (for constrained problems) and the mass matrix (for time-dependent problems). More...
virtual void	assemble_all_affine_operators ()
	Assemble and store all Q_a affine operators as well as the inner-product matrix. More...
virtual void	assemble_all_affine_vectors ()
	Assemble and store the affine RHS vectors. More...
virtual void	assemble_all_output_vectors ()
	Assemble and store the output vectors. More...
virtual void	compute_output_dual_innerprods ()
	Compute and store the dual norm of each output functional. More...
virtual void	compute_Fq_representor_innerprods (bool compute_inner_products=true)
	Compute the terms that are combined ‘online’ to determine the dual norm of the residual. More...
virtual void	enrich_RB_space ()
	Add a new basis function to the RB space. More...
virtual void	update_system ()
	Update the system after enriching the RB space; this calls a series of functions to update the system properly. More...
virtual Real	get_RB_error_bound ()
virtual void	update_RB_system_matrices ()
	Compute the reduced basis matrices for the current basis. More...
virtual void	update_residual_terms (bool compute_inner_products=true)
	Compute the terms that are combined ‘online’ to determine the dual norm of the residual. More...
virtual void	init_context (FEMContext &)
	Initialize the FEMContext prior to performing an element loop. More...
bool	get_convergence_assertion_flag () const
	Getter for the flag determining if convergence should be checked after each solve. More...
void	check_convergence (LinearSolver< Number > &input_solver)
	Check if the linear solver reports convergence. More...
unsigned int	get_current_training_parameter_index () const
	Get/set the current training parameter index. More...
void	set_current_training_parameter_index (unsigned int index)
const std::vector< Number > &	get_evaluated_thetas (unsigned int training_parameter_index) const
	Return the evaluated theta functions at the given training parameter index. More...
virtual void	preevaluate_thetas ()
void	reset_preevaluate_thetas_completed ()
	Reset the _preevaluate_thetas_completed flag to false. More...
virtual void	init_data ()
	Initializes the member data fields associated with the system, so that, e.g., assemble() may be used. More...
RBParameters	get_params_from_training_set (unsigned int global_index)
	Return the RBParameters in index global_index of the global training set. More...
void	set_params_from_training_set (unsigned int global_index)
	Set parameters to the RBParameters stored in index global_index of the global training set. More...
virtual void	set_params_from_training_set_and_broadcast (unsigned int global_index)
	Load the specified training parameter and then broadcast to all processors. More...
virtual void	add_matrices () override
	Adds the system matrix. More...
template<typename T >
void	setup_static_condensation_preconditioner (T &solver)
	Sets up the static condensation preconditioner for the supplied solver. More...
void	project_vector (NumericVector< Number > &, int is_adjoint=-1, std::optional< ConstElemRange > active_local_range=std::nullopt, std::optional< std::vector< unsigned int >> variable_numbers=std::nullopt) const
	Projects the vector defined on the old mesh onto the new mesh. More...
void	project_vector (const NumericVector< Number > &, NumericVector< Number > &, int is_adjoint=-1, std::optional< ConstElemRange > active_local_range=std::nullopt, std::optional< std::vector< unsigned int >> variable_numbers=std::nullopt) const
	Projects the vector defined on the old mesh onto the new mesh. More...
virtual void	init_matrices ()
	Initializes the matrices associated with this system. More...
bool	can_add_matrices () const
virtual bool	condense_constrained_dofs () const
	Whether this object should condense out constrained degrees of freedom. More...
void	increment_constructor_count (const std::string &name) noexcept
	Increments the construction counter. More...
void	increment_constructor_count (const std::string &name) noexcept
	Increments the construction counter. More...
void	increment_destructor_count (const std::string &name) noexcept
	Increments the destruction counter. More...
void	increment_destructor_count (const std::string &name) noexcept
	Increments the destruction counter. More...

Static Protected Member Functions
static void	get_global_max_error_pair (const Parallel::Communicator &communicator, std::pair< numeric_index_type, Real > &error_pair)
	Static function to return the error pair (index,error) that is corresponds to the largest error on all processors. More...

Protected Attributes
unsigned int	Nmax
	Maximum number of reduced basis functions we are willing to use. More...
unsigned int	delta_N
	The number of basis functions that we add at each greedy step. More...
bool	output_dual_innerprods_computed
	A boolean flag to indicate whether or not the output dual norms have already been computed — used to make sure that we don't recompute them unnecessarily. More...
bool	assert_convergence
	A boolean flag to indicate whether to check for proper convergence after each solve. More...
bool	quiet_mode
	Flag to indicate whether we print out extra information during the Offline stage. More...
bool	serial_training_set
	This boolean flag indicates whether or not the training set should be the same on all processors. More...
bool	_normalize_solution_snapshots
	Set this boolean to true if we want to normalize solution snapshots used in training to have norm of 1. More...
std::unique_ptr< NumericVector< Number > >	inner_product_storage_vector
	We keep an extra temporary vector that is useful for performing inner products (avoids unnecessary memory allocation/deallocation). More...
unsigned int	_n_linear_iterations
	The number of linear iterations required to solve the linear system Ax=b. More...
Real	_final_linear_residual
	The final residual for the linear system Ax=b. More...
ShellMatrix< Number > *	_shell_matrix
	User supplies shell matrix or nullptr if no shell matrix is used. More...
const SystemSubset *	_subset
	The current subset on which to solve (or nullptr if none). More...
SubsetSolveMode	_subset_solve_mode
	If restrict-solve-to-subset mode is active, this member decides what happens with the dofs outside the subset. More...
const Parallel::Communicator &	_communicator

Static Protected Attributes
static Counts	_counts
	Actually holds the data. More...
static Counts	_counts
	Actually holds the data. More...
static Threads::atomic< unsigned int >	_n_objects
	The number of objects. More...
static Threads::atomic< unsigned int >	_n_objects
	The number of objects. More...
static Threads::spin_mutex	_mutex
	Mutual exclusion object to enable thread-safe reference counting. More...
static Threads::spin_mutex	_mutex
	Mutual exclusion object to enable thread-safe reference counting. More...
static bool	_enable_print_counter = true
	Flag to control whether reference count information is printed when print_info is called. More...
static bool	_enable_print_counter = true
	Flag to control whether reference count information is printed when print_info is called. More...

Private Attributes
RBEvaluation *	rb_eval
	The current RBEvaluation object we are using to perform the Evaluation stage of the reduced basis method. More...
RBAssemblyExpansion *	rb_assembly_expansion
	This member holds the (parameter independent) assembly functors that define the "affine expansion" of the PDE that we are solving. More...
ElemAssembly *	inner_product_assembly
	Pointer to inner product assembly. More...
bool	use_energy_inner_product
	Boolean to indicate whether we're using the energy inner-product. More...
std::vector< Number >	energy_inner_product_coeffs
	We may optionally want to use the "energy inner-product" rather than the inner-product assembly specified in inner_product_assembly. More...
std::vector< std::unique_ptr< SparseMatrix< Number > > >	Aq_vector
	Vector storing the Q_a matrices from the affine expansion. More...
std::vector< std::unique_ptr< NumericVector< Number > > >	Fq_vector
	Vector storing the Q_f vectors in the affine decomposition of the right-hand side. More...
std::vector< std::vector< std::unique_ptr< NumericVector< Number > > > >	outputs_vector
	The libMesh vectors that define the output functionals. More...
std::vector< std::unique_ptr< SparseMatrix< Number > > >	non_dirichlet_Aq_vector
	We may also need a second set of matrices/vectors that do not have the Dirichlet boundary conditions enforced. More...
std::vector< std::unique_ptr< NumericVector< Number > > >	non_dirichlet_Fq_vector
std::vector< std::vector< std::unique_ptr< NumericVector< Number > > > >	non_dirichlet_outputs_vector
std::unique_ptr< SparseMatrix< Number > >	non_dirichlet_inner_product_matrix
Real	rel_training_tolerance
	Relative and absolute tolerances for training reduced basis using the Greedy scheme. More...
Real	abs_training_tolerance
bool	normalize_rb_bound_in_greedy
	This boolean indicates if we normalize the RB error in the greedy using RBEvaluation::get_error_bound_normalization(). More...
std::string	RB_training_type
	This string indicates the type of training that we will use. More...
std::vector< std::unique_ptr< NumericVector< Number > > >	_untransformed_basis_functions
	In cases where we have dof transformations such as a change of coordinates at some nodes we need to store an extra set of basis functions which have not had dof transformations applied to them. More...
std::unique_ptr< NumericVector< Number > >	_untransformed_solution
	We also store a copy of the untransformed solution in order to create _untransformed_basis_functions. More...
bool	_preevaluate_thetas_flag
	Flag to indicate if we preevaluate the theta functions. More...
bool	_preevaluate_thetas_completed
	Flag to indicate if the preevaluate_thetas function has been called, since this allows us to avoid calling preevaluate_thetas more than once, which is typically unnecessary. More...
unsigned int	_current_training_parameter_index
	The current training parameter index during reduced basis training. More...
std::vector< std::vector< Number > >	_evaluated_thetas
	Storage of evaluated theta functions at a set of parameters. More...

Detailed Description

This class is part of the rbOOmit framework.

RBConstruction implements the Construction stage of the certified reduced basis method for steady-state elliptic parametrized PDEs.

Author

David J. Knezevic

Date

2009

Definition at line 53 of file rb_construction.h.

Member Typedef Documentation

const_matrices_iterator

typedef std::map<std::string, std::unique_ptr<SparseMatrix<Number> >, std::less<> >::const_iterator libMesh::System::const_matrices_iterator

inherited

Definition at line 830 of file system.h.

const_vectors_iterator

typedef std::map<std::string, std::unique_ptr<NumericVector<Number> >, std::less<> >::const_iterator libMesh::System::const_vectors_iterator

inherited

Definition at line 804 of file system.h.

Counts [1/2]

typedef std::map<std::string, std::pair<unsigned int, unsigned int> > libMesh::ReferenceCounter::Counts

protectedinherited

Data structure to log the information.

The log is identified by the class name.

Definition at line 119 of file reference_counter.h.

Counts [2/2]

typedef std::map<std::string, std::pair<unsigned int, unsigned int> > libMesh::ReferenceCounter::Counts

protectedinherited

Data structure to log the information.

The log is identified by the class name.

Definition at line 119 of file reference_counter.h.

GradientFunctionPointer

typedef Gradient(* libMesh::System::GradientFunctionPointer) (const Point &p, const Parameters &parameters, const std::string &sys_name, const std::string &unknown_name)

inherited

Definition at line 555 of file system.h.

matrices_iterator

typedef std::map<std::string, std::unique_ptr<SparseMatrix<Number> >, std::less<> >::iterator libMesh::System::matrices_iterator

inherited

Matrix iterator typedefs.

Definition at line 829 of file system.h.

Parent

typedef RBConstructionBase<LinearImplicitSystem> libMesh::RBConstruction::Parent

The type of the parent.

Definition at line 130 of file rb_construction.h.

sys_type

typedef RBConstruction libMesh::RBConstruction::sys_type

The type of system.

Definition at line 79 of file rb_construction.h.

ValueFunctionPointer

typedef Number(* libMesh::System::ValueFunctionPointer) (const Point &p, const Parameters &Parameters, const std::string &sys_name, const std::string &unknown_name)

inherited

Projects arbitrary functions onto the current solution.

The function value fptr and its gradient gptr are represented by function pointers. A gradient gptr is only required/used for projecting onto finite element spaces with continuous derivatives.

Definition at line 551 of file system.h.

vectors_iterator

typedef std::map<std::string, std::unique_ptr<NumericVector<Number> >, std::less<> >::iterator libMesh::System::vectors_iterator

inherited

Vector iterator typedefs.

Definition at line 803 of file system.h.

Constructor & Destructor Documentation

RBConstruction() [1/3]

libMesh::RBConstruction::RBConstruction	(	EquationSystems &	es,
		const std::string &	name,
		const unsigned int	number
	)

Constructor.

Optionally initializes required data structures.

Definition at line 66 of file rb_construction.C.

References libMesh::System::assemble_before_solve.

  : Parent(es, name_in, number_in),
    inner_product_solver(LinearSolver<Number>::build(es.comm())),
    extra_linear_solver(nullptr),
    inner_product_matrix(SparseMatrix<Number>::build(es.comm())),
    skip_residual_in_train_reduced_basis(false),
    exit_on_repeated_greedy_parameters(true),
    impose_internal_fluxes(false),
    skip_degenerate_sides(true),
    compute_RB_inner_product(false),
    store_dirichlet_operators(true),
    store_non_dirichlet_operators(false),
    store_untransformed_basis(false),
    use_empty_rb_solve_in_greedy(true),
    Fq_representor_innerprods_computed(false),
    Nmax(0),
    delta_N(1),
    output_dual_innerprods_computed(false),
    assert_convergence(true),
    rb_eval(nullptr),
    inner_product_assembly(nullptr),
    use_energy_inner_product(false),
    rel_training_tolerance(1.e-4),
    abs_training_tolerance(1.e-12),
    normalize_rb_bound_in_greedy(false),
    RB_training_type("Greedy"),
    _preevaluate_thetas_flag(false),
    _preevaluate_thetas_completed(false)
{
  // set assemble_before_solve flag to false
  // so that we control matrix assembly.
  assemble_before_solve = false;
}

RBConstruction() [2/3]

libMesh::RBConstruction::RBConstruction ( RBConstruction &&  )

default

Special functions.

• This class has the same restrictions/defaults as its base class.
• Destructor is defaulted out-of-line

RBConstruction() [3/3]

libMesh::RBConstruction::RBConstruction ( const RBConstruction &  )

delete

~RBConstruction()

libMesh::RBConstruction::~RBConstruction ( )

virtualdefault

Member Function Documentation

activate()

void libMesh::System::activate ( )

inlineinherited

Activates the system.

Only active systems are solved.

Definition at line 2441 of file system.h.

References libMesh::System::_active.

{
  _active = true;
}

active()

bool libMesh::System::active ( ) const

inlineinherited

Returns

true if the system is active, false otherwise. An active system will be solved.

Definition at line 2433 of file system.h.

References libMesh::System::_active.

{
  return _active;
}

add_adjoint_rhs()

NumericVector< Number > & libMesh::System::add_adjoint_rhs ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's adjoint rhs vectors, by default the one corresponding to the first qoi. Creates the vector if it doesn't already exist.

Definition at line 1284 of file system.C.

References libMesh::System::add_vector().

Referenced by libMesh::ExplicitSystem::assemble_qoi_derivative(), and libMesh::FEMSystem::assemble_qoi_derivative().

{
  std::ostringstream adjoint_rhs_name;
  adjoint_rhs_name << "adjoint_rhs" << i;

  return this->add_vector(adjoint_rhs_name.str(), false);
}

add_adjoint_solution()

NumericVector< Number > & libMesh::System::add_adjoint_solution ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's adjoint solution vectors, by default the one corresponding to the first qoi. Creates the vector if it doesn't already exist.

Definition at line 1220 of file system.C.

References libMesh::System::add_vector(), and libMesh::System::set_vector_as_adjoint().

Referenced by libMesh::ImplicitSystem::adjoint_solve().

{
  std::ostringstream adjoint_name;
  adjoint_name << "adjoint_solution" << i;

  NumericVector<Number> & returnval = this->add_vector(adjoint_name.str());
  this->set_vector_as_adjoint(adjoint_name.str(), i);
  return returnval;
}

add_matrices()

void libMesh::ImplicitSystem::add_matrices ( )

overrideprotectedvirtualinherited

Adds the system matrix.

Reimplemented from libMesh::System.

Definition at line 113 of file implicit_system.C.

References libMesh::System::add_matrices(), libMesh::System::add_matrix(), libMesh::libmesh_assert(), libMesh::ImplicitSystem::matrix, and libMesh::System::n_matrices().

{
  Parent::add_matrices();

  // Possible that we cleared the _matrices but
  // forgot to update the matrix pointer?
  if (this->n_matrices() == 0)
    matrix = nullptr;

  // Only need to add the matrix if it isn't there
  // already!
  if (matrix == nullptr)
    matrix = &(this->add_matrix ("System Matrix"));

  libmesh_assert(matrix);
}

add_matrix() [1/3]

SparseMatrix< Number > & libMesh::System::add_matrix ( std::string_view  mat_name, ParallelType  type = PARALLEL, MatrixBuildType  mat_build_type = MatrixBuildType::AUTOMATIC  )

inherited

Adds the additional matrix mat_name to this system.

Only allowed prior to assemble(). All additional matrices have the same sparsity pattern as the matrix used during solution. When not System but the user wants to initialize the main/system matrix, then all the additional matrices, if existent, have to be initialized by the user, too.

This non-template method will add a derived matrix type corresponding to the solver package. If the user wishes to specify the matrix type to add, use the templated add_matrix method instead

Parameters

mat_name	A name for the matrix
type	The serial/parallel/ghosted type of the matrix
mat_build_type	The matrix type to build

Definition at line 998 of file system.C.

References libMesh::System::_matrices, libMesh::System::_matrix_types, libMesh::SparseMatrix< T >::build(), libMesh::ParallelObject::comm(), libMesh::default_solver_package(), libMesh::DIAGONAL, libMesh::System::get_dof_map(), libMesh::System::get_mesh(), libMesh::DofMap::get_static_condensation(), libMesh::System::has_static_condensation(), libMesh::if(), libMesh::System::late_matrix_init(), and libMesh::libmesh_assert().

Referenced by libMesh::ImplicitSystem::add_matrices(), libMesh::EigenSystem::add_matrices(), alternative_fe_assembly(), libMesh::ClawSystem::assemble_boundary_condition_matrices(), libMesh::ImplicitSystem::create_static_condensation_system_matrix(), form_matrixA(), libMesh::EigenTimeSolver::init(), main(), and libMesh::NewmarkSystem::NewmarkSystem().

{
  parallel_object_only();

  libmesh_assert(this->comm().verify(std::string(mat_name)));
  libmesh_assert(this->comm().verify(int(type)));
  libmesh_assert(this->comm().verify(int(mat_build_type)));

  // Return the matrix if it is already there.
  if (auto it = this->_matrices.find(mat_name);
      it != this->_matrices.end())
    return *it->second;

  // Otherwise build the matrix to return.
  std::unique_ptr<SparseMatrix<Number>> matrix;
  if (this->has_static_condensation())
    {
      if (mat_build_type == MatrixBuildType::DIAGONAL)
        libmesh_error_msg(
            "We do not currently support static condensation of the diagonal matrix type");
      matrix = std::make_unique<StaticCondensation>(this->get_mesh(),
                                                    *this,
                                                    this->get_dof_map(),
                                                    this->get_dof_map().get_static_condensation());
    }
  else
    matrix = SparseMatrix<Number>::build(this->comm(), libMesh::default_solver_package());
  auto & mat = *matrix;

  _matrices.emplace(mat_name, std::move(matrix));

  _matrix_types.emplace(mat_name, type);

  // Initialize it first if we've already initialized the others.
  this->late_matrix_init(mat, type);

  return mat;
}

add_matrix() [2/3]

template<template< typename > class MatrixType>

SparseMatrix< Number > & libMesh::System::add_matrix ( std::string_view  mat_name, ParallelType  type = PARALLEL  )

inlineinherited

Adds the additional matrix mat_name to this system.

Only allowed prior to assemble(). All additional matrices have the same sparsity pattern as the matrix used during solution. When not System but the user wants to initialize the main/system matrix, then all the additional matrices, if existent, have to be initialized by the user, too.

This method will create add a derived matrix of type MatrixType<Number>. One can use the non-templated add_matrix method to add a matrix corresponding to the default solver package

Parameters

mat_name	A name for the matrix
type	The serial/parallel/ghosted type of the matrix

Definition at line 2646 of file system.h.

References libMesh::System::_matrices, libMesh::System::_matrix_types, and libMesh::System::late_matrix_init().

{
  // Return the matrix if it is already there.
  auto it = this->_matrices.find(mat_name);
  if (it != this->_matrices.end())
    return *it->second;

  // Otherwise build the matrix to return.
  auto pr = _matrices.emplace(mat_name, std::make_unique<MatrixType<Number>>(this->comm()));
  _matrix_types.emplace(mat_name, type);

  SparseMatrix<Number> & mat = *(pr.first->second);

  // Initialize it first if we've already initialized the others.
  this->late_matrix_init(mat, type);

  return mat;
}

add_matrix() [3/3]

SparseMatrix< Number > & libMesh::System::add_matrix ( std::string_view  mat_name, std::unique_ptr< SparseMatrix< Number >>  matrix, ParallelType  type = PARALLEL  )

inherited

Adds the additional matrix mat_name to this system.

Only allowed prior to assemble(). All additional matrices have the same sparsity pattern as the matrix used during solution. When not System but the user wants to initialize the main/system matrix, then all the additional matrices, if existent, have to be initialized by the user, too.

Parameters

mat_name	A name for the matrix
matrix	The matrix we are handing over the System for ownership
type	The serial/parallel/ghosted type of the matrix

Definition at line 1041 of file system.C.

References libMesh::System::_matrices, libMesh::System::_matrix_types, libMesh::ParallelObject::comm(), libMesh::System::late_matrix_init(), and libMesh::libmesh_assert().

{
  parallel_object_only();

  const std::string namestr{mat_name};

  libmesh_assert(this->comm().verify(namestr));
  libmesh_assert(this->comm().verify(int(type)));

  SparseMatrix<Number> & mat = *matrix;

  _matrices[namestr] = std::move(matrix);
  _matrix_types[namestr] = type;

  // Initialize it first if we've already initialized the others.
  this->late_matrix_init(mat, type);

  return mat;
}

add_scaled_Aq()

void libMesh::RBConstruction::add_scaled_Aq	(	Number	scalar,
		unsigned int	q_a,
		SparseMatrix< Number > *	input_matrix,
		bool	symmetrize
	)

Add the scaled q^th affine matrix to input_matrix.

If symmetrize==true, then we symmetrize Aq before adding it.

Definition at line 1056 of file rb_construction.C.

References libMesh::SparseMatrix< T >::add(), add_scaled_matrix_and_vector(), libMesh::SparseMatrix< T >::close(), libMesh::RBAssemblyExpansion::get_A_assembly(), get_Aq(), get_rb_theta_expansion(), and rb_assembly_expansion.

Referenced by libMesh::RBSCMConstruction::add_scaled_symm_Aq().

{
  LOG_SCOPE("add_scaled_Aq()", "RBConstruction");

  libmesh_error_msg_if(q_a >= get_rb_theta_expansion().get_n_A_terms(),
                       "Error: We must have q < Q_a in add_scaled_Aq.");

  if (!symmetrize)
    {
      input_matrix->add(scalar, *get_Aq(q_a));
      input_matrix->close();
    }
  else
    {
      add_scaled_matrix_and_vector(scalar,
                                   &rb_assembly_expansion->get_A_assembly(q_a),
                                   input_matrix,
                                   nullptr,
                                   symmetrize);
    }
}

add_scaled_matrix_and_vector()

void libMesh::RBConstruction::add_scaled_matrix_and_vector ( Number  scalar, ElemAssembly *  elem_assembly, SparseMatrix< Number > *  input_matrix, NumericVector< Number > *  input_vector, bool  symmetrize = false, bool  apply_dof_constraints = true  )

protected

This function loops over the mesh and applies the specified interior and/or boundary assembly routines, then adds the scaled result to input_matrix and/or input_vector.

If symmetrize==true then we assemble the symmetric part of the matrix, 0.5*(A + A^T)

Definition at line 641 of file rb_construction.C.

References libMesh::DofMap::_dof_coupling, libMesh::SparseMatrix< T >::add_matrix(), libMesh::NumericVector< T >::add_vector(), libMesh::FEAbstract::attach_quadrature_rule(), libMesh::ElemAssembly::boundary_assembly(), build_context(), libMesh::NumericVector< T >::close(), libMesh::SparseMatrix< T >::close(), libMesh::ParallelObject::comm(), libMesh::DofMap::constrain_element_matrix_and_vector(), libMesh::FEType::default_quadrature_rule(), libMesh::DGFEMContext::dg_terms_are_active(), libMesh::DenseMatrix< T >::el(), libMesh::FEMContext::elem_fe_reinit(), libMesh::DiffContext::get_dof_indices(), libMesh::System::get_dof_map(), libMesh::FEMContext::get_elem(), libMesh::DGFEMContext::get_elem_elem_jacobian(), libMesh::DiffContext::get_elem_jacobian(), libMesh::DGFEMContext::get_elem_neighbor_jacobian(), libMesh::DiffContext::get_elem_residual(), libMesh::FEMContext::get_element_fe(), libMesh::FEAbstract::get_fe_type(), libMesh::System::get_mesh(), libMesh::DGFEMContext::get_neighbor_dof_indices(), libMesh::DGFEMContext::get_neighbor_elem_jacobian(), libMesh::DGFEMContext::get_neighbor_neighbor_jacobian(), libMesh::ElemAssembly::get_nodal_values(), libMesh::FEMContext::get_side(), libMesh::DenseMatrix< T >::get_transpose(), impose_internal_fluxes, init_context(), libMesh::ElemAssembly::interior_assembly(), libMesh::Tri3Subdivision::is_ghost(), libMesh::DenseMatrixBase< T >::m(), TIMPI::Communicator::max(), mesh, libMesh::DenseMatrixBase< T >::n(), libMesh::Elem::n_sides(), libMesh::System::n_vars(), libMesh::Elem::neighbor_ptr(), libMesh::NODEELEM, post_process_elem_matrix_and_vector(), libMesh::FEMContext::pre_fe_reinit(), libMesh::DofObject::processor_id(), TIMPI::Communicator::rank(), libMesh::FEMContext::side, libMesh::DGFEMContext::side_fe_reinit(), libMesh::Elem::side_ptr(), skip_degenerate_sides, libMesh::TRI3SUBDIVISION, and libMesh::MeshTools::volume().

Referenced by add_scaled_Aq(), assemble_all_output_vectors(), assemble_Aq_matrix(), assemble_Fq_vector(), assemble_inner_product_matrix(), libMesh::TransientRBConstruction::assemble_L2_matrix(), and libMesh::TransientRBConstruction::assemble_Mq_matrix().

{
  LOG_SCOPE("add_scaled_matrix_and_vector()", "RBConstruction");

  bool assemble_matrix = (input_matrix != nullptr);
  bool assemble_vector = (input_vector != nullptr);

  if (!assemble_matrix && !assemble_vector)
    return;

  const MeshBase & mesh = this->get_mesh();

  // First add any node-based terms (e.g. point loads)

  // Make a std::set of all the nodes that are in 1 or more
  // nodesets. We only want to call get_nodal_values() once per Node
  // per ElemAssembly object, regardless of how many nodesets it
  // appears in.
  std::set<dof_id_type> nodes_with_nodesets;
  for (const auto & t : mesh.get_boundary_info().build_node_list())
    nodes_with_nodesets.insert(std::get<0>(t));

  // It's possible for the node assembly loop below to throw an
  // exception on one (or some subset) of the processors. In that
  // case, we stop assembling on the processor(s) that threw and let
  // the other processors finish assembly. Then we synchronize to
  // check whether an exception was thrown on _any_ processor, and if
  // so, re-throw it on _all_ processors.
  int nodal_assembly_threw = 0;

  libmesh_try
  {

  for (const auto & id : nodes_with_nodesets)
    {
      const Node & node = mesh.node_ref(id);

      // If node is on this processor, then all dofs on node are too
      // so we can do the add below safely
      if (node.processor_id() == this->comm().rank())
        {
          // Get the values to add to the rhs vector
          std::vector<dof_id_type> nodal_dof_indices;
          DenseMatrix<Number> nodal_matrix;
          DenseVector<Number> nodal_rhs;
          elem_assembly->get_nodal_values(nodal_dof_indices,
                                          nodal_matrix,
                                          nodal_rhs,
                                          *this,
                                          node);

          // Perform any required user-defined postprocessing on
          // the matrix and rhs.
          //
          // TODO: We need to postprocess node matrices and vectors
          // in some cases (e.g. when rotations are applied to
          // nodes), but since we don't have a FEMContext at this
          // point we would need to have a different interface
          // taking the DenseMatrix, DenseVector, and probably the
          // current node that we are on...
          // this->post_process_elem_matrix_and_vector(nodal_matrix, nodal_rhs);

          if (!nodal_dof_indices.empty())
            {
              if (apply_dof_constraints)
                {
                  // Apply constraints, e.g. Dirichlet and periodic constraints
                  this->get_dof_map().constrain_element_matrix_and_vector(
                    nodal_matrix,
                    nodal_rhs,
                    nodal_dof_indices,
                    /*asymmetric_constraint_rows*/ false);
                }

              // Scale and add to global matrix and/or vector
              nodal_matrix *= scalar;
              nodal_rhs    *= scalar;

              if (assemble_vector)
                input_vector->add_vector(nodal_rhs, nodal_dof_indices);

              if (assemble_matrix)
                input_matrix->add_matrix(nodal_matrix, nodal_dof_indices);
            }
        }
    }
  } // libmesh_try
  libmesh_catch(...)
  {
    nodal_assembly_threw = 1;
  }

  // Check for exceptions on any procs and if there is one, re-throw
  // it on all procs.
  this->comm().max(nodal_assembly_threw);

  if (nodal_assembly_threw)
    libmesh_error_msg("Error during assembly in RBConstruction::add_scaled_matrix_and_vector()");

  std::unique_ptr<DGFEMContext> c = this->build_context();
  DGFEMContext & context  = cast_ref<DGFEMContext &>(*c);

  this->init_context(context);

  // It's possible for the assembly loop below to throw an exception
  // on one (or some subset) of the processors. This can happen when
  // e.g. the mesh contains one or more elements with negative
  // Jacobian. In that case, we stop assembling on the processor(s)
  // that threw and let the other processors finish assembly. Then we
  // synchronize to check whether an exception was thrown on _any_
  // processor, and if so, re-throw it on _all_ processors. This way,
  // we make it easier for callers to handle exceptions in parallel
  // during assembly: they can simply assume that the code either
  // throws on all procs or on no procs.
  int assembly_threw = 0;

  libmesh_try
  {

  for (const auto & elem : mesh.active_local_element_ptr_range())
    {
      const ElemType elemtype = elem->type();

      if(elemtype == NODEELEM)
        {
          // We assume that we do not perform any assembly directly on
          // NodeElems, so we skip the assembly calls.

          // However, in a spline basis with Dirichlet constraints on
          // spline nodes, a constrained matrix has to take those
          // nodes into account.
          if (!apply_dof_constraints)
            continue;
        }

      // Subdivision elements need special care:
      // - skip ghost elements
      // - init special quadrature rule
      std::unique_ptr<QBase> qrule;
      if (elemtype == TRI3SUBDIVISION)
        {
          const Tri3Subdivision * gh_elem = static_cast<const Tri3Subdivision *> (elem);
          if (gh_elem->is_ghost())
            continue ;
          // A Gauss quadrature rule for numerical integration.
          // For subdivision shell elements, a single Gauss point per
          // element is sufficient, hence we use extraorder = 0.
          const int extraorder = 0;
          FEBase * elem_fe = nullptr;
          context.get_element_fe( 0, elem_fe );

          qrule = elem_fe->get_fe_type().default_quadrature_rule (2, extraorder);

          // Tell the finite element object to use our quadrature rule.
          elem_fe->attach_quadrature_rule (qrule.get());
        }

      context.pre_fe_reinit(*this, elem);

      // Do nothing in case there are no dof_indices on the current element
      if ( context.get_dof_indices().empty() )
        continue;

      context.elem_fe_reinit();

      if (elemtype != NODEELEM)
        {
          elem_assembly->interior_assembly(context);

          const unsigned char n_sides = context.get_elem().n_sides();
          for (context.side = 0; context.side != n_sides; ++context.side)
            {
              // May not need to apply fluxes on non-boundary elements
              if ((context.get_elem().neighbor_ptr(context.get_side()) != nullptr) && !impose_internal_fluxes)
                continue;

              // skip degenerate sides with zero area
              if( (context.get_elem().side_ptr(context.get_side())->volume() <= 0.) && skip_degenerate_sides)
                continue;

              context.side_fe_reinit();
              elem_assembly->boundary_assembly(context);

              if (context.dg_terms_are_active())
                {
                  input_matrix->add_matrix (context.get_elem_elem_jacobian(),
                                            context.get_dof_indices(),
                                            context.get_dof_indices());

                  input_matrix->add_matrix (context.get_elem_neighbor_jacobian(),
                                            context.get_dof_indices(),
                                            context.get_neighbor_dof_indices());

                  input_matrix->add_matrix (context.get_neighbor_elem_jacobian(),
                                            context.get_neighbor_dof_indices(),
                                            context.get_dof_indices());

                  input_matrix->add_matrix (context.get_neighbor_neighbor_jacobian(),
                                            context.get_neighbor_dof_indices(),
                                            context.get_neighbor_dof_indices());
                }
            }
        }

      // Do any required user post-processing before symmetrizing and/or applying
      // constraints.
      //
      // We only do this if apply_dof_constraints is true because we want to be
      // able to set apply_dof_constraints=false in order to obtain a matrix
      // A with no dof constraints or dof transformations, as opposed to C^T A C,
      // which includes constraints and/or dof transformations. Here C refers to
      // the matrix that imposes dof constraints and transformations on the
      // solution u.
      //
      // Matrices such as A are what we store in our "non_dirichlet" operators, and
      // they are useful for computing terms such as (C u_i)^T A (C u_j) (e.g. see
      // update_RB_system_matrices()),  where C u is the result of a "truth_solve",
      // which includes calls to both enforce_constraints_exactly() and
      // post_process_truth_solution(). If we use C^T A C to compute these terms then
      // we would "double apply" the matrix C, which can give incorrect results.
      if (apply_dof_constraints)
        this->post_process_elem_matrix_and_vector(context);

      // Need to symmetrize before imposing
      // periodic constraints
      if (assemble_matrix && symmetrize)
        {
          DenseMatrix<Number> Ke_transpose;
          context.get_elem_jacobian().get_transpose(Ke_transpose);
          context.get_elem_jacobian() += Ke_transpose;
          context.get_elem_jacobian() *= 0.5;
        }

      // As discussed above, we can set apply_dof_constraints=false to
      // get A instead of C^T A C
      if (apply_dof_constraints)
        {
          // Apply constraints, e.g. Dirichlet and periodic constraints
          this->get_dof_map().constrain_element_matrix_and_vector
            (context.get_elem_jacobian(),
             context.get_elem_residual(),
             context.get_dof_indices(),
             /*asymmetric_constraint_rows*/ false );
        }

      // Scale and add to global matrix and/or vector
      context.get_elem_jacobian() *= scalar;
      context.get_elem_residual() *= scalar;

      if (assemble_matrix)
        {

          CouplingMatrix * coupling_matrix = get_dof_map()._dof_coupling;
          if (!coupling_matrix)
            {
              // If we haven't defined a _dof_coupling matrix then just add
              // the whole matrix
              input_matrix->add_matrix (context.get_elem_jacobian(),
                                        context.get_dof_indices() );
            }
          else
            {
              // Otherwise we should only add the relevant submatrices
              for (unsigned int var1=0; var1<n_vars(); var1++)
                {
                  ConstCouplingRow ccr(var1, *coupling_matrix);
                  for (const auto & var2 : ccr)
                    {
                      unsigned int sub_m = context.get_elem_jacobian( var1, var2 ).m();
                      unsigned int sub_n = context.get_elem_jacobian( var1, var2 ).n();
                      DenseMatrix<Number> sub_jac(sub_m, sub_n);
                      for (unsigned int row=0; row<sub_m; row++)
                        for (unsigned int col=0; col<sub_n; col++)
                          {
                            sub_jac(row,col) = context.get_elem_jacobian( var1, var2 ).el(row,col);
                          }
                      input_matrix->add_matrix (sub_jac,
                                                context.get_dof_indices(var1),
                                                context.get_dof_indices(var2) );
                    }
                }
            }

        }

      if (assemble_vector)
        input_vector->add_vector (context.get_elem_residual(),
                                  context.get_dof_indices() );
    } // end for (elem)

  } // libmesh_try
  libmesh_catch(...)
  {
    assembly_threw = 1;
  }

  // Note: regardless of whether any procs threw during assembly (and
  // thus didn't finish assembling), we should not leave the matrix
  // and vector in an inconsistent state, since it may be possible to
  // recover from the exception. Therefore, we close them now. The
  // assumption here is that the nature of the exception does not
  // prevent the matrix and vector from still being assembled (albeit
  // with incomplete data).
  if (assemble_matrix)
    input_matrix->close();
  if (assemble_vector)
    input_vector->close();

  // Check for exceptions on any procs and if there is one, re-throw
  // it on all procs.
  this->comm().max(assembly_threw);

  if (assembly_threw)
    libmesh_error_msg("Error during assembly in RBConstruction::add_scaled_matrix_and_vector()");
}

add_sensitivity_rhs()

NumericVector< Number > & libMesh::System::add_sensitivity_rhs ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's sensitivity rhs vectors, by default the one corresponding to the first parameter. Creates the vector if it doesn't already exist.

Definition at line 1314 of file system.C.

References libMesh::System::add_vector().

Referenced by libMesh::ImplicitSystem::assemble_residual_derivatives().

{
  std::ostringstream sensitivity_rhs_name;
  sensitivity_rhs_name << "sensitivity_rhs" << i;

  return this->add_vector(sensitivity_rhs_name.str(), false);
}

add_sensitivity_solution()

NumericVector< Number > & libMesh::System::add_sensitivity_solution ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's solution sensitivity vectors, by default the one corresponding to the first parameter. Creates the vector if it doesn't already exist.

Definition at line 1169 of file system.C.

References libMesh::System::add_vector().

Referenced by libMesh::ImplicitSystem::sensitivity_solve().

{
  std::ostringstream sensitivity_name;
  sensitivity_name << "sensitivity_solution" << i;

  return this->add_vector(sensitivity_name.str());
}

add_variable() [1/2]

unsigned int libMesh::System::add_variable ( std::string_view  var, const FEType &  type, const std::set< subdomain_id_type > *const  active_subdomains = nullptr  )

inherited

Adds the variable var to the list of variables for this system.

If active_subdomains is either nullptr (the default) or points to an empty set, then it will be assumed that var has no subdomain restrictions

Returns

The index number for the new variable.

Definition at line 1344 of file system.C.

References libMesh::DofMap::add_variable(), and libMesh::System::get_dof_map().

Referenced by libMesh::DifferentiableSystem::add_second_order_dot_vars(), libMesh::System::add_variable(), assemble_and_solve(), OverlappingTestBase::init(), SolidSystem::init_data(), CurlCurlSystem::init_data(), SimpleEIMConstruction::init_data(), libMesh::AdvectionSystem::init_data(), HeatSystem::init_data(), SimpleRBConstruction::init_data(), libMesh::VariationalSmootherSystem::init_data(), NonManifoldCouplingTestBase::init_es(), main(), libMesh::ErrorVector::plot_error(), libMesh::System::read_header(), libMesh::PetscPreconditioner< T >::set_hypre_ads_data(), libMesh::PetscPreconditioner< T >::set_hypre_ams_data(), RationalMapTest< elem_type >::setUp(), SlitMeshRefinedSystemTest::setUp(), FETestBase< order, family, elem_type, N_x >::setUp(), WriteVecAndScalar::setupTests(), SystemsTest::simpleSetup(), MultiEvaluablePredTest::test(), ConstraintOperatorTest::test1DCoarseningNewNodes(), ConstraintOperatorTest::test1DCoarseningOperator(), MeshFunctionTest::test_bad_gradient_var_with_out_of_mesh_value(), MeshFunctionTest::test_bad_hessian_var_with_out_of_mesh_value(), MeshfunctionDFEM::test_mesh_function_dfem(), MeshfunctionDFEM::test_mesh_function_dfem_grad(), MeshFunctionTest::test_p_level(), ProjectSolutionTest::test_partial_project_solution(), MeshFunctionTest::test_subdomain_id_sets(), SystemsTest::testAssemblyWithDgFemContext(), DofMapTest::testBadElemFECombo(), EquationSystemsTest::testBadVarNames(), SystemsTest::testBlockRestrictedVarNDofs(), SystemsTest::testBoundaryProjectCube(), DofMapTest::testConstraintLoopDetection(), MeshInputTest::testCopyElementSolutionImpl(), MeshInputTest::testCopyElementVectorImpl(), MeshInputTest::testCopyNodalSolutionImpl(), ConstraintOperatorTest::testCoreform(), DefaultCouplingTest::testCoupling(), PointNeighborCouplingTest::testCoupling(), SystemsTest::testDofCouplingWithVarGroups(), DofMapTest::testDofOwner(), MeshInputTest::testDynaReadPatch(), MeshInputTest::testExodusWriteElementDataFromDiscontinuousNodalData(), SystemsTest::testFirstScalarNumber(), MeshAssignTest::testMeshMoveAssign(), PeriodicBCTest::testPeriodicBC(), EquationSystemsTest::testPostInitAddElem(), EquationSystemsTest::testPostInitAddRealSystem(), SystemsTest::testProjectCubeWithMeshFunction(), MeshInputTest::testProjectionRegression(), SystemsTest::testProjectMatrix1D(), SystemsTest::testProjectMatrix2D(), SystemsTest::testProjectMatrix3D(), InfFERadialTest::testRefinement(), EquationSystemsTest::testRefineThenReinitPreserveFlags(), EquationSystemsTest::testReinitWithNodeElem(), EquationSystemsTest::testRepartitionThenReinit(), EquationSystemsTest::testSelectivePRefine(), SystemsTest::testSetSystemParameterOverEquationSystem(), BoundaryInfoTest::testShellFaceConstraints(), MeshInputTest::testSingleElementImpl(), DisjointNeighborTest::testTempJump(), DisjointNeighborTest::testTempJumpRefine(), WriteVecAndScalar::testWriteExodus(), and WriteVecAndScalar::testWriteNemesis().

{
  return this->get_dof_map().add_variable(*this, var, type, active_subdomains);
}

add_variable() [2/2]

unsigned int libMesh::System::add_variable ( std::string_view  var, const Order  order = FIRST, const FEFamily  family = LAGRANGE, const std::set< subdomain_id_type > *const  active_subdomains = nullptr, const bool  p_refinement = true  )

inherited

Adds the variable var to the list of variables for this system.

Same as before, but assumes LAGRANGE as default value for FEType.family. If active_subdomains is either nullptr (the default) or points to an empty set, then it will be assumed that var has no subdomain restrictions. If p_refinement is false, then even when on an Elem with non-zero p_level() this variable will not be p-refined.

Definition at line 1353 of file system.C.

References libMesh::System::add_variable().

{
  return this->add_variable(var,
                            FEType(order, family).set_p_refinement(p_refinement),
                            active_subdomains);
}

add_variable_array()

unsigned int libMesh::System::add_variable_array ( const std::vector< std::string > &  vars, const FEType &  type, const std::set< subdomain_id_type > *const  active_subdomains = nullptr  )

inherited

Adds variables vars to the list of variables for this system.

If active_subdomains is either nullptr (the default) or points to an empty set, then it will be assumed that the vars have no subdomain restrictions. This API will end up calling this->add_variables(). However, we will additionally store data that can be leveraged by the DofMap to build degrees of freedom containers corresponding to all the variables in this variable array

An 'array variable' is simply a sequence of contiguous variable numbers defined by pair where the first member of the pair is the first number in the variable sequence and the second member of the pair is the number of the last variable in the sequence plus one. Array variables may be used in tandem with variable grouping by downstream code to build optimized physics kernels since each variable in the array will have the same shape functions.

Returns

The index number for the last of the new variables.

Definition at line 1386 of file system.C.

References libMesh::DofMap::add_variable_array(), and libMesh::System::get_dof_map().

Referenced by DofMapTest::testArrayDofIndicesWithType().

{
  return this->get_dof_map().add_variable_array(*this, vars, type, active_subdomains);
}

add_variables() [1/2]

unsigned int libMesh::System::add_variables ( const std::vector< std::string > &  vars, const FEType &  type, const std::set< subdomain_id_type > *const  active_subdomains = nullptr  )

inherited

Adds the variables vars to the list of variables for this system.

If active_subdomains is either nullptr (the default) or points to an empty set, then it will be assumed that the vars have no subdomain restrictions

Returns

The index number for the last of the new variables.

Definition at line 1366 of file system.C.

References libMesh::DofMap::add_variables(), and libMesh::System::get_dof_map().

Referenced by libMesh::System::add_variables(), and SystemsTest::test100KVariables().

{
  return this->get_dof_map().add_variables(*this, vars, type, active_subdomains);
}

add_variables() [2/2]

unsigned int libMesh::System::add_variables ( const std::vector< std::string > &  vars, const Order  order = FIRST, const FEFamily  family = LAGRANGE, const std::set< subdomain_id_type > *const  active_subdomains = nullptr, const bool  p_refinement = true  )

inherited

Adds the variable var to the list of variables for this system.

Same as before, but assumes LAGRANGE as default value for FEType.family. If active_subdomains is either nullptr (the default) or points to an empty set, then it will be assumed that var has no subdomain restrictions. If p_refinement is false, then even when on an Elem with non-zero p_level() this variable will not be p-refined.

Definition at line 1375 of file system.C.

References libMesh::System::add_variables().

{
  return this->add_variables(vars,
                             FEType(order, family).set_p_refinement(p_refinement),
                             active_subdomains);
}

add_vector()

NumericVector< Number > & libMesh::System::add_vector ( std::string_view  vec_name, const bool  projections = true, const ParallelType  type = PARALLEL  )

inherited

Adds the additional vector vec_name to this system.

All the additional vectors are similarly distributed, like the solution, and initialized to zero.

By default vectors added by add_vector are projected to changed grids by reinit(). To zero them instead (more efficient), pass "false" as the second argument

If the vector already exists, the existing vector is returned. after any upgrade to the projections or type has been made. We only handle upgrades (projections false->true, or type PARALLEL->GHOSTED) in this fashion, not downgrades, on the theory that if two codes have differing needs we want to support the union of those needs, not the intersection. Downgrades can only be accomplished manually, via set_vector_preservation() or by setting a vector type() and re-initializing.

Definition at line 756 of file system.C.

References libMesh::System::_dof_map, libMesh::System::_is_initialized, libMesh::System::_vector_is_adjoint, libMesh::System::_vector_projections, libMesh::System::_vectors, libMesh::NumericVector< T >::build(), libMesh::NumericVector< T >::close(), libMesh::NumericVector< T >::closed(), libMesh::ParallelObject::comm(), libMesh::default_solver_package(), libMesh::GHOSTED, libMesh::NumericVector< T >::initialized(), libMesh::libmesh_assert(), libMesh::System::n_dofs(), libMesh::System::n_local_dofs(), libMesh::ParallelObject::n_processors(), libMesh::PARALLEL, libMesh::SERIAL, libMesh::NumericVector< T >::set_type(), libMesh::NumericVector< T >::swap(), and libMesh::NumericVector< T >::type().

Referenced by libMesh::System::add_adjoint_rhs(), libMesh::System::add_adjoint_solution(), libMesh::System::add_sensitivity_rhs(), libMesh::System::add_sensitivity_solution(), libMesh::ExplicitSystem::add_system_rhs(), libMesh::System::add_weighted_sensitivity_adjoint_solution(), libMesh::System::add_weighted_sensitivity_solution(), alternative_fe_assembly(), libMesh::AdjointRefinementEstimator::estimate_error(), fe_assembly(), form_functionA(), form_functionB(), libMesh::SecondOrderUnsteadySolver::init(), libMesh::UnsteadySolver::init(), libMesh::UnsteadySolver::init_adjoints(), libMesh::TimeSolver::init_adjoints(), libMesh::OptimizationSystem::init_data(), libMesh::ContinuationSystem::init_data(), main(), libMesh::NewmarkSystem::NewmarkSystem(), libMesh::System::read_header(), libMesh::FrequencySystem::set_frequencies(), libMesh::FrequencySystem::set_frequencies_by_range(), libMesh::FrequencySystem::set_frequencies_by_steps(), SystemsTest::testAddVectorProjChange(), SystemsTest::testAddVectorTypeChange(), SystemsTest::testPostInitAddVector(), and SystemsTest::testPostInitAddVectorTypeChange().

{
  parallel_object_only();

  libmesh_assert(this->comm().verify(std::string(vec_name)));
  libmesh_assert(this->comm().verify(int(type)));
  libmesh_assert(this->comm().verify(projections));

  // Return the vector if it is already there.
  if (auto it = this->_vectors.find(vec_name);
      it != this->_vectors.end())
    {
      // If the projection setting has *upgraded*, change it.
      if (projections) // only do expensive lookup if needed
        libmesh_map_find(_vector_projections, vec_name) = projections;

      NumericVector<Number> & vec = *it->second;

      // If we're in serial, our vectors are effectively SERIAL, so
      // we'll ignore any type setting.  If we're in parallel, we
      // might have a type change to deal with.

      if (this->n_processors() > 1)
        {
          // If the type setting has changed in a way we can't
          // perceive as an upgrade or a downgrade, scream.
          libmesh_assert_equal_to(type == SERIAL,
                                  vec.type() == SERIAL);

          // If the type setting has *upgraded*, change it.
          if (type == GHOSTED && vec.type() == PARALLEL)
            {
              // A *really* late upgrade is expensive, but better not
              // to risk zeroing data.
              if (vec.initialized())
                {
                  if (!vec.closed())
                    vec.close();

                  // Ideally we'd move parallel coefficients and then
                  // add ghosted coefficients, but copy and swap is
                  // simpler.  If anyone actually ever uses this case
                  // for real we can look into optimizing it.
                  auto new_vec = NumericVector<Number>::build(this->comm());
#ifdef LIBMESH_ENABLE_GHOSTED
                  new_vec->init (this->n_dofs(), this->n_local_dofs(),
                                 _dof_map->get_send_list(), /*fast=*/false,
                                 GHOSTED);
#else
                  libmesh_error_msg("Cannot initialize ghosted vectors when they are not enabled.");
#endif

                  *new_vec = vec;
                  vec.swap(*new_vec);
                }
              else
                // The PARALLEL vec is not yet initialized, so we can
                // just "upgrade" it to GHOSTED.
                vec.set_type(type);
            }
        }

      // Any upgrades are done; we're happy here.
      return vec;
    }

  // Otherwise, build the vector. The following emplace() is
  // guaranteed to succeed because, if we made it here, we don't
  // already have a vector named "vec_name". We pass the user's
  // requested ParallelType directly to NumericVector::build() so
  // that, even if the vector is not initialized now, it will get the
  // right type when it is initialized later.
  auto pr =
    _vectors.emplace(vec_name,
                     NumericVector<Number>::build(this->comm(),
                                                  libMesh::default_solver_package(),
                                                  type));
  auto buf = pr.first->second.get();
  _vector_projections.emplace(vec_name, projections);

  // Vectors are primal by default
  _vector_is_adjoint.emplace(vec_name, -1);

  // Initialize it if necessary
  if (_is_initialized)
    {
      if (type == GHOSTED)
        {
#ifdef LIBMESH_ENABLE_GHOSTED
          buf->init (this->n_dofs(), this->n_local_dofs(),
                     _dof_map->get_send_list(), /*fast=*/false,
                     GHOSTED);
#else
          libmesh_error_msg("Cannot initialize ghosted vectors when they are not enabled.");
#endif
        }
      else
        buf->init (this->n_dofs(), this->n_local_dofs(), false, type);
    }

  return *buf;
}

add_weighted_sensitivity_adjoint_solution()

NumericVector< Number > & libMesh::System::add_weighted_sensitivity_adjoint_solution ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's weighted sensitivity adjoint solution vectors, by default the one corresponding to the first qoi. Creates the vector if it doesn't already exist.

Definition at line 1252 of file system.C.

References libMesh::System::add_vector(), and libMesh::System::set_vector_as_adjoint().

Referenced by libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve().

{
  std::ostringstream adjoint_name;
  adjoint_name << "weighted_sensitivity_adjoint_solution" << i;

  NumericVector<Number> & returnval = this->add_vector(adjoint_name.str());
  this->set_vector_as_adjoint(adjoint_name.str(), i);
  return returnval;
}

add_weighted_sensitivity_solution()

NumericVector< Number > & libMesh::System::add_weighted_sensitivity_solution ( )

inherited

Returns

A reference to the solution of the last weighted sensitivity solve Creates the vector if it doesn't already exist.

Definition at line 1199 of file system.C.

References libMesh::System::add_vector().

Referenced by libMesh::ImplicitSystem::weighted_sensitivity_solve().

{
  return this->add_vector("weighted_sensitivity_solution");
}

adjoint_qoi_parameter_sensitivity()

void libMesh::ImplicitSystem::adjoint_qoi_parameter_sensitivity ( const QoISet &  qoi_indices, const ParameterVector &  parameters, SensitivityData &  sensitivities  )

overridevirtualinherited

Solves for the derivative of each of the system's quantities of interest q in qoi[qoi_indices] with respect to each parameter in parameters, placing the result for qoi i and parameter j into sensitivities[i][j].

Uses adjoint_solve() and the adjoint sensitivity method.

Currently uses finite differenced derivatives (partial q / partial p) and (partial R / partial p).

Reimplemented from libMesh::System.

Definition at line 517 of file implicit_system.C.

References libMesh::ImplicitSystem::adjoint_solve(), libMesh::SensitivityData::allocate_data(), libMesh::ExplicitSystem::assemble_qoi(), libMesh::ImplicitSystem::assemble_residual_derivatives(), libMesh::NumericVector< T >::dot(), libMesh::System::get_qoi_values(), libMesh::System::get_sensitivity_rhs(), libMesh::QoISet::has_index(), libMesh::System::is_adjoint_already_solved(), libMesh::System::n_qois(), libMesh::Real, libMesh::ParameterVector::size(), and libMesh::TOLERANCE.

Referenced by libMesh::UnsteadySolver::integrate_adjoint_sensitivity(), and main().

{
  ParameterVector & parameters_vec =
    const_cast<ParameterVector &>(parameters_in);

  const unsigned int Np = cast_int<unsigned int>
    (parameters_vec.size());
  const unsigned int Nq = this->n_qois();

  // An introduction to the problem:
  //
  // Residual R(u(p),p) = 0
  // partial R / partial u = J = system matrix
  //
  // This implies that:
  // d/dp(R) = 0
  // (partial R / partial p) +
  // (partial R / partial u) * (partial u / partial p) = 0

  // We first do an adjoint solve:
  // J^T * z = (partial q / partial u)
  // if we haven't already or dont have an initial condition for the adjoint
  if (!this->is_adjoint_already_solved())
    {
      this->adjoint_solve(qoi_indices);
    }

  this->assemble_residual_derivatives(parameters_in);

  // Get ready to fill in sensitivities:
  sensitivities.allocate_data(qoi_indices, *this, parameters_vec);

  // We use the identities:
  // dq/dp = (partial q / partial p) + (partial q / partial u) *
  //         (partial u / partial p)
  // dq/dp = (partial q / partial p) + (J^T * z) *
  //         (partial u / partial p)
  // dq/dp = (partial q / partial p) + z * J *
  //         (partial u / partial p)

  // Leading to our final formula:
  // dq/dp = (partial q / partial p) - z * (partial R / partial p)

  // In the case of adjoints with heterogenous Dirichlet boundary
  // function phi, where
  // q :=  S(u) - R(u,phi)
  // the final formula works out to:
  // dq/dp = (partial S / partial p) - z * (partial R / partial p)
  // Because we currently have no direct access to
  // (partial S / partial p), we use the identity
  // (partial S / partial p) = (partial q / partial p) +
  //                           phi * (partial R / partial p)
  // to derive an equivalent equation:
  // dq/dp = (partial q / partial p) - (z-phi) * (partial R / partial p)

  // Since z-phi degrees of freedom are zero for constrained indices,
  // we can use the same constrained -(partial R / partial p) that we
  // use for forward sensitivity solves, taking into account the
  // differing sign convention.
  //
  // Since that vector is constrained, its constrained indices are
  // zero, so its product with phi is zero, so we can neglect the
  // evaluation of phi terms.

  for (unsigned int j=0; j != Np; ++j)
    {
      // We currently get partial derivatives via central differencing

      // (partial q / partial p) ~= (q(p+dp)-q(p-dp))/(2*dp)
      // (partial R / partial p) ~= (rhs(p+dp) - rhs(p-dp))/(2*dp)

      Number old_parameter = *parameters_vec[j];

      const Real delta_p =
        TOLERANCE * std::max(std::abs(old_parameter), 1e-3);

      *parameters_vec[j] = old_parameter - delta_p;
      this->assemble_qoi(qoi_indices);
      const std::vector<Number> qoi_minus = this->get_qoi_values();

      NumericVector<Number> & neg_partialR_partialp = this->get_sensitivity_rhs(j);

      *parameters_vec[j] = old_parameter + delta_p;
      this->assemble_qoi(qoi_indices);
      const std::vector<Number> qoi_plus = this->get_qoi_values();

      std::vector<Number> partialq_partialp(Nq, 0);
      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          partialq_partialp[i] = (qoi_plus[i] - qoi_minus[i]) / (2.*delta_p);

      // Don't leave the parameter changed
      *parameters_vec[j] = old_parameter;

      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          sensitivities[i][j] = partialq_partialp[i] +
            neg_partialR_partialp.dot(this->get_adjoint_solution(i));
    }

  // All parameters_vec have been reset.
  // Reset the original qoi.

  this->assemble_qoi(qoi_indices);
}

adjoint_solve()

std::pair< unsigned int, Real > libMesh::ImplicitSystem::adjoint_solve ( const QoISet &  qoi_indices = QoISet() )

overridevirtualinherited

Assembles & solves the linear system (dR/du)^T*z = dq/du, for those quantities of interest q specified by qoi_indices.

Leave qoi_indices empty to solve all adjoint problems.

Returns

A pair with the total number of linear iterations performed and the (sum of the) final residual norms

Reimplemented from libMesh::System.

Reimplemented in libMesh::DifferentiableSystem.

Definition at line 196 of file implicit_system.C.

References libMesh::System::add_adjoint_solution(), libMesh::LinearSolver< T >::adjoint_solve(), libMesh::System::assemble_before_solve, libMesh::ExplicitSystem::assemble_qoi_derivative(), libMesh::ImplicitSystem::assembly(), libMesh::DofMap::enforce_adjoint_constraints_exactly(), libMesh::System::get_adjoint_rhs(), libMesh::System::get_adjoint_solution(), libMesh::System::get_dof_map(), libMesh::ImplicitSystem::get_linear_solve_parameters(), libMesh::ImplicitSystem::get_linear_solver(), libMesh::QoISet::has_index(), libMesh::make_range(), libMesh::ImplicitSystem::matrix, and libMesh::System::n_qois().

Referenced by libMesh::ImplicitSystem::adjoint_qoi_parameter_sensitivity(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::ImplicitSystem::qoi_parameter_hessian(), and libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product().

{
  // Log how long the linear solve takes.
  LOG_SCOPE("adjoint_solve()", "ImplicitSystem");

  if (this->assemble_before_solve)
    // Assemble the linear system
    this->assembly (/* get_residual = */ false,
                    /* get_jacobian = */ true);

  // The adjoint problem is linear
  LinearSolver<Number> * solver = this->get_linear_solver();

  // Reset and build the RHS from the QOI derivative
  this->assemble_qoi_derivative(qoi_indices,
                                /* include_liftfunc = */ false,
                                /* apply_constraints = */ true);

  // Our iteration counts and residuals will be sums of the individual
  // results
  std::pair<unsigned int, Real> solver_params =
    this->get_linear_solve_parameters();
  std::pair<unsigned int, Real> totalrval = std::make_pair(0,0.0);

  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      {
        const std::pair<unsigned int, Real> rval =
          solver->adjoint_solve (*matrix, this->add_adjoint_solution(i),
                                 this->get_adjoint_rhs(i),
                                 double(solver_params.second),
                                 solver_params.first);

        totalrval.first  += rval.first;
        totalrval.second += rval.second;
      }

  // The linear solver may not have fit our constraints exactly
#ifdef LIBMESH_ENABLE_CONSTRAINTS
  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      this->get_dof_map().enforce_adjoint_constraints_exactly
        (this->get_adjoint_solution(i), i);
#endif

  return totalrval;
}

allocate_data_structures()

void libMesh::RBConstruction::allocate_data_structures ( )

protectedvirtual

Helper function that actually allocates all the data structures required by this class.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 514 of file rb_construction.C.

References Aq_vector, libMesh::DofMap::attach_matrix(), libMesh::SparseMatrix< T >::build(), libMesh::NumericVector< T >::build(), libMesh::ParallelObject::comm(), Fq_representor, Fq_representor_innerprods, Fq_vector, libMesh::System::get_dof_map(), libMesh::RBThetaExpansion::get_n_A_terms(), libMesh::RBThetaExpansion::get_n_F_terms(), libMesh::RBThetaExpansion::get_n_output_terms(), libMesh::RBThetaExpansion::get_n_outputs(), get_rb_theta_expansion(), inner_product_matrix, libMesh::System::n_dofs(), libMesh::System::n_local_dofs(), non_dirichlet_Aq_vector, non_dirichlet_Fq_vector, non_dirichlet_inner_product_matrix, non_dirichlet_outputs_vector, output_dual_innerprods, outputs_vector, libMesh::PARALLEL, store_dirichlet_operators, store_non_dirichlet_operators, and truth_outputs.

Referenced by libMesh::TransientRBConstruction::allocate_data_structures(), and initialize_rb_construction().

{
  // Resize vectors for storing mesh-dependent data
  Aq_vector.resize(get_rb_theta_expansion().get_n_A_terms());
  Fq_vector.resize(get_rb_theta_expansion().get_n_F_terms());

  // Resize the Fq_representors and initialize each to nullptr.
  // These are basis independent and hence stored here, whereas
  // the Aq_representors are stored in RBEvaluation
  Fq_representor.resize(get_rb_theta_expansion().get_n_F_terms());

  // Initialize vectors for the inner products of the Fq representors
  // These are basis independent and therefore stored here.
  unsigned int Q_f_hat = get_rb_theta_expansion().get_n_F_terms()*(get_rb_theta_expansion().get_n_F_terms()+1)/2;
  Fq_representor_innerprods.resize(Q_f_hat);

  // Resize the output vectors
  outputs_vector.resize(get_rb_theta_expansion().get_n_outputs());
  for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
    outputs_vector[n].resize( get_rb_theta_expansion().get_n_output_terms(n) );

  // Resize the output dual norm vectors
  output_dual_innerprods.resize(get_rb_theta_expansion().get_n_outputs());
  for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
    {
      unsigned int Q_l_hat = get_rb_theta_expansion().get_n_output_terms(n)*(get_rb_theta_expansion().get_n_output_terms(n)+1)/2;
      output_dual_innerprods[n].resize(Q_l_hat);
    }

  {
    DofMap & dof_map = this->get_dof_map();

    if (store_dirichlet_operators)
      {
        dof_map.attach_matrix(*inner_product_matrix);
        inner_product_matrix->init();
        inner_product_matrix->zero();
      }

    if(store_non_dirichlet_operators)
      {
        // We also need a non-Dirichlet inner-product matrix
        non_dirichlet_inner_product_matrix = SparseMatrix<Number>::build(this->comm());
        dof_map.attach_matrix(*non_dirichlet_inner_product_matrix);
        non_dirichlet_inner_product_matrix->init();
        non_dirichlet_inner_product_matrix->zero();
      }

    if (store_dirichlet_operators)
      for (unsigned int q=0; q<get_rb_theta_expansion().get_n_A_terms(); q++)
        {
          // Initialize the memory for the matrices
          Aq_vector[q] = SparseMatrix<Number>::build(this->comm());
          dof_map.attach_matrix(*Aq_vector[q]);
          Aq_vector[q]->init();
          Aq_vector[q]->zero();
        }

    // We also need to initialize a second set of non-Dirichlet operators
    if (store_non_dirichlet_operators)
      {
        non_dirichlet_Aq_vector.resize(get_rb_theta_expansion().get_n_A_terms());
        for (unsigned int q=0; q<get_rb_theta_expansion().get_n_A_terms(); q++)
          {
            // Initialize the memory for the matrices
            non_dirichlet_Aq_vector[q] = SparseMatrix<Number>::build(this->comm());
            dof_map.attach_matrix(*non_dirichlet_Aq_vector[q]);
            non_dirichlet_Aq_vector[q]->init();
            non_dirichlet_Aq_vector[q]->zero();
          }
      }
  }

  // Initialize the vectors
  if (store_dirichlet_operators)
    for (unsigned int q=0; q<get_rb_theta_expansion().get_n_F_terms(); q++)
      {
        // Initialize the memory for the vectors
        Fq_vector[q] = NumericVector<Number>::build(this->comm());
        Fq_vector[q]->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
      }

  // We also need to initialize a second set of non-Dirichlet operators
  if (store_non_dirichlet_operators)
    {
      non_dirichlet_Fq_vector.resize(get_rb_theta_expansion().get_n_F_terms());
      for (unsigned int q=0; q<get_rb_theta_expansion().get_n_F_terms(); q++)
        {
          // Initialize the memory for the vectors
          non_dirichlet_Fq_vector[q] = NumericVector<Number>::build(this->comm());
          non_dirichlet_Fq_vector[q]->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
        }
    }

  if (store_dirichlet_operators)
    for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
      for (unsigned int q_l=0; q_l<get_rb_theta_expansion().get_n_output_terms(n); q_l++)
        {
          // Initialize the memory for the truth output vectors
          outputs_vector[n][q_l] = NumericVector<Number>::build(this->comm());
          outputs_vector[n][q_l]->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
        }

  if (store_non_dirichlet_operators)
    {
      non_dirichlet_outputs_vector.resize(get_rb_theta_expansion().get_n_outputs());
      for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
        {
          non_dirichlet_outputs_vector[n].resize( get_rb_theta_expansion().get_n_output_terms(n) );
          for (unsigned int q_l=0; q_l<get_rb_theta_expansion().get_n_output_terms(n); q_l++)
            {
              // Initialize the memory for the truth output vectors
              non_dirichlet_outputs_vector[n][q_l] = NumericVector<Number>::build(this->comm());
              non_dirichlet_outputs_vector[n][q_l]->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
            }
        }
    }

  // Resize truth_outputs vector
  truth_outputs.resize(this->get_rb_theta_expansion().get_n_outputs());
}

assemble()

virtual void libMesh::LinearImplicitSystem::assemble ( )

inlineoverridevirtualinherited

Prepares matrix and _dof_map for matrix assembly.

Does not actually assemble anything. For matrix assembly, use the assemble() in derived classes. Should be overridden in derived classes.

Reimplemented from libMesh::ImplicitSystem.

Reimplemented in libMesh::FrequencySystem, and libMesh::NewmarkSystem.

Definition at line 115 of file linear_implicit_system.h.

References libMesh::ImplicitSystem::assemble().

Referenced by libMesh::NewmarkSystem::assemble(), libMesh::FrequencySystem::assemble(), libMesh::LinearImplicitSystem::assembly(), and libMesh::LinearImplicitSystem::solve().

{ ImplicitSystem::assemble(); }

assemble_affine_expansion()

void libMesh::RBConstruction::assemble_affine_expansion ( bool  skip_matrix_assembly, bool  skip_vector_assembly  )

virtual

Assemble the matrices and vectors for this system.

Optionally skip matrix or vector assembly (e.g. we may want to read data in from disk instead).

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 496 of file rb_construction.C.

References assemble_all_affine_operators(), assemble_all_affine_vectors(), assemble_all_output_vectors(), and assemble_misc_matrices().

Referenced by libMesh::TransientRBConstruction::assemble_affine_expansion(), and initialize_rb_construction().

{
  if (!skip_matrix_assembly)
    {
      // Assemble and store all of the matrices
      this->assemble_misc_matrices();
      this->assemble_all_affine_operators();
    }

  if (!skip_vector_assembly)
    {
      // Assemble and store all of the vectors
      this->assemble_all_affine_vectors();
      this->assemble_all_output_vectors();
    }
}

assemble_all_affine_operators()

void libMesh::RBConstruction::assemble_all_affine_operators ( )

protectedvirtual

Assemble and store all Q_a affine operators as well as the inner-product matrix.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 1096 of file rb_construction.C.

References assemble_Aq_matrix(), get_Aq(), libMesh::RBThetaExpansion::get_n_A_terms(), get_non_dirichlet_Aq(), get_rb_theta_expansion(), libMesh::out, store_dirichlet_operators, and store_non_dirichlet_operators.

Referenced by assemble_affine_expansion(), and libMesh::TransientRBConstruction::assemble_all_affine_operators().

{
  if (store_dirichlet_operators)
    for (unsigned int q_a=0; q_a<get_rb_theta_expansion().get_n_A_terms(); q_a++)
      {
        libMesh::out << "Assembling affine operator " << (q_a+1) << " of "
                    << get_rb_theta_expansion().get_n_A_terms() << std::endl;
        assemble_Aq_matrix(q_a, get_Aq(q_a));
      }

  if (store_non_dirichlet_operators)
    {
      for (unsigned int q_a=0; q_a<get_rb_theta_expansion().get_n_A_terms(); q_a++)
        {
          libMesh::out << "Assembling non-Dirichlet affine operator " << (q_a+1) << " of "
                       << get_rb_theta_expansion().get_n_A_terms() << std::endl;
          assemble_Aq_matrix(q_a, get_non_dirichlet_Aq(q_a), false);
        }
    }
}

assemble_all_affine_vectors()

void libMesh::RBConstruction::assemble_all_affine_vectors ( )

protectedvirtual

Assemble and store the affine RHS vectors.

Definition at line 1117 of file rb_construction.C.

References assemble_Fq_vector(), get_Fq(), libMesh::RBThetaExpansion::get_n_F_terms(), get_non_dirichlet_Fq(), get_rb_theta_expansion(), libMesh::out, store_dirichlet_operators, and store_non_dirichlet_operators.

Referenced by assemble_affine_expansion().

{
  if (store_dirichlet_operators)
    for (unsigned int q_f=0; q_f<get_rb_theta_expansion().get_n_F_terms(); q_f++)
      {
        libMesh::out << "Assembling affine vector " << (q_f+1) << " of "
                    << get_rb_theta_expansion().get_n_F_terms() << std::endl;
        assemble_Fq_vector(q_f, get_Fq(q_f));
      }

  if (store_non_dirichlet_operators)
    {
      for (unsigned int q_f=0; q_f<get_rb_theta_expansion().get_n_F_terms(); q_f++)
        {
          libMesh::out << "Assembling non-Dirichlet affine vector " << (q_f+1) << " of "
                       << get_rb_theta_expansion().get_n_F_terms() << std::endl;
          assemble_Fq_vector(q_f, get_non_dirichlet_Fq(q_f), false);
        }
    }

}

assemble_all_output_vectors()

void libMesh::RBConstruction::assemble_all_output_vectors ( )

protectedvirtual

Assemble and store the output vectors.

Definition at line 1156 of file rb_construction.C.

References add_scaled_matrix_and_vector(), libMesh::RBThetaExpansion::get_n_output_terms(), libMesh::RBThetaExpansion::get_n_outputs(), get_non_dirichlet_output_vector(), libMesh::RBAssemblyExpansion::get_output_assembly(), get_output_vector(), get_rb_theta_expansion(), libMesh::out, rb_assembly_expansion, store_dirichlet_operators, store_non_dirichlet_operators, and libMesh::NumericVector< T >::zero().

Referenced by assemble_affine_expansion().

{
  if (store_dirichlet_operators)
    for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
      for (unsigned int q_l=0; q_l<get_rb_theta_expansion().get_n_output_terms(n); q_l++)
        {
          libMesh::out << "Assembling output vector, (" << (n+1) << "," << (q_l+1)
                      << ") of (" << get_rb_theta_expansion().get_n_outputs()
                      << "," << get_rb_theta_expansion().get_n_output_terms(n) << ")"
                      << std::endl;
          get_output_vector(n, q_l)->zero();
          add_scaled_matrix_and_vector(1., &rb_assembly_expansion->get_output_assembly(n,q_l),
                                      nullptr,
                                      get_output_vector(n,q_l),
                                      false, /* symmetrize */
                                      true   /* apply_dof_constraints */);
        }

  if (store_non_dirichlet_operators)
    {
      for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
        for (unsigned int q_l=0; q_l<get_rb_theta_expansion().get_n_output_terms(n); q_l++)
          {
            libMesh::out << "Assembling non-Dirichlet output vector, (" << (n+1) << "," << (q_l+1)
                         << ") of (" << get_rb_theta_expansion().get_n_outputs()
                         << "," << get_rb_theta_expansion().get_n_output_terms(n) << ")"
                         << std::endl;
            get_non_dirichlet_output_vector(n, q_l)->zero();
            add_scaled_matrix_and_vector(1., &rb_assembly_expansion->get_output_assembly(n,q_l),
                                         nullptr,
                                         get_non_dirichlet_output_vector(n,q_l),
                                         false, /* symmetrize */
                                         false  /* apply_dof_constraints */);
          }
    }
}

assemble_Aq_matrix()

void libMesh::RBConstruction::assemble_Aq_matrix	(	unsigned int	q,
		SparseMatrix< Number > *	input_matrix,
		bool	apply_dof_constraints = true
	)

Assemble the q^th affine matrix and store it in input_matrix.

Definition at line 1039 of file rb_construction.C.

References add_scaled_matrix_and_vector(), libMesh::RBAssemblyExpansion::get_A_assembly(), get_rb_theta_expansion(), rb_assembly_expansion, and libMesh::SparseMatrix< T >::zero().

Referenced by assemble_all_affine_operators().

{
  libmesh_error_msg_if(q >= get_rb_theta_expansion().get_n_A_terms(),
                       "Error: We must have q < Q_a in assemble_Aq_matrix.");

  input_matrix->zero();

  add_scaled_matrix_and_vector(1.,
                               &rb_assembly_expansion->get_A_assembly(q),
                               input_matrix,
                               nullptr,
                               false, /* symmetrize */
                               apply_dof_constraints);
}

assemble_Fq_vector()

void libMesh::RBConstruction::assemble_Fq_vector	(	unsigned int	q,
		NumericVector< Number > *	input_vector,
		bool	apply_dof_constraints = true
	)

Assemble the q^th affine vector and store it in input_matrix.

Definition at line 1139 of file rb_construction.C.

References add_scaled_matrix_and_vector(), libMesh::RBAssemblyExpansion::get_F_assembly(), get_rb_theta_expansion(), rb_assembly_expansion, and libMesh::NumericVector< T >::zero().

Referenced by assemble_all_affine_vectors().

{
  libmesh_error_msg_if(q >= get_rb_theta_expansion().get_n_F_terms(),
                       "Error: We must have q < Q_f in assemble_Fq_vector.");

  input_vector->zero();

  add_scaled_matrix_and_vector(1.,
                               &rb_assembly_expansion->get_F_assembly(q),
                               nullptr,
                               input_vector,
                               false,             /* symmetrize */
                               apply_dof_constraints /* apply_dof_constraints */);
}

assemble_inner_product_matrix()

void libMesh::RBConstruction::assemble_inner_product_matrix	(	SparseMatrix< Number > *	input_matrix,
		bool	apply_dof_constraints = true
	)

Assemble the inner product matrix and store it in input_matrix.

Definition at line 1006 of file rb_construction.C.

References add_scaled_matrix_and_vector(), energy_inner_product_coeffs, libMesh::RBAssemblyExpansion::get_A_assembly(), libMesh::RBThetaExpansion::get_n_A_terms(), get_rb_theta_expansion(), inner_product_assembly, rb_assembly_expansion, use_energy_inner_product, and libMesh::SparseMatrix< T >::zero().

Referenced by assemble_misc_matrices().

{
  input_matrix->zero();

  if(!use_energy_inner_product)
  {
    add_scaled_matrix_and_vector(1.,
                                 inner_product_assembly,
                                 input_matrix,
                                 nullptr,
                                 false, /* symmetrize */
                                 apply_dof_constraints);
  }
  else
  {
    libmesh_error_msg_if(energy_inner_product_coeffs.size() != get_rb_theta_expansion().get_n_A_terms(),
                         "Error: invalid number of entries in energy_inner_product_coeffs.");

    // We symmetrize below so that we may use the energy inner-product even in cases
    // where the A_q are not symmetric.
    for (unsigned int q_a=0; q_a<get_rb_theta_expansion().get_n_A_terms(); q_a++)
      {
        add_scaled_matrix_and_vector(energy_inner_product_coeffs[q_a],
                                     &rb_assembly_expansion->get_A_assembly(q_a),
                                     input_matrix,
                                     nullptr,
                                     true, /* symmetrize */
                                     apply_dof_constraints);
      }
  }
}

assemble_misc_matrices()

void libMesh::RBConstruction::assemble_misc_matrices ( )

protectedvirtual

Assemble and store all the inner-product matrix, the constraint matrix (for constrained problems) and the mass matrix (for time-dependent problems).

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 1081 of file rb_construction.C.

References assemble_inner_product_matrix(), inner_product_matrix, non_dirichlet_inner_product_matrix, libMesh::out, store_dirichlet_operators, and store_non_dirichlet_operators.

Referenced by assemble_affine_expansion(), and libMesh::TransientRBConstruction::assemble_misc_matrices().

{
  if (store_dirichlet_operators)
    {
      libMesh::out << "Assembling inner product matrix" << std::endl;
      assemble_inner_product_matrix(inner_product_matrix.get());
    }

  if (store_non_dirichlet_operators)
    {
      libMesh::out << "Assembling non-Dirichlet inner product matrix" << std::endl;
      assemble_inner_product_matrix(non_dirichlet_inner_product_matrix.get(), false);
    }
}

assemble_qoi()

void libMesh::ExplicitSystem::assemble_qoi ( const QoISet &  qoi_indices = QoISet() )

overridevirtualinherited

Prepares qoi for quantity of interest assembly, then calls user qoi function.

Can be overridden in derived classes.

Reimplemented from libMesh::System.

Reimplemented in libMesh::FEMSystem.

Definition at line 54 of file explicit_system.C.

References libMesh::System::assemble_qoi(), libMesh::QoISet::has_index(), libMesh::make_range(), libMesh::System::n_qois(), and libMesh::System::set_qoi().

Referenced by libMesh::ImplicitSystem::adjoint_qoi_parameter_sensitivity(), libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity(), libMesh::Euler2Solver::integrate_qoi_timestep(), libMesh::EulerSolver::integrate_qoi_timestep(), libMesh::SteadySolver::integrate_qoi_timestep(), libMesh::ImplicitSystem::qoi_parameter_hessian(), and libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product().

{
  // The user quantity of interest assembly gets to expect to
  // accumulate on initially zero values
  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      this->set_qoi(i, 0);

  Parent::assemble_qoi (qoi_indices);
}

assemble_qoi_derivative()

void libMesh::ExplicitSystem::assemble_qoi_derivative ( const QoISet &  qoi_indices = QoISet(), bool  include_liftfunc = true, bool  apply_constraints = true  )

overridevirtualinherited

Prepares adjoint_rhs for quantity of interest derivative assembly, then calls user qoi derivative function.

Can be overridden in derived classes.

Reimplemented from libMesh::System.

Reimplemented in libMesh::FEMSystem.

Definition at line 67 of file explicit_system.C.

References libMesh::System::add_adjoint_rhs(), libMesh::System::assemble_qoi_derivative(), libMesh::QoISet::has_index(), libMesh::make_range(), libMesh::System::n_qois(), and libMesh::NumericVector< T >::zero().

Referenced by libMesh::ImplicitSystem::adjoint_solve(), libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity(), libMesh::ImplicitSystem::qoi_parameter_hessian(), libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product(), and libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve().

{
  // The user quantity of interest derivative assembly gets to expect
  // to accumulate on initially zero vectors
  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      this->add_adjoint_rhs(i).zero();

  Parent::assemble_qoi_derivative (qoi_indices, include_liftfunc,
                                   apply_constraints);
}

assemble_residual_derivatives()

void libMesh::ImplicitSystem::assemble_residual_derivatives ( const ParameterVector &  parameters )

overridevirtualinherited

Residual parameter derivative function.

Uses finite differences by default.

This will assemble the sensitivity rhs vectors to hold -(partial R / partial p_i), making them ready to solve the forward sensitivity equation.

Can be overridden in derived classes.

Reimplemented from libMesh::System.

Definition at line 474 of file implicit_system.C.

References libMesh::System::add_sensitivity_rhs(), libMesh::ImplicitSystem::assembly(), libMesh::NumericVector< T >::close(), libMesh::Real, libMesh::ExplicitSystem::rhs, libMesh::ParameterVector::size(), and libMesh::TOLERANCE.

Referenced by libMesh::ImplicitSystem::adjoint_qoi_parameter_sensitivity(), and libMesh::ImplicitSystem::sensitivity_solve().

{
  ParameterVector & parameters_vec =
    const_cast<ParameterVector &>(parameters_in);

  const unsigned int Np = cast_int<unsigned int>
    (parameters_vec.size());

  for (unsigned int p=0; p != Np; ++p)
    {
      NumericVector<Number> & sensitivity_rhs = this->add_sensitivity_rhs(p);

      // Approximate -(partial R / partial p) by
      // (R(p-dp) - R(p+dp)) / (2*dp)

      Number old_parameter = *parameters_vec[p];

      const Real delta_p =
        TOLERANCE * std::max(std::abs(old_parameter), 1e-3);

      *parameters_vec[p] -= delta_p;

      //      this->assembly(true, false, true);
      this->assembly(true, false, false);
      this->rhs->close();
      sensitivity_rhs = *this->rhs;

      *parameters_vec[p] = old_parameter + delta_p;

      //      this->assembly(true, false, true);
      this->assembly(true, false, false);
      this->rhs->close();

      sensitivity_rhs -= *this->rhs;
      sensitivity_rhs /= (2*delta_p);
      sensitivity_rhs.close();

      *parameters_vec[p] = old_parameter;
    }
}

assembly()

void libMesh::LinearImplicitSystem::assembly ( bool  get_residual, bool  get_jacobian, bool  apply_heterogeneous_constraints = false, bool  apply_no_constraints = false  )

overridevirtualinherited

Assembles a residual in rhs and/or a jacobian in matrix, as requested.

Reimplemented from libMesh::ImplicitSystem.

Definition at line 367 of file linear_implicit_system.C.

References libMesh::NumericVector< T >::add_vector(), libMesh::LinearImplicitSystem::assemble(), libMesh::NumericVector< T >::close(), libMesh::SparseMatrix< T >::close(), libMesh::ImplicitSystem::matrix, libMesh::ExplicitSystem::rhs, and libMesh::System::solution.

{
  // Residual R(u(p),p) := A(p)*u(p) - b(p)
  // partial R / partial u = A

  this->assemble();
  this->rhs->close();
  this->matrix->close();

  *(this->rhs) *= -1.0;
  this->rhs->add_vector(*this->solution, *this->matrix);
}

attach_assemble_function()

void libMesh::System::attach_assemble_function ( void   fptrEquationSystems &es, const std::string &name )

inherited

Register a user function to use in assembling the system matrix and RHS.

Definition at line 1954 of file system.C.

References libMesh::System::_assemble_system_function, libMesh::System::_assemble_system_object, fptr(), and libMesh::libmesh_assert().

Referenced by assemble_and_solve(), main(), ConstraintOperatorTest::test1DCoarseningNewNodes(), ConstraintOperatorTest::test1DCoarseningOperator(), SystemsTest::testAssemblyWithDgFemContext(), ConstraintOperatorTest::testCoreform(), SystemsTest::testDofCouplingWithVarGroups(), PeriodicBCTest::testPeriodicBC(), DisjointNeighborTest::testTempJump(), and DisjointNeighborTest::testTempJumpRefine().

{
  libmesh_assert(fptr);

  if (_assemble_system_object != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both assembly function and object!");

      _assemble_system_object = nullptr;
    }

  _assemble_system_function = fptr;
}

attach_assemble_object()

void libMesh::System::attach_assemble_object ( System::Assembly &  assemble_in )

inherited

Register a user object to use in assembling the system matrix and RHS.

Definition at line 1971 of file system.C.

References libMesh::System::_assemble_system_function, and libMesh::System::_assemble_system_object.

Referenced by main().

{
  if (_assemble_system_function != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both assembly object and function!");

      _assemble_system_function = nullptr;
    }

  _assemble_system_object = &assemble_in;
}

attach_constraint_function()

void libMesh::System::attach_constraint_function ( void   fptrEquationSystems &es, const std::string &name )

inherited

Register a user function for imposing constraints.

Definition at line 1985 of file system.C.

References libMesh::System::_constrain_system_function, libMesh::System::_constrain_system_object, fptr(), and libMesh::libmesh_assert().

{
  libmesh_assert(fptr);

  if (_constrain_system_object != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both constraint function and object!");

      _constrain_system_object = nullptr;
    }

  _constrain_system_function = fptr;
}

attach_constraint_object()

void libMesh::System::attach_constraint_object ( System::Constraint &  constrain )

inherited

Register a user object for imposing constraints.

Definition at line 2002 of file system.C.

References libMesh::System::_constrain_system_function, and libMesh::System::_constrain_system_object.

Referenced by libMesh::VariationalMeshSmoother::setup(), and DofMapTest::testConstraintLoopDetection().

{
  if (_constrain_system_function != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both constraint object and function!");

      _constrain_system_function = nullptr;
    }

  _constrain_system_object = &constrain;
}

attach_init_function()

void libMesh::System::attach_init_function ( void   fptrEquationSystems &es, const std::string &name )

inherited

Register a user function to use in initializing the system.

Definition at line 1923 of file system.C.

References libMesh::System::_init_system_function, libMesh::System::_init_system_object, fptr(), and libMesh::libmesh_assert().

Referenced by main().

{
  libmesh_assert(fptr);

  if (_init_system_object != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both initialization function and object!");

      _init_system_object = nullptr;
    }

  _init_system_function = fptr;
}

attach_init_object()

void libMesh::System::attach_init_object ( System::Initialization &  init_in )

inherited

Register a user class to use to initialize the system.

Note

This is exclusive with the attach_init_function.

Definition at line 1940 of file system.C.

References libMesh::System::_init_system_function, and libMesh::System::_init_system_object.

{
  if (_init_system_function != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both initialization object and function!");

      _init_system_function = nullptr;
    }

  _init_system_object = &init_in;
}

attach_QOI_derivative()

void libMesh::System::attach_QOI_derivative ( void   fptrEquationSystems &es, const std::string &name, const QoISet &qoi_indices, bool include_liftfunc, bool apply_constraints )

inherited

Register a user function for evaluating derivatives of a quantity of interest with respect to test functions, whose values should be placed in System::rhs.

Definition at line 2059 of file system.C.

References libMesh::System::_qoi_evaluate_derivative_function, libMesh::System::_qoi_evaluate_derivative_object, fptr(), and libMesh::libmesh_assert().

{
  libmesh_assert(fptr);

  if (_qoi_evaluate_derivative_object != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both QOI derivative function and object!");

      _qoi_evaluate_derivative_object = nullptr;
    }

  _qoi_evaluate_derivative_function = fptr;
}

attach_QOI_derivative_object()

void libMesh::System::attach_QOI_derivative_object ( QOIDerivative &  qoi_derivative )

inherited

Register a user object for evaluating derivatives of a quantity of interest with respect to test functions, whose values should be placed in System::rhs.

Definition at line 2076 of file system.C.

References libMesh::System::_qoi_evaluate_derivative_function, and libMesh::System::_qoi_evaluate_derivative_object.

{
  if (_qoi_evaluate_derivative_function != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both QOI derivative object and function!");

      _qoi_evaluate_derivative_function = nullptr;
    }

  _qoi_evaluate_derivative_object = &qoi_derivative;
}

attach_QOI_function()

void libMesh::System::attach_QOI_function ( void   fptrEquationSystems &es, const std::string &name, const QoISet &qoi_indices )

inherited

Register a user function for evaluating the quantities of interest, whose values should be placed in System::qoi.

Definition at line 2027 of file system.C.

References libMesh::System::_qoi_evaluate_function, libMesh::System::_qoi_evaluate_object, fptr(), and libMesh::libmesh_assert().

{
  libmesh_assert(fptr);

  if (_qoi_evaluate_object != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both QOI function and object!");

      _qoi_evaluate_object = nullptr;
    }

  _qoi_evaluate_function = fptr;
}

attach_QOI_object()

void libMesh::System::attach_QOI_object ( QOI &  qoi )

inherited

Register a user object for evaluating the quantities of interest, whose values should be placed in System::qoi.

Definition at line 2045 of file system.C.

References libMesh::System::_qoi_evaluate_function, and libMesh::System::_qoi_evaluate_object.

{
  if (_qoi_evaluate_function != nullptr)
    {
      libmesh_warning("WARNING:  Cannot specify both QOI object and function!");

      _qoi_evaluate_function = nullptr;
    }

  _qoi_evaluate_object = &qoi_in;
}

attach_shell_matrix()

void libMesh::LinearImplicitSystem::attach_shell_matrix ( ShellMatrix< Number > *  shell_matrix )

inherited

This function enables the user to provide a shell matrix, i.e.

a matrix that is not stored element-wise, but as a function. When you register your shell matrix using this function, calling solve() will no longer use the matrix member but the registered shell matrix instead. You can reset this behaviour to its original state by supplying a nullptr to this function.

Definition at line 165 of file linear_implicit_system.C.

References libMesh::LinearImplicitSystem::_shell_matrix.

Referenced by libMesh::LinearImplicitSystem::detach_shell_matrix(), and main().

{
  _shell_matrix = shell_matrix;
}

boundary_project_solution() [1/2]

void libMesh::System::boundary_project_solution ( const std::set< boundary_id_type > &  b, const std::vector< unsigned int > &  variables, FunctionBase< Number > *  f, FunctionBase< Gradient > *  g = nullptr, std::optional< ConstElemRange >  active_local_range = std::nullopt  )

inherited

Projects arbitrary boundary functions onto a vector of degree of freedom values for the current system.

This method projects an arbitrary boundary function onto the solution via L2 projections and nodal interpolations on each element.

Only degrees of freedom which affect the function's trace on a boundary in the set b are affected. Only degrees of freedom associated with the variables listed in the vector variables are projected. The function value f and its gradient g are user-provided cloneable functors. A gradient g is only required/used for projecting onto finite element spaces with continuous derivatives. elem_range active_local_range, if provided, indicates the range of elements over which to perform the projection.

Definition at line 1305 of file system_projection.C.

References b.

Referenced by SystemsTest::testBoundaryProjectCube().

{
  this->boundary_project_vector(b, variables, *solution, f, g, -1 /*is_adjoint*/, active_local_range);

  solution->localize(*current_local_solution);
}

boundary_project_solution() [2/2]

void libMesh::System::boundary_project_solution ( const std::set< boundary_id_type > &  b, const std::vector< unsigned int > &  variables, ValueFunctionPointer  fptr, GradientFunctionPointer  gptr, const Parameters &  parameters, std::optional< ConstElemRange >  active_local_range = std::nullopt  )

inherited

Projects arbitrary boundary functions onto a vector of degree of freedom values for the current system.

This method projects components of an arbitrary boundary function onto the solution via L2 projections and nodal interpolations on each element.

Only degrees of freedom which affect the function's trace on a boundary in the set b are affected. Only degrees of freedom associated with the variables listed in the vector variables are projected. The function value fptr and its gradient gptr are represented by function pointers. A gradient gptr is only required/used for projecting onto finite element spaces with continuous derivatives. elem_range active_local_range, if provided, indicates the range of elements over which to perform the projection.

Definition at line 1286 of file system_projection.C.

References b, fptr(), and gptr().

{
  WrappedFunction<Number> f(*this, fptr, &function_parameters);
  WrappedFunction<Gradient> g(*this, gptr, &function_parameters);
  this->boundary_project_solution(b, variables, &f, &g, active_local_range);
}

boundary_project_vector() [1/2]

void libMesh::System::boundary_project_vector ( const std::set< boundary_id_type > &  b, const std::vector< unsigned int > &  variables, NumericVector< Number > &  new_vector, FunctionBase< Number > *  f, FunctionBase< Gradient > *  g = nullptr, int  is_adjoint = -1, std::optional< ConstElemRange >  active_local_range = std::nullopt  ) const

inherited

Projects arbitrary boundary functions onto a vector of degree of freedom values for the current system.

This method projects an arbitrary function via L2 projections and nodal interpolations on each element.

Only degrees of freedom which affect the function's trace on a boundary in the set b are affected. Only degrees of freedom associated with the variables listed in the vector variables are projected. The function value f and its gradient g are user-provided cloneable functors. A gradient g is only required/used for projecting onto finite element spaces with continuous derivatives. elem_range active_local_range, if provided, indicates the range of elements over which to perform the projection.

Constrain the new vector using the requested adjoint rather than primal constraints if is_adjoint is non-negative.

Definition at line 1343 of file system_projection.C.

References b, libMesh::NumericVector< T >::close(), libMesh::libmesh_ignore(), and libMesh::Threads::parallel_for().

{
  LOG_SCOPE ("boundary_project_vector()", "System");

  if (!active_local_range)
  {
    active_local_range.emplace
      (this->get_mesh().active_local_elements_begin(),
       this->get_mesh().active_local_elements_end());
  }

  Threads::parallel_for
    (active_local_range.value(),
     BoundaryProjectSolution(b, variables, *this, f, g,
                             this->get_equation_systems().parameters,
                             new_vector)
     );

  // We don't do SCALAR dofs when just projecting the boundary, so
  // we're done here.

  new_vector.close();

#ifdef LIBMESH_ENABLE_CONSTRAINTS
  if (is_adjoint == -1)
    this->get_dof_map().enforce_constraints_exactly(*this, &new_vector);
  else if (is_adjoint >= 0)
    this->get_dof_map().enforce_adjoint_constraints_exactly(new_vector,
                                                            is_adjoint);
#else
  libmesh_ignore(is_adjoint);
#endif
}

boundary_project_vector() [2/2]

void libMesh::System::boundary_project_vector ( const std::set< boundary_id_type > &  b, const std::vector< unsigned int > &  variables, ValueFunctionPointer  fptr, GradientFunctionPointer  gptr, const Parameters &  parameters, NumericVector< Number > &  new_vector, int  is_adjoint = -1, std::optional< ConstElemRange >  active_local_range = std::nullopt  ) const

inherited

Projects arbitrary boundary functions onto a vector of degree of freedom values for the current system.

This method projects an arbitrary boundary function via L2 projections and nodal interpolations on each element.

Only degrees of freedom which affect the function's trace on a boundary in the set b are affected. Only degrees of freedom associated with the variables listed in the vector variables are projected. The function value fptr and its gradient gptr are represented by function pointers. A gradient gptr is only required/used for projecting onto finite element spaces with continuous derivatives. elem_range active_local_range, if provided, indicates the range of elements over which to perform the projection.

Constrain the new vector using the requested adjoint rather than primal constraints if is_adjoint is non-negative.

Definition at line 1324 of file system_projection.C.

References b, fptr(), and gptr().

{
  WrappedFunction<Number> f(*this, fptr, &function_parameters);
  WrappedFunction<Gradient> g(*this, gptr, &function_parameters);
  this->boundary_project_vector(b, variables, new_vector, &f, &g,
                                is_adjoint, active_local_range);
}

broadcast_parameters()

void libMesh::RBConstructionBase< LinearImplicitSystem >::broadcast_parameters ( const unsigned int  proc_id )

inherited

Broadcasts parameters from processor proc_id to all processors.

This broadcasts the RBParameters object from .get_parameters(), and then sets it on all processors with .set_parameters().

Definition at line 720 of file rb_construction_base.C.

Referenced by compute_max_error_bound().

{
  libmesh_assert_less (proc_id, this->n_processors());

  // create a copy of the current parameters
  RBParameters current_parameters = get_parameters();
  libmesh_error_msg_if(current_parameters.n_samples()!=1,
      "Only single-sample RBParameter objects can be broadcast.");

  // Serialize the current_parameters to current_parameters_vector in order to broadcast.
  // We handle multiple samples and vector values.
  // However, the vector values are assumed to remain the same size across samples.
  const std::size_t nparams = current_parameters.n_parameters();
  const std::size_t nsamples = current_parameters.n_samples();

  // First we get the sizes of all the parameter value vectors.
  std::vector<std::size_t> param_value_sizes;
  param_value_sizes.reserve(nparams);
  for (const auto & pr : current_parameters)
    param_value_sizes.push_back(pr.second[0].size());

  // Broadcast the sizes vector and reserve memory.
  this->comm().broadcast(param_value_sizes, proc_id);
  std::size_t buffsize = std::accumulate(param_value_sizes.cbegin(), param_value_sizes.cend(), 0ul);
  std::vector<Real> serialized_parameters;
  serialized_parameters.reserve(buffsize);

  // Then we serialize the parameters/sample/value vectors into a single vector.
  for (const auto & pr : current_parameters)
    {
      for (const auto sample_idx : make_range(nsamples))
        serialized_parameters.insert(serialized_parameters.end(),
                                     pr.second[sample_idx].cbegin(),
                                     pr.second[sample_idx].cend());
    }

  // Do the broadcasts.
  this->comm().broadcast(serialized_parameters, proc_id);

  // Deserialize into the copy of the RBParameters object.
  std::size_t param_idx = 0;
  auto val_idx = serialized_parameters.cbegin();
  for (const auto & pr : current_parameters)
    {
      const std::size_t param_value_size = param_value_sizes[param_idx];
      for (const auto sample_idx: make_range(nsamples))
        {
          auto end_val_idx = std::next(val_idx,param_value_size);
          RBParameter sample_val(val_idx, end_val_idx);
          current_parameters.set_value(pr.first, sample_idx, sample_val);
          val_idx = end_val_idx;
        }
      ++param_idx;
    }

  // Overwrite the parameters globally.
  set_parameters(current_parameters);
}

build_context()

std::unique_ptr< DGFEMContext > libMesh::RBConstruction::build_context ( )

protectedvirtual

Builds a DGFEMContext object with enough information to do evaluations on each element.

We use DGFEMContext since it allows for both DG and continuous Galerkin formulations.

Definition at line 636 of file rb_construction.C.

Referenced by add_scaled_matrix_and_vector().

{
  return std::make_unique<DGFEMContext>(*this);
}

build_zero_dirichlet_boundary_object()

std::unique_ptr< DirichletBoundary > libMesh::RBConstruction::build_zero_dirichlet_boundary_object ( )

static

It's helpful to be able to generate a DirichletBoundary that stores a ZeroFunction in order to impose Dirichlet boundary conditions.

Definition at line 2527 of file rb_construction.C.

Referenced by SimpleRBConstruction::init_data().

{
  ZeroFunction<> zf;

  std::set<boundary_id_type> dirichlet_ids;
  std::vector<unsigned int> variables;

  // The DirichletBoundary constructor clones zf, so it's OK that zf is only in local scope
  return std::make_unique<DirichletBoundary>(dirichlet_ids, variables, &zf);
}

calculate_norm() [1/2]

Real libMesh::System::calculate_norm ( const NumericVector< Number > &  v, unsigned int  var, FEMNormType  norm_type, std::set< unsigned int > *  skip_dimensions = nullptr  ) const

inherited

Returns

A norm of variable var in the vector v, in the specified norm (e.g. L2, L_INF, H1)

Definition at line 1506 of file system.C.

References libMesh::DISCRETE_L1, libMesh::DISCRETE_L2, libMesh::DISCRETE_L_INF, libMesh::System::discrete_var_norm(), libMesh::L2, libMesh::System::n_vars(), and libMesh::Real.

Referenced by libMesh::TwostepTimeSolver::adjoint_solve(), libMesh::AdaptiveTimeSolver::calculate_norm(), libMesh::UnsteadySolver::du(), main(), output_norms(), ConstraintOperatorTest::testCoreform(), and MeshInputTest::testProjectionRegression().

{
  //short circuit to save time
  if (norm_type == DISCRETE_L1 ||
      norm_type == DISCRETE_L2 ||
      norm_type == DISCRETE_L_INF)
    return discrete_var_norm(v,var,norm_type);

  // Not a discrete norm
  std::vector<FEMNormType> norms(this->n_vars(), L2);
  std::vector<Real> weights(this->n_vars(), 0.0);
  norms[var] = norm_type;
  weights[var] = 1.0;
  Real val = this->calculate_norm(v, SystemNorm(norms, weights), skip_dimensions);
  return val;
}

calculate_norm() [2/2]

Real libMesh::System::calculate_norm ( const NumericVector< Number > &  v, const SystemNorm &  norm, std::set< unsigned int > *  skip_dimensions = nullptr  ) const

inherited

Returns

A norm of the vector v, using component_norm and component_scale to choose and weight the norms of each variable.

Definition at line 1528 of file system.C.

References libMesh::System::_dof_map, libMesh::System::_mesh, libMesh::FEGenericBase< OutputType >::build(), libMesh::NumericVector< T >::build(), libMesh::ParallelObject::comm(), libMesh::FEType::default_quadrature_rule(), dim, libMesh::DISCRETE_L1, libMesh::DISCRETE_L2, libMesh::DISCRETE_L_INF, libMesh::System::discrete_var_norm(), libMesh::DofMap::dof_indices(), libMesh::MeshBase::elem_dimensions(), libMesh::Utility::enum_to_string(), libMesh::FEInterface::field_type(), libMesh::System::get_dof_map(), libMesh::System::get_mesh(), libMesh::GHOSTED, libMesh::H1, libMesh::H1_SEMINORM, libMesh::H2, libMesh::H2_SEMINORM, libMesh::L1, libMesh::NumericVector< T >::l1_norm(), libMesh::L2, libMesh::NumericVector< T >::l2_norm(), libMesh::L_INF, libMesh::libmesh_assert(), libMesh::NumericVector< T >::linfty_norm(), libMesh::NumericVector< T >::local_size(), libMesh::NumericVector< T >::localize(), libMesh::make_range(), TIMPI::Communicator::max(), libMesh::System::n_vars(), libMesh::TensorTools::norm(), libMesh::TensorTools::norm_sq(), libMesh::Real, libMesh::NumericVector< T >::size(), TIMPI::Communicator::sum(), libMesh::TYPE_SCALAR, libMesh::TYPE_VECTOR, libMesh::DofMap::variable_type(), libMesh::W1_INF_SEMINORM, libMesh::W2_INF_SEMINORM, and libMesh::SystemNorm::weight().

{
  // This function must be run on all processors at once
  parallel_object_only();

  LOG_SCOPE ("calculate_norm()", "System");

  // Zero the norm before summation
  Real v_norm = 0.;

  if (norm.is_discrete())
    {
      //Check to see if all weights are 1.0 and all types are equal
      FEMNormType norm_type0 = norm.type(0);
      unsigned int check_var = 0, check_end = this->n_vars();
      for (; check_var != check_end; ++check_var)
        if ((norm.weight(check_var) != 1.0) || (norm.type(check_var) != norm_type0))
          break;

      //All weights were 1.0 so just do the full vector discrete norm
      if (check_var == this->n_vars())
        {
          if (norm_type0 == DISCRETE_L1)
            return v.l1_norm();
          if (norm_type0 == DISCRETE_L2)
            return v.l2_norm();
          if (norm_type0 == DISCRETE_L_INF)
            return v.linfty_norm();
          else
            libmesh_error_msg("Invalid norm_type0 = " << Utility::enum_to_string(norm_type0));
        }

      for (auto var : make_range(this->n_vars()))
        {
          // Skip any variables we don't need to integrate
          if (norm.weight(var) == 0.0)
            continue;

          v_norm += norm.weight(var) * discrete_var_norm(v, var, norm.type(var));
        }

      return v_norm;
    }

  // Localize the potentially parallel vector
  std::unique_ptr<NumericVector<Number>> local_v = NumericVector<Number>::build(this->comm());
  local_v->init(v.size(), v.local_size(), _dof_map->get_send_list(),
                true, GHOSTED);
  v.localize (*local_v, _dof_map->get_send_list());

  // I'm not sure how best to mix Hilbert norms on some variables (for
  // which we'll want to square then sum then square root) with norms
  // like L_inf (for which we'll just want to take an absolute value
  // and then sum).
  bool using_hilbert_norm = true,
    using_nonhilbert_norm = true;

  // Loop over all variables
  for (auto var : make_range(this->n_vars()))
    {
      // Skip any variables we don't need to integrate
      Real norm_weight_sq = norm.weight_sq(var);
      if (norm_weight_sq == 0.0)
        continue;
      Real norm_weight = norm.weight(var);

      // Check for unimplemented norms (rather than just returning 0).
      FEMNormType norm_type = norm.type(var);
      if ((norm_type==H1) ||
          (norm_type==H2) ||
          (norm_type==L2) ||
          (norm_type==H1_SEMINORM) ||
          (norm_type==H2_SEMINORM))
        {
          if (!using_hilbert_norm)
            libmesh_not_implemented();
          using_nonhilbert_norm = false;
        }
      else if ((norm_type==L1) ||
               (norm_type==L_INF) ||
               (norm_type==W1_INF_SEMINORM) ||
               (norm_type==W2_INF_SEMINORM))
        {
          if (!using_nonhilbert_norm)
            libmesh_not_implemented();
          using_hilbert_norm = false;
        }
      else
        libmesh_not_implemented();

      const FEType & fe_type = this->get_dof_map().variable_type(var);

      // Allow space for dims 0-3, and for both scalar and vector
      // elements, even if we don't use them all
      std::vector<std::unique_ptr<FEBase>> fe_ptrs(4);
      std::vector<std::unique_ptr<FEVectorBase>> vec_fe_ptrs(4);
      std::vector<std::unique_ptr<QBase>> q_rules(4);

      const std::set<unsigned char> & elem_dims = _mesh.elem_dimensions();

      // Prepare finite elements for each dimension present in the mesh
      for (const auto & dim : elem_dims)
        {
          if (skip_dimensions && skip_dimensions->find(dim) != skip_dimensions->end())
            continue;

          // Construct quadrature and finite element objects
          q_rules[dim] = fe_type.default_quadrature_rule (dim);

          const FEFieldType field_type = FEInterface::field_type(fe_type);
          if (field_type == TYPE_SCALAR)
            {
              fe_ptrs[dim] = FEBase::build(dim, fe_type);
              fe_ptrs[dim]->attach_quadrature_rule (q_rules[dim].get());
            }
          else
            {
              vec_fe_ptrs[dim] = FEVectorBase::build(dim, fe_type);
              vec_fe_ptrs[dim]->attach_quadrature_rule (q_rules[dim].get());
              libmesh_assert_equal_to(field_type, TYPE_VECTOR);
            }

        }

      std::vector<dof_id_type> dof_indices;

      // Begin the loop over the elements
      for (const auto & elem : this->get_mesh().active_local_element_ptr_range())
        {
          const unsigned int dim = elem->dim();

          // One way for implementing this would be to exchange the fe with the FEInterface- class.
          // However, it needs to be discussed whether integral-norms make sense for infinite elements.
          // or in which sense they could make sense.
          if (elem->infinite() )
            libmesh_not_implemented();

          if (skip_dimensions && skip_dimensions->find(dim) != skip_dimensions->end())
            continue;

          QBase * qrule = q_rules[dim].get();
          libmesh_assert(qrule);

          this->get_dof_map().dof_indices (elem, dof_indices, var);

          auto element_calculation = [&dof_indices, &elem,
               norm_type, norm_weight, norm_weight_sq, &qrule,
               &local_v, &v_norm](auto & fe) {
          typedef typename std::remove_reference<decltype(fe)>::type::OutputShape OutputShape;
          typedef typename TensorTools::MakeNumber<OutputShape>::type OutputNumberShape;
          typedef typename std::remove_reference<decltype(fe)>::type::OutputGradient OutputGradient;
          typedef typename TensorTools::MakeNumber<OutputGradient>::type OutputNumberGradient;

          const std::vector<Real> &                     JxW = fe.get_JxW();
          const std::vector<std::vector<OutputShape>> * phi = nullptr;
          if (norm_type == H1 ||
              norm_type == H2 ||
              norm_type == L2 ||
              norm_type == L1 ||
              norm_type == L_INF)
            phi = &(fe.get_phi());

          const std::vector<std::vector<OutputGradient>> * dphi = nullptr;
          if (norm_type == H1 ||
              norm_type == H2 ||
              norm_type == H1_SEMINORM ||
              norm_type == W1_INF_SEMINORM)
            dphi = &(fe.get_dphi());

#ifdef LIBMESH_ENABLE_SECOND_DERIVATIVES
          typedef typename std::remove_reference<decltype(fe)>::type::OutputTensor OutputTensor;

          const std::vector<std::vector<OutputTensor>> *  d2phi = nullptr;
          if (norm_type == H2 ||
              norm_type == H2_SEMINORM ||
              norm_type == W2_INF_SEMINORM)
            d2phi = &(fe.get_d2phi());
#endif

          fe.reinit (elem);

          const unsigned int n_qp = qrule->n_points();

          const unsigned int n_sf = cast_int<unsigned int>
            (dof_indices.size());

          // Begin the loop over the Quadrature points.
          for (unsigned int qp=0; qp<n_qp; qp++)
            {
              if (norm_type == L1)
                {
                  OutputNumberShape u_h = 0.;
                  for (unsigned int i=0; i != n_sf; ++i)
                    u_h += (*phi)[i][qp] * (*local_v)(dof_indices[i]);
                  v_norm += norm_weight *
                    JxW[qp] * TensorTools::norm(u_h);
                }

              if (norm_type == L_INF)
                {
                  OutputNumberShape u_h = 0.;
                  for (unsigned int i=0; i != n_sf; ++i)
                    u_h += (*phi)[i][qp] * (*local_v)(dof_indices[i]);
                  v_norm = std::max(v_norm, norm_weight * TensorTools::norm(u_h));
                }

              if (norm_type == H1 ||
                  norm_type == H2 ||
                  norm_type == L2)
                {
                  OutputNumberShape u_h = 0.;
                  for (unsigned int i=0; i != n_sf; ++i)
                    u_h += (*phi)[i][qp] * (*local_v)(dof_indices[i]);
                  v_norm += norm_weight_sq *
                    JxW[qp] * TensorTools::norm_sq(u_h);
                }

              if (norm_type == H1 ||
                  norm_type == H2 ||
                  norm_type == H1_SEMINORM)
                {
                  OutputNumberGradient grad_u_h;
                  for (unsigned int i=0; i != n_sf; ++i)
                    grad_u_h.add_scaled((*dphi)[i][qp], (*local_v)(dof_indices[i]));
                  v_norm += norm_weight_sq *
                    JxW[qp] * grad_u_h.norm_sq();
                }

              if (norm_type == W1_INF_SEMINORM)
                {
                  OutputNumberGradient grad_u_h;
                  for (unsigned int i=0; i != n_sf; ++i)
                    grad_u_h.add_scaled((*dphi)[i][qp], (*local_v)(dof_indices[i]));
                  v_norm = std::max(v_norm, norm_weight * grad_u_h.norm());
                }

#ifdef LIBMESH_ENABLE_SECOND_DERIVATIVES
              typedef typename TensorTools::MakeNumber<OutputTensor>::type OutputNumberTensor;

              if (norm_type == H2 ||
                  norm_type == H2_SEMINORM)
                {
                  OutputNumberTensor hess_u_h;
                  for (unsigned int i=0; i != n_sf; ++i)
                    hess_u_h.add_scaled((*d2phi)[i][qp], (*local_v)(dof_indices[i]));
                  v_norm += norm_weight_sq *
                    JxW[qp] * hess_u_h.norm_sq();
                }

              if (norm_type == W2_INF_SEMINORM)
                {
                  OutputNumberTensor hess_u_h;
                  for (unsigned int i=0; i != n_sf; ++i)
                    hess_u_h.add_scaled((*d2phi)[i][qp], (*local_v)(dof_indices[i]));
                  v_norm = std::max(v_norm, norm_weight * hess_u_h.norm());
                }
#endif
            }
          };

          FEBase * scalar_fe = fe_ptrs[dim].get();
          FEVectorBase * vec_fe = vec_fe_ptrs[dim].get();

          if (scalar_fe)
            {
              libmesh_assert(!vec_fe);
              element_calculation(*scalar_fe);
            }

          if (vec_fe)
            {
              libmesh_assert(!scalar_fe);
              element_calculation(*vec_fe);
            }
        }
    }

  if (using_hilbert_norm)
    {
      this->comm().sum(v_norm);
      v_norm = std::sqrt(v_norm);
    }
  else
    {
      this->comm().max(v_norm);
    }

  return v_norm;
}

can_add_matrices()

bool libMesh::System::can_add_matrices ( ) const

inlineprotectedinherited

Returns

Whether or not matrices can still be added without expensive per-matrix initialization.

Definition at line 2028 of file system.h.

References libMesh::System::_matrices_initialized.

Referenced by libMesh::EigenSystem::set_eigenproblem_type().

{ return !_matrices_initialized; }

check_convergence()

void libMesh::RBConstruction::check_convergence ( LinearSolver< Number > &  input_solver )

protected

Check if the linear solver reports convergence.

Throw an error when that is not the case.

Definition at line 2736 of file rb_construction.C.

References libMesh::LinearSolver< T >::get_converged_reason().

Referenced by compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), enrich_basis_from_rhs_terms(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

{
  libMesh::LinearConvergenceReason conv_flag;

  conv_flag = input_solver.get_converged_reason();

  libmesh_error_msg_if(conv_flag < 0, "Convergence error. Error id: " << conv_flag);
}

check_if_zero_truth_solve()

bool libMesh::RBConstruction::check_if_zero_truth_solve ( ) const

virtual

Returns

true if the most recent truth solve gave a zero solution.

Definition at line 461 of file rb_construction.C.

References libMesh::System::solution.

Referenced by train_reduced_basis_with_greedy().

{
  return (solution->l2_norm() == 0.);
}

clear()

void libMesh::RBConstruction::clear ( )

overridevirtual

Clear all the data structures associated with the system.

Reimplemented from libMesh::RBConstructionBase< LinearImplicitSystem >.

Reimplemented in libMesh::TransientSystem< RBConstruction >, and libMesh::TransientRBConstruction.

Definition at line 104 of file rb_construction.C.

References Aq_vector, libMesh::RBConstructionBase< LinearImplicitSystem >::clear(), Fq_representor, Fq_representor_innerprods_computed, Fq_vector, non_dirichlet_Aq_vector, non_dirichlet_Fq_vector, non_dirichlet_outputs_vector, outputs_vector, store_dirichlet_operators, and store_non_dirichlet_operators.

{
  LOG_SCOPE("clear()", "RBConstruction");

  Parent::clear();

  if (store_dirichlet_operators)
    {
      Aq_vector.clear();
      Fq_vector.clear();
      outputs_vector.clear();
    }

  if (store_non_dirichlet_operators)
    {
      non_dirichlet_Aq_vector.clear();
      non_dirichlet_Fq_vector.clear();
      non_dirichlet_outputs_vector.clear();
    }

  // Also delete the Fq representors
  Fq_representor.clear();

  // Set Fq_representor_innerprods_computed flag to false now
  // that we've cleared the Fq representors
  Fq_representor_innerprods_computed = false;
}

comm()

const Parallel::Communicator& libMesh::ParallelObject::comm ( ) const

inlineinherited

Returns

A reference to the Parallel::Communicator object used by this mesh.

Definition at line 97 of file parallel_object.h.

References libMesh::ParallelObject::_communicator.

Referenced by libMesh::__libmesh_petsc_diff_solver_jacobian(), libMesh::__libmesh_petsc_diff_solver_monitor(), libMesh::__libmesh_petsc_diff_solver_residual(), libMesh::__libmesh_tao_equality_constraints(), libMesh::__libmesh_tao_equality_constraints_jacobian(), libMesh::__libmesh_tao_gradient(), libMesh::__libmesh_tao_hessian(), libMesh::__libmesh_tao_inequality_constraints(), libMesh::__libmesh_tao_inequality_constraints_jacobian(), libMesh::__libmesh_tao_objective(), libMesh::MeshRefinement::_coarsen_elements(), libMesh::ExactSolution::_compute_error(), libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::Partitioner::_find_global_index_by_pid_map(), libMesh::BoundaryInfo::_find_id_maps(), libMesh::PetscLinearSolver< Number >::_petsc_shell_matrix_get_diagonal(), libMesh::SlepcEigenSolver< libMesh::Number >::_petsc_shell_matrix_get_diagonal(), libMesh::PetscLinearSolver< Number >::_petsc_shell_matrix_mult(), libMesh::SlepcEigenSolver< libMesh::Number >::_petsc_shell_matrix_mult(), libMesh::PetscLinearSolver< Number >::_petsc_shell_matrix_mult_add(), libMesh::MeshRefinement::_refine_elements(), libMesh::MeshRefinement::_smooth_flags(), libMesh::DofMap::add_constraints_to_send_list(), add_cube_convex_hull_to_mesh(), libMesh::PetscDMWrapper::add_dofs_helper(), libMesh::PetscDMWrapper::add_dofs_to_section(), libMesh::TransientRBConstruction::add_IC_to_RB_space(), libMesh::RBEIMEvaluation::add_interpolation_data(), libMesh::CondensedEigenSystem::add_matrices(), libMesh::EigenSystem::add_matrices(), libMesh::System::add_matrix(), add_scaled_matrix_and_vector(), libMesh::DofMap::add_variable(), libMesh::DofMap::add_variables(), libMesh::System::add_vector(), libMesh::MeshTools::Modification::all_tri(), libMesh::LaplaceMeshSmoother::allgather_graph(), libMesh::DofMap::allgather_recursive_constraints(), libMesh::TransientRBConstruction::allocate_data_structures(), allocate_data_structures(), libMesh::TransientRBConstruction::assemble_affine_expansion(), libMesh::AdvectionSystem::assemble_claw_rhs(), libMesh::FEMSystem::assemble_qoi(), libMesh::Nemesis_IO::assert_symmetric_cmaps(), libMesh::MeshCommunication::assign_global_indices(), libMesh::Partitioner::assign_partitioning(), libMesh::MeshTools::Generation::build_extrusion(), libMesh::Partitioner::build_graph(), libMesh::InfElemBuilder::build_inf_elem(), libMesh::BoundaryInfo::build_node_list_from_side_list(), libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::PetscDMWrapper::build_section(), libMesh::PetscDMWrapper::build_sf(), libMesh::MeshBase::cache_elem_data(), libMesh::System::calculate_norm(), libMesh::DofMap::check_dirichlet_bcid_consistency(), libMesh::MeshBase::complete_preparation(), compute_Fq_representor_innerprods(), compute_max_error_bound(), libMesh::Nemesis_IO_Helper::compute_num_global_elem_blocks(), libMesh::Nemesis_IO_Helper::compute_num_global_nodesets(), libMesh::Nemesis_IO_Helper::compute_num_global_sidesets(), compute_output_dual_innerprods(), compute_residual_dual_norm_slow(), libMesh::RBSCMConstruction::compute_SCM_bounds_on_training_set(), libMesh::DofMap::computed_sparsity_already(), libMesh::Problem_Interface::computeF(), libMesh::Problem_Interface::computeJacobian(), libMesh::Problem_Interface::computePreconditioner(), libMesh::ContinuationSystem::ContinuationSystem(), libMesh::MeshBase::copy_constraint_rows(), libMesh::ExodusII_IO::copy_elemental_solution(), libMesh::ExodusII_IO::copy_nodal_solution(), libMesh::ExodusII_IO::copy_scalar_solution(), libMesh::CondensedEigenSystem::copy_super_to_sub(), libMesh::MeshTools::correct_node_proc_ids(), libMesh::MeshTools::create_bounding_box(), libMesh::DofMap::create_dof_constraints(), libMesh::MeshTools::create_nodal_bounding_box(), libMesh::MeshRefinement::create_parent_error_vector(), libMesh::MeshTools::create_processor_bounding_box(), libMesh::MeshTools::create_subdomain_bounding_box(), libMesh::PetscMatrix< T >::create_submatrix_nosort(), create_wrapped_function(), libMesh::MeshCommunication::delete_remote_elements(), libMesh::RBEIMEvaluation::distribute_bfs(), DMlibMeshFunction(), DMlibMeshJacobian(), DMlibMeshSetSystem_libMesh(), DMVariableBounds_libMesh(), libMesh::DTKSolutionTransfer::DTKSolutionTransfer(), libMesh::MeshRefinement::eliminate_unrefined_patches(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_interiors(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_nodes(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_sides(), libMesh::TransientRBConstruction::enrich_RB_space(), libMesh::EpetraVector< T >::EpetraVector(), AssembleOptimization::equality_constraints(), libMesh::PatchRecoveryErrorEstimator::estimate_error(), libMesh::WeightedPatchRecoveryErrorEstimator::estimate_error(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::ExactErrorEstimator::estimate_error(), libMesh::SmoothnessEstimator::estimate_smoothness(), libMesh::MeshRefinement::flag_elements_by_elem_fraction(), libMesh::MeshRefinement::flag_elements_by_error_fraction(), libMesh::MeshRefinement::flag_elements_by_error_tolerance(), libMesh::MeshRefinement::flag_elements_by_mean_stddev(), libMesh::MeshRefinement::flag_elements_by_nelem_target(), libMesh::RBEIMEvaluation::gather_bfs(), libMesh::DofMap::gather_constraints(), libMesh::MeshfreeInterpolation::gather_remote_data(), libMesh::CondensedEigenSystem::get_eigenpair(), libMesh::RBEIMEvaluation::get_eim_basis_function_node_value(), libMesh::RBEIMEvaluation::get_eim_basis_function_side_value(), libMesh::RBEIMEvaluation::get_eim_basis_function_value(), libMesh::System::get_info(), libMesh::MeshBase::get_info(), libMesh::DofMap::get_info(), libMesh::RBEIMEvaluation::get_interior_basis_functions_as_vecs(), libMesh::ImplicitSystem::get_linear_solver(), libMesh::RBEIMConstruction::get_max_abs_value(), libMesh::RBEIMConstruction::get_node_max_abs_value(), libMesh::RBEIMEvaluation::get_parametrized_function_node_value(), libMesh::RBEIMEvaluation::get_parametrized_function_side_value(), libMesh::RBEIMEvaluation::get_parametrized_function_value(), libMesh::RBEIMConstruction::get_random_point(), libMesh::MeshTetInterface::improve_hull_integrity(), AssembleOptimization::inequality_constraints(), AssembleOptimization::inequality_constraints_jacobian(), libMesh::LocationMap< T >::init(), libMesh::TimeSolver::init(), libMesh::StaticCondensation::init(), libMesh::SystemSubsetBySubdomain::init(), libMesh::PetscDMWrapper::init_and_attach_petscdm(), libMesh::AdvectionSystem::init_data(), libMesh::ClawSystem::init_data(), libMesh::PetscDMWrapper::init_petscdm(), libMesh::ExodusII_IO_Helper::initialize(), libMesh::OptimizationSystem::initialize_equality_constraints_storage(), libMesh::OptimizationSystem::initialize_inequality_constraints_storage(), libMesh::RBEIMConstruction::initialize_parametrized_functions_in_training_set(), libMesh::RBEIMConstruction::inner_product(), integrate_function(), libMesh::MeshTools::libmesh_assert_consistent_distributed(), libMesh::MeshTools::libmesh_assert_consistent_distributed_nodes(), libMesh::MeshTools::libmesh_assert_contiguous_dof_ids(), libMesh::MeshTools::libmesh_assert_equal_connectivity(), libMesh::MeshTools::libmesh_assert_equal_points(), libMesh::MeshTools::libmesh_assert_parallel_consistent_new_node_procids(), libMesh::MeshTools::libmesh_assert_parallel_consistent_procids< Elem >(), libMesh::MeshTools::libmesh_assert_parallel_consistent_procids< Node >(), libMesh::MeshTools::libmesh_assert_topology_consistent_procids< Node >(), libMesh::MeshTools::libmesh_assert_valid_boundary_ids(), libMesh::MeshTools::libmesh_assert_valid_constraint_rows(), libMesh::MeshTools::libmesh_assert_valid_dof_ids(), libMesh::MeshTools::libmesh_assert_valid_neighbors(), libMesh::DistributedMesh::libmesh_assert_valid_parallel_flags(), libMesh::DistributedMesh::libmesh_assert_valid_parallel_object_ids(), libMesh::DistributedMesh::libmesh_assert_valid_parallel_p_levels(), libMesh::MeshTools::libmesh_assert_valid_refinement_flags(), libMesh::MeshTools::libmesh_assert_valid_unique_ids(), libMesh::libmesh_petsc_linesearch_shellfunc(), libMesh::libmesh_petsc_preconditioner_apply(), libMesh::libmesh_petsc_recalculate_monitor(), libMesh::libmesh_petsc_snes_fd_residual(), libMesh::libmesh_petsc_snes_jacobian(), libMesh::libmesh_petsc_snes_mffd_interface(), libMesh::libmesh_petsc_snes_mffd_residual(), libMesh::libmesh_petsc_snes_postcheck(), libMesh::libmesh_petsc_snes_precheck(), libMesh::libmesh_petsc_snes_residual(), libMesh::libmesh_petsc_snes_residual_helper(), libMesh::MeshRefinement::limit_level_mismatch_at_edge(), libMesh::MeshRefinement::limit_level_mismatch_at_node(), libMesh::MeshRefinement::limit_overrefined_boundary(), libMesh::MeshRefinement::limit_underrefined_boundary(), libMesh::LinearImplicitSystem::LinearImplicitSystem(), main(), libMesh::MeshRefinement::make_coarsening_compatible(), libMesh::MeshCommunication::make_elems_parallel_consistent(), libMesh::MeshRefinement::make_flags_parallel_consistent(), libMesh::MeshCommunication::make_new_node_proc_ids_parallel_consistent(), libMesh::MeshCommunication::make_new_nodes_parallel_consistent(), libMesh::MeshCommunication::make_node_bcids_parallel_consistent(), libMesh::MeshCommunication::make_node_ids_parallel_consistent(), libMesh::MeshCommunication::make_node_proc_ids_parallel_consistent(), libMesh::MeshCommunication::make_node_unique_ids_parallel_consistent(), libMesh::MeshCommunication::make_nodes_parallel_consistent(), libMesh::MeshCommunication::make_p_levels_parallel_consistent(), libMesh::MeshRefinement::make_refinement_compatible(), libMesh::TransientRBConstruction::mass_matrix_scaled_matvec(), libMesh::FEMSystem::mesh_position_set(), libMesh::TriangulatorInterface::MeshedHole::MeshedHole(), LinearElasticityWithContact::move_mesh(), libMesh::DistributedMesh::n_active_elem(), libMesh::MeshTools::n_active_levels(), libMesh::BoundaryInfo::n_boundary_conds(), libMesh::MeshTools::n_connected_components(), libMesh::DofMap::n_constrained_dofs(), libMesh::MeshBase::n_constraint_rows(), libMesh::DofMap::n_dofs(), libMesh::DofMap::n_dofs_per_processor(), libMesh::BoundaryInfo::n_edge_conds(), libMesh::CondensedEigenSystem::n_global_non_condensed_dofs(), libMesh::MeshTools::n_levels(), MixedOrderTest::n_neighbor_links(), libMesh::BoundaryInfo::n_nodeset_conds(), libMesh::SparsityPattern::Build::n_nonzeros(), libMesh::MeshTools::n_p_levels(), libMesh::BoundaryInfo::n_shellface_conds(), libMesh::RBEIMEvaluation::node_distribute_bfs(), libMesh::RBEIMEvaluation::node_gather_bfs(), libMesh::RBEIMConstruction::node_inner_product(), libMesh::PetscVector< libMesh::Number >::operator=(), libMesh::MeshBase::operator==(), libMesh::DistributedMesh::parallel_max_elem_id(), libMesh::DistributedMesh::parallel_max_node_id(), libMesh::ReplicatedMesh::parallel_max_unique_id(), libMesh::DistributedMesh::parallel_max_unique_id(), libMesh::DistributedMesh::parallel_n_elem(), libMesh::DistributedMesh::parallel_n_nodes(), libMesh::SparsityPattern::Build::parallel_sync(), libMesh::BoundaryInfo::parallel_sync_node_ids(), libMesh::BoundaryInfo::parallel_sync_side_ids(), libMesh::MeshTools::paranoid_n_levels(), libMesh::Partitioner::partition(), libMesh::Partitioner::partition_unpartitioned_elements(), libMesh::petsc_auto_fieldsplit(), libMesh::System::point_gradient(), libMesh::System::point_hessian(), libMesh::System::point_value(), libMesh::MeshBase::print_constraint_rows(), libMesh::DofMap::print_dof_constraints(), libMesh::DofMap::process_mesh_constraint_rows(), libMesh::Partitioner::processor_pairs_to_interface_nodes(), libMesh::InterMeshProjection::project_system_vectors(), FEMParameters::read(), libMesh::Nemesis_IO::read(), libMesh::XdrIO::read(), libMesh::EquationSystems::read(), libMesh::ExodusII_IO::read_header(), libMesh::CheckpointIO::read_header(), libMesh::XdrIO::read_header(), libMesh::System::read_header(), libMesh::RBEIMEvaluation::read_in_interior_basis_functions(), libMesh::RBEIMEvaluation::read_in_node_basis_functions(), libMesh::RBEIMEvaluation::read_in_side_basis_functions(), libMesh::RBEvaluation::read_in_vectors_from_multiple_files(), libMesh::TransientRBConstruction::read_riesz_representors_from_files(), read_riesz_representors_from_files(), libMesh::System::read_SCALAR_dofs(), libMesh::XdrIO::read_serialized_bc_names(), libMesh::XdrIO::read_serialized_bcs_helper(), libMesh::System::read_serialized_blocked_dof_objects(), libMesh::XdrIO::read_serialized_connectivity(), libMesh::XdrIO::read_serialized_nodes(), libMesh::XdrIO::read_serialized_nodesets(), libMesh::XdrIO::read_serialized_subdomain_names(), libMesh::System::read_serialized_vector(), libMesh::Nemesis_IO_Helper::read_var_names_impl(), MeshFunctionTest::read_variable_info_from_output_data(), libMesh::MeshBase::recalculate_n_partitions(), libMesh::MeshRefinement::refine_and_coarsen_elements(), libMesh::SimplexRefiner::refine_via_edges(), libMesh::StaticCondensationDofMap::reinit(), libMesh::BoundaryInfo::remove_edge_id(), libMesh::BoundaryInfo::remove_node_id(), libMesh::BoundaryInfo::remove_shellface_id(), libMesh::BoundaryInfo::remove_side_id(), libMesh::DistributedMesh::renumber_dof_objects(), libMesh::DistributedMesh::renumber_nodes_and_elements(), LinearElasticityWithContact::residual_and_jacobian(), OverlappingAlgebraicGhostingTest::run_ghosting_test(), OverlappingCouplingGhostingTest::run_sparsity_pattern_test(), scale_mesh_and_plot(), libMesh::DofMap::scatter_constraints(), libMesh::CheckpointIO::select_split_config(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::send_and_insert_dof_values(), libMesh::TransientRBConstruction::set_error_temporal_data(), libMesh::Partitioner::set_interface_node_processor_ids_BFS(), libMesh::Partitioner::set_interface_node_processor_ids_linear(), libMesh::Partitioner::set_interface_node_processor_ids_petscpartitioner(), libMesh::Partitioner::set_node_processor_ids(), libMesh::DofMap::set_nonlocal_dof_objects(), libMesh::Partitioner::set_parent_processor_ids(), libMesh::PetscDMWrapper::set_point_range_in_section(), libMesh::PetscDiffSolver::setup_petsc_data(), libMesh::RBEIMEvaluation::side_distribute_bfs(), libMesh::RBEIMEvaluation::side_gather_bfs(), libMesh::RBEIMConstruction::side_inner_product(), libMesh::Partitioner::single_partition(), libMesh::LaplaceMeshSmoother::smooth(), libMesh::VariationalMeshSmoother::smooth(), libMesh::ClawSystem::solve_conservation_law(), libMesh::split_mesh(), libMesh::RBEIMConstruction::store_eim_solutions_for_training_set(), libMesh::MeshBase::subdomain_ids(), libMesh::BoundaryInfo::sync(), libMesh::MeshBase::sync_subdomain_name_map(), ConstraintOperatorTest::test1DCoarseningNewNodes(), ConstraintOperatorTest::test1DCoarseningOperator(), MeshFunctionTest::test_bad_gradient_var_with_out_of_mesh_value(), MeshFunctionTest::test_bad_hessian_var_with_out_of_mesh_value(), libMesh::MeshRefinement::test_level_one(), MeshfunctionDFEM::test_mesh_function_dfem(), MeshfunctionDFEM::test_mesh_function_dfem_grad(), MeshFunctionTest::test_p_level(), libMesh::MeshRefinement::test_unflagged(), DofMapTest::testBadElemFECombo(), SystemsTest::testBlockRestrictedVarNDofs(), BoundaryInfoTest::testBoundaryOnChildrenErrors(), VolumeTest::testC0PolygonMethods(), VolumeTest::testC0PolyhedronMethods(), ConstraintOperatorTest::testCoreform(), ConnectedComponentsTest::testEdge(), MeshInputTest::testExodusIGASidesets(), MeshTriangulationTest::testFoundCenters(), PointLocatorTest::testLocator(), BoundaryInfoTest::testMesh(), PointLocatorTest::testPlanar(), MeshTriangulationTest::testPoly2TriRefinementBase(), SystemsTest::testProjectCubeWithMeshFunction(), BoundaryInfoTest::testRenumber(), BoundaryInfoTest::testSelectiveRenumber(), CheckpointIOTest::testSplitter(), MeshInputTest::testTetgenIO(), MeshTriangulationTest::testTriangulatorInterp(), MeshTriangulationTest::testTriangulatorMeshedHoles(), MeshTriangulationTest::testTriangulatorRoundHole(), MeshSmootherTest::testVariationalSmoother(), libMesh::MeshTools::total_weight(), train_reduced_basis_with_POD(), libMesh::MeshFunctionSolutionTransfer::transfer(), libMesh::MeshfreeSolutionTransfer::transfer(), libMesh::Poly2TriTriangulator::triangulate(), libMesh::TransientRBConstruction::truth_assembly(), truth_assembly(), libMesh::MeshRefinement::uniformly_coarsen(), update_current_local_solution(), libMesh::TransientRBConstruction::update_RB_initial_condition_all_N(), libMesh::TransientRBConstruction::update_RB_system_matrices(), update_RB_system_matrices(), libMesh::TransientRBConstruction::update_residual_terms(), update_residual_terms(), libMesh::MeshTools::volume(), libMesh::STLIO::write(), libMesh::NameBasedIO::write(), libMesh::XdrIO::write(), libMesh::VTKIO::write_nodal_data(), libMesh::RBEIMEvaluation::write_out_interior_basis_functions(), libMesh::RBEIMEvaluation::write_out_node_basis_functions(), libMesh::RBEIMEvaluation::write_out_side_basis_functions(), libMesh::RBEvaluation::write_out_vectors(), libMesh::TransientRBConstruction::write_riesz_representors_to_files(), write_riesz_representors_to_files(), libMesh::System::write_SCALAR_dofs(), libMesh::XdrIO::write_serialized_bcs_helper(), libMesh::System::write_serialized_blocked_dof_objects(), libMesh::XdrIO::write_serialized_connectivity(), libMesh::XdrIO::write_serialized_nodes(), libMesh::XdrIO::write_serialized_nodesets(), libMesh::RBDataSerialization::RBEvaluationSerialization::write_to_file(), libMesh::RBDataSerialization::TransientRBEvaluationSerialization::write_to_file(), libMesh::RBDataSerialization::RBEIMEvaluationSerialization::write_to_file(), and libMesh::RBDataSerialization::RBSCMEvaluationSerialization::write_to_file().

  { return _communicator; }

compare()

bool libMesh::System::compare ( const System &  other_system, const Real  threshold, const bool  verbose  ) const

virtualinherited

Returns

true when the other system contains identical data, up to the given threshold. Outputs some diagnostic info when verbose is set.

Definition at line 606 of file system.C.

References libMesh::System::_is_initialized, libMesh::System::_sys_name, libMesh::System::_vectors, libMesh::System::get_vector(), libMesh::libmesh_assert(), libMesh::System::n_vectors(), libMesh::System::name(), libMesh::out, and libMesh::System::solution.

{
  // we do not care for matrices, but for vectors
  libmesh_assert (_is_initialized);
  libmesh_assert (other_system._is_initialized);

  if (verbose)
    {
      libMesh::out << "  Systems \"" << _sys_name << "\"" << std::endl;
      libMesh::out << "   comparing matrices not supported." << std::endl;
      libMesh::out << "   comparing names...";
    }

  // compare the name: 0 means identical
  const int name_result = _sys_name.compare(other_system.name());
  if (verbose)
    {
      if (name_result == 0)
        libMesh::out << " identical." << std::endl;
      else
        libMesh::out << "  names not identical." << std::endl;
      libMesh::out << "   comparing solution vector...";
    }


  // compare the solution: -1 means identical
  const int solu_result = solution->compare (*other_system.solution.get(),
                                             threshold);

  if (verbose)
    {
      if (solu_result == -1)
        libMesh::out << " identical up to threshold." << std::endl;
      else
        libMesh::out << "  first difference occurred at index = "
                     << solu_result << "." << std::endl;
    }


  // safety check, whether we handle at least the same number
  // of vectors
  std::vector<int> ov_result;

  if (this->n_vectors() != other_system.n_vectors())
    {
      if (verbose)
        {
          libMesh::out << "   Fatal difference. This system handles "
                       << this->n_vectors() << " add'l vectors," << std::endl
                       << "   while the other system handles "
                       << other_system.n_vectors()
                       << " add'l vectors." << std::endl
                       << "   Aborting comparison." << std::endl;
        }
      return false;
    }
  else if (this->n_vectors() == 0)
    {
      // there are no additional vectors...
      ov_result.clear ();
    }
  else
    {
      // compare other vectors
      for (auto & [vec_name, vec] : _vectors)
        {
          if (verbose)
            libMesh::out << "   comparing vector \""
                         << vec_name << "\" ...";

          // assume they have the same name
          const NumericVector<Number> & other_system_vector =
            other_system.get_vector(vec_name);

          ov_result.push_back(vec->compare(other_system_vector, threshold));

          if (verbose)
            {
              if (ov_result[ov_result.size()-1] == -1)
                libMesh::out << " identical up to threshold." << std::endl;
              else
                libMesh::out << " first difference occurred at" << std::endl
                             << "   index = " << ov_result[ov_result.size()-1] << "." << std::endl;
            }
        }
    } // finished comparing additional vectors


  bool overall_result;

  // sum up the results
  if ((name_result==0) && (solu_result==-1))
    {
      if (ov_result.size()==0)
        overall_result = true;
      else
        {
          bool ov_identical;
          unsigned int n    = 0;
          do
            {
              ov_identical = (ov_result[n]==-1);
              n++;
            }
          while (ov_identical && n<ov_result.size());
          overall_result = ov_identical;
        }
    }
  else
    overall_result = false;

  if (verbose)
    {
      libMesh::out << "   finished comparisons, ";
      if (overall_result)
        libMesh::out << "found no differences." << std::endl << std::endl;
      else
        libMesh::out << "found differences." << std::endl << std::endl;
    }

  return overall_result;
}

compute_Fq_representor_innerprods()

void libMesh::RBConstruction::compute_Fq_representor_innerprods ( bool  compute_inner_products = true )

protectedvirtual

Compute the terms that are combined ‘online’ to determine the dual norm of the residual.

Here we compute the terms associated with the right-hand side. These terms are basis independent, hence we separate them from the rest of the calculations that are done in update_residual_terms. By default, inner product terms are also computed, but you can turn this feature off e.g. if you are already reading in that data from files.

Definition at line 2186 of file rb_construction.C.

References libMesh::NumericVector< T >::add(), assert_convergence, libMesh::NumericVector< T >::build(), check_convergence(), libMesh::ParallelObject::comm(), libMesh::LinearImplicitSystem::final_linear_residual(), Fq_representor, libMesh::RBEvaluation::Fq_representor_innerprods, Fq_representor_innerprods, Fq_representor_innerprods_computed, get_Fq(), libMesh::RBThetaExpansion::get_n_F_terms(), get_non_dirichlet_inner_product_matrix_if_avail(), get_rb_evaluation(), get_rb_theta_expansion(), libMesh::Utility::get_timestamp(), inner_product_matrix, inner_product_solver, libMesh::RBConstructionBase< LinearImplicitSystem >::inner_product_storage_vector, libMesh::RBConstructionBase< LinearImplicitSystem >::is_quiet(), libMesh::libmesh_assert(), libMesh::System::n_dofs(), libMesh::LinearImplicitSystem::n_linear_iterations(), libMesh::System::n_local_dofs(), libMesh::out, libMesh::PARALLEL, libMesh::ExplicitSystem::rhs, libMesh::System::solution, solve_for_matrix_and_rhs(), libMesh::SparseMatrix< T >::vector_mult(), and libMesh::NumericVector< T >::zero().

Referenced by recompute_all_residual_terms(), and train_reduced_basis_with_greedy().

{

  // Skip calculations if we've already computed the Fq_representors
  if (!Fq_representor_innerprods_computed)
    {
      // Only log if we get to here
      LOG_SCOPE("compute_Fq_representor_innerprods()", "RBConstruction");

      for (unsigned int q_f=0; q_f<get_rb_theta_expansion().get_n_F_terms(); q_f++)
        {
          if (!Fq_representor[q_f])
            {
              Fq_representor[q_f] = NumericVector<Number>::build(this->comm());
              Fq_representor[q_f]->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
            }

          libmesh_assert(Fq_representor[q_f]->size()       == this->n_dofs()       &&
                         Fq_representor[q_f]->local_size() == this->n_local_dofs() );

          rhs->zero();
          rhs->add(1., *get_Fq(q_f));

          if (!is_quiet())
            libMesh::out << "Starting solve q_f=" << q_f
                         << " in RBConstruction::update_residual_terms() at "
                         << Utility::get_timestamp() << std::endl;

          solve_for_matrix_and_rhs(*inner_product_solver, *inner_product_matrix, *rhs);

          if (assert_convergence)
            check_convergence(*inner_product_solver);

          if (!is_quiet())
            {
              libMesh::out << "Finished solve q_f=" << q_f
                           << " in RBConstruction::update_residual_terms() at "
                           << Utility::get_timestamp() << std::endl;

              libMesh::out << this->n_linear_iterations()
                           << " iterations, final residual "
                           << this->final_linear_residual() << std::endl;
            }

          *Fq_representor[q_f] = *solution;
        }

      if (compute_inner_products)
        {
          unsigned int q=0;

          for (unsigned int q_f1=0; q_f1<get_rb_theta_expansion().get_n_F_terms(); q_f1++)
            {
              get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*inner_product_storage_vector, *Fq_representor[q_f1]);

              for (unsigned int q_f2=q_f1; q_f2<get_rb_theta_expansion().get_n_F_terms(); q_f2++)
                {
                  Fq_representor_innerprods[q] = inner_product_storage_vector->dot(*Fq_representor[q_f2]);

                  q++;
                }
            }
        } // end if (compute_inner_products)

      Fq_representor_innerprods_computed = true;
    }

  get_rb_evaluation().Fq_representor_innerprods = Fq_representor_innerprods;
}

compute_max_error_bound()

Real libMesh::RBConstruction::compute_max_error_bound ( )

virtual

(i) Compute the a posteriori error bound for each set of parameters in the training set, (ii) set current_parameters to the parameters that maximize the error bound, and (iii) return the maximum error bound.

Definition at line 1833 of file rb_construction.C.

References libMesh::RBConstructionBase< LinearImplicitSystem >::broadcast_parameters(), libMesh::ParallelObject::comm(), libMesh::RBConstructionBase< LinearImplicitSystem >::get_first_local_training_index(), libMesh::RBConstructionBase< LinearImplicitSystem >::get_global_max_error_pair(), libMesh::RBConstructionBase< LinearImplicitSystem >::get_last_local_training_index(), libMesh::RBConstructionBase< LinearImplicitSystem >::get_local_n_training_samples(), libMesh::RBParametrized::get_n_params(), get_preevaluate_thetas_flag(), get_RB_error_bound(), get_rb_evaluation(), libMesh::ParallelObject::processor_id(), libMesh::Real, libMesh::RBConstructionBase< LinearImplicitSystem >::serial_training_set, set_current_training_parameter_index(), libMesh::RBConstructionBase< LinearImplicitSystem >::set_params_from_training_set(), TIMPI::Communicator::sum(), and training_error_bounds.

Referenced by train_reduced_basis_with_greedy().

{
  LOG_SCOPE("compute_max_error_bound()", "RBConstruction");

  // Treat the case with no parameters in a special way
  if (get_n_params() == 0)
    {
      Real max_val;
      if (std::numeric_limits<Real>::has_infinity)
        {
          max_val = std::numeric_limits<Real>::infinity();
        }
      else
        {
          max_val = std::numeric_limits<Real>::max();
        }

      // Make sure we do at least one solve, but otherwise return a zero error bound
      // when we have no parameters
      return (get_rb_evaluation().get_n_basis_functions() == 0) ? max_val : 0.;
    }

  training_error_bounds.resize(this->get_local_n_training_samples());

  // keep track of the maximum error
  unsigned int max_err_index = 0;
  Real max_err = 0.;

  numeric_index_type first_index = get_first_local_training_index();
  for (unsigned int i=0; i<get_local_n_training_samples(); i++)
    {
      // Load training parameter i, this is only loaded
      // locally since the RB solves are local.
      set_params_from_training_set( first_index+i );

      // In case we pre-evaluate the theta functions,
      // also keep track of the current training parameter index.
      if (get_preevaluate_thetas_flag())
        set_current_training_parameter_index(first_index+i);


      training_error_bounds[i] = get_RB_error_bound();

      if (training_error_bounds[i] > max_err)
        {
          max_err_index = i;
          max_err = training_error_bounds[i];
        }
    }

  std::pair<numeric_index_type, Real> error_pair(first_index+max_err_index, max_err);
  get_global_max_error_pair(this->comm(),error_pair);

  // If we have a serial training set (i.e. a training set that is the same on all processors)
  // just set the parameters on all processors
  if (serial_training_set)
    {
      set_params_from_training_set( error_pair.first );
    }
  // otherwise, broadcast the parameter that produced the maximum error
  else
    {
      unsigned int root_id=0;
      if ((get_first_local_training_index() <= error_pair.first) &&
          (error_pair.first < get_last_local_training_index()))
        {
          set_params_from_training_set( error_pair.first );
          root_id = this->processor_id();
        }

      this->comm().sum(root_id); // root_id is only non-zero on one processor
      broadcast_parameters(root_id);
    }

  return error_pair.second;
}

compute_output_dual_innerprods()

void libMesh::RBConstruction::compute_output_dual_innerprods ( )

protectedvirtual

Compute and store the dual norm of each output functional.

Definition at line 2091 of file rb_construction.C.

References libMesh::NumericVector< T >::add(), assert_convergence, libMesh::NumericVector< T >::build(), check_convergence(), libMesh::ParallelObject::comm(), libMesh::LinearImplicitSystem::final_linear_residual(), libMesh::LinearImplicitSystem::get_linear_solver(), get_matrix_for_output_dual_solves(), libMesh::RBThetaExpansion::get_n_output_terms(), libMesh::RBThetaExpansion::get_n_outputs(), get_output_vector(), get_rb_evaluation(), get_rb_theta_expansion(), libMesh::Utility::get_timestamp(), libMesh::RBConstructionBase< LinearImplicitSystem >::inner_product_storage_vector, libMesh::RBConstructionBase< LinearImplicitSystem >::is_quiet(), libMesh::ImplicitSystem::linear_solver, libMesh::System::n_dofs(), libMesh::LinearImplicitSystem::n_linear_iterations(), libMesh::System::n_local_dofs(), libMesh::out, libMesh::RBEvaluation::output_dual_innerprods, output_dual_innerprods, output_dual_innerprods_computed, libMesh::PARALLEL, libMesh::ExplicitSystem::rhs, libMesh::System::solution, solve_for_matrix_and_rhs(), libMesh::SparseMatrix< T >::vector_mult(), and libMesh::NumericVector< T >::zero().

Referenced by train_reduced_basis_with_greedy().

{
  // Skip calculations if we've already computed the output dual norms
  if (!output_dual_innerprods_computed)
    {
      // Short circuit if we don't have any outputs
      if (get_rb_theta_expansion().get_n_outputs() == 0)
        {
          output_dual_innerprods_computed = true;
          return;
        }

      // Only log if we get to here
      LOG_SCOPE("compute_output_dual_innerprods()", "RBConstruction");

      libMesh::out << "Compute output dual inner products" << std::endl;

      // Find out the largest value of Q_l
      unsigned int max_Q_l = 0;
      for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
        max_Q_l = (get_rb_theta_expansion().get_n_output_terms(n) > max_Q_l) ? get_rb_theta_expansion().get_n_output_terms(n) : max_Q_l;

      std::vector<std::unique_ptr<NumericVector<Number>>> L_q_representor(max_Q_l);
      for (unsigned int q=0; q<max_Q_l; q++)
        {
          L_q_representor[q] = NumericVector<Number>::build(this->comm());
          L_q_representor[q]->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
        }

      for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
        {
          for (unsigned int q_l=0; q_l<get_rb_theta_expansion().get_n_output_terms(n); q_l++)
            {
              rhs->zero();
              rhs->add(1., *get_output_vector(n,q_l));

              if (!is_quiet())
                libMesh::out << "Starting solve n=" << n << ", q_l=" << q_l
                             << " in RBConstruction::compute_output_dual_innerprods() at "
                             << Utility::get_timestamp() << std::endl;

              // Use the main linear solver here instead of the inner_product solver, since
              // get_matrix_for_output_dual_solves() may not return the inner product matrix.
              solve_for_matrix_and_rhs(*get_linear_solver(), get_matrix_for_output_dual_solves(), *rhs);

              // We possibly perform multiple solves here with the same matrix, hence
              // set reuse_preconditioner(true) (and set it back to false again below
              // at the end of this function).
              linear_solver->reuse_preconditioner(true);

              if (assert_convergence)
                check_convergence(*get_linear_solver());

              if (!is_quiet())
                {
                  libMesh::out << "Finished solve n=" << n << ", q_l=" << q_l
                               << " in RBConstruction::compute_output_dual_innerprods() at "
                               << Utility::get_timestamp() << std::endl;

                  libMesh::out << this->n_linear_iterations()
                               << " iterations, final residual "
                               << this->final_linear_residual() << std::endl;
                }

              *L_q_representor[q_l] = *solution;
            }

          unsigned int q=0;
          for (unsigned int q_l1=0; q_l1<get_rb_theta_expansion().get_n_output_terms(n); q_l1++)
            {
              get_matrix_for_output_dual_solves().vector_mult(*inner_product_storage_vector, *L_q_representor[q_l1]);

              for (unsigned int q_l2=q_l1; q_l2<get_rb_theta_expansion().get_n_output_terms(n); q_l2++)
                {
                  output_dual_innerprods[n][q] = L_q_representor[q_l2]->dot(*inner_product_storage_vector);
                  libMesh::out << "output_dual_innerprods[" << n << "][" << q << "] = " << output_dual_innerprods[n][q] << std::endl;

                  q++;
                }
            }
        }

      // We may not need to use linear_solver again (e.g. this would happen if we use
      // extra_linear_solver for the truth_solves). As a result, let's clear linear_solver
      // to release any memory it may be taking up. If we do need it again, it will
      // be initialized when necessary.
      linear_solver->clear();
      linear_solver->reuse_preconditioner(false);

      output_dual_innerprods_computed = true;
    }

  get_rb_evaluation().output_dual_innerprods = output_dual_innerprods;
}

compute_residual_dual_norm_slow()

Real libMesh::RBConstruction::compute_residual_dual_norm_slow

(

const unsigned int

N

)

The slow (but simple, non-error prone) way to compute the residual dual norm.

Useful for error checking.

Definition at line 2275 of file rb_construction.C.

References _untransformed_basis_functions, libMesh::SparseMatrix< T >::add(), libMesh::NumericVector< T >::add(), libMesh::NumericVector< T >::build(), libMesh::ParallelObject::comm(), libMesh::RBEvaluation::get_n_basis_functions(), get_non_dirichlet_inner_product_matrix_if_avail(), get_rb_evaluation(), inner_product_matrix, inner_product_solver, libMesh::RBConstructionBase< LinearImplicitSystem >::inner_product_storage_vector, libMesh::libmesh_real(), libMesh::ImplicitSystem::matrix, libMesh::System::n_dofs(), libMesh::System::n_local_dofs(), libMesh::PARALLEL, libMesh::ExplicitSystem::rhs, libMesh::System::solution, solve_for_matrix_and_rhs(), store_untransformed_basis, truth_assembly(), libMesh::SparseMatrix< T >::vector_mult(), and libMesh::SparseMatrix< T >::zero().

{
  LOG_SCOPE("compute_residual_dual_norm_slow()", "RBConstruction");

  // Put the residual in rhs in order to compute the norm of the Riesz representor
  // Note that this only works in serial since otherwise each processor will
  // have a different parameter value during the Greedy training.

  std::unique_ptr<NumericVector<Number>> RB_sol = NumericVector<Number>::build(comm());
  RB_sol->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);

  std::unique_ptr<NumericVector<Number>> temp = NumericVector<Number>::build(comm());
  temp->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);

  if (store_untransformed_basis)
    {
      libmesh_assert_equal_to(_untransformed_basis_functions.size(), get_rb_evaluation().get_n_basis_functions());
    }

  for (unsigned int i=0; i<N; i++)
  {
    if (!store_untransformed_basis)
      {
        RB_sol->add(get_rb_evaluation().RB_solution(i), get_rb_evaluation().get_basis_function(i));
      }
    else
      {
        RB_sol->add(get_rb_evaluation().RB_solution(i), *_untransformed_basis_functions[i]);
      }
  }

  this->truth_assembly();
  matrix->vector_mult(*temp, *RB_sol);
  rhs->add(-1., *temp);

  // Then solve to get the Reisz representor
  matrix->zero();
  matrix->add(1., *inner_product_matrix);

  solve_for_matrix_and_rhs(*inner_product_solver, *inner_product_matrix, *rhs);
  get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*inner_product_storage_vector, *solution);
  Number slow_residual_norm_sq = solution->dot(*inner_product_storage_vector);

  return std::sqrt( libmesh_real(slow_residual_norm_sq) );
}

condense_constrained_dofs()

virtual bool libMesh::System::condense_constrained_dofs ( ) const

inlineprotectedvirtualinherited

Whether this object should condense out constrained degrees of freedom.

Reimplemented in libMesh::CondensedEigenSystem.

Definition at line 2059 of file system.h.

Referenced by libMesh::EigenSystem::init_matrices().

{ return false; }

create_static_condensation()

void libMesh::LinearImplicitSystem::create_static_condensation ( )

overridevirtualinherited

Request that static condensation be performed for this system.

Reimplemented from libMesh::ImplicitSystem.

Definition at line 63 of file linear_implicit_system.C.

References libMesh::ImplicitSystem::create_static_condensation(), libMesh::ImplicitSystem::linear_solver, and libMesh::ImplicitSystem::setup_static_condensation_preconditioner().

{
  Parent::create_static_condensation();
  this->setup_static_condensation_preconditioner(*linear_solver);
}

current_solution()

Number libMesh::System::current_solution ( const dof_id_type  global_dof_number ) const

inherited

Returns

The current solution for the specified global DOF.

Definition at line 162 of file system.C.

References libMesh::System::_dof_map, and libMesh::System::current_local_solution.

Referenced by libMesh::ExactSolution::_compute_error(), libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::HPCoarsenTest::add_projection(), compute_stresses(), LinearElasticityWithContact::compute_stresses(), LinearElasticity::compute_stresses(), LargeDeformationElasticity::compute_stresses(), libMesh::ExactErrorEstimator::estimate_error(), main(), libMesh::WeightedPatchRecoveryErrorEstimator::EstimateError::operator()(), libMesh::SmoothnessEstimator::EstimateSmoothness::operator()(), libMesh::PatchRecoveryErrorEstimator::EstimateError::operator()(), libMesh::HPCoarsenTest::select_refinement(), SolidSystem::side_time_derivative(), libMesh::EnsightIO::write_scalar_ascii(), and libMesh::EnsightIO::write_vector_ascii().

{
  // Check the sizes
  libmesh_assert_less (global_dof_number, _dof_map->n_dofs());
  libmesh_assert_less (global_dof_number, current_local_solution->size());

  return (*current_local_solution)(global_dof_number);
}

deactivate()

void libMesh::System::deactivate ( )

inlineinherited

Deactivates the system.

Only active systems are solved.

Definition at line 2449 of file system.h.

References libMesh::System::_active.

{
  _active = false;
}

detach_shell_matrix()

void libMesh::LinearImplicitSystem::detach_shell_matrix ( )

inlineinherited

Detaches a shell matrix.

Same as attach_shell_matrix(nullptr).

Definition at line 176 of file linear_implicit_system.h.

References libMesh::LinearImplicitSystem::attach_shell_matrix().

Referenced by main().

{ attach_shell_matrix(nullptr); }

disable_cache()

void libMesh::ImplicitSystem::disable_cache ( )

overridevirtualinherited

Avoids use of any cached data that might affect any solve result.

Should be overridden in derived systems.

Reimplemented from libMesh::System.

Definition at line 132 of file implicit_system.C.

References libMesh::System::assemble_before_solve, libMesh::ImplicitSystem::get_linear_solver(), and libMesh::LinearSolver< T >::reuse_preconditioner().

Referenced by libMesh::DifferentiableSystem::pop_physics(), and libMesh::DifferentiableSystem::push_physics().

                                    {
  this->assemble_before_solve = true;
  this->get_linear_solver()->reuse_preconditioner(false);
}

disable_print_counter_info() [1/2]

void libMesh::ReferenceCounter::disable_print_counter_info ( )

staticinherited

Definition at line 100 of file reference_counter.C.

References libMesh::ReferenceCounter::_enable_print_counter.

{
  _enable_print_counter = false;
  return;
}

disable_print_counter_info() [2/2]

void libMesh::ReferenceCounter::disable_print_counter_info ( )

staticinherited

Definition at line 100 of file reference_counter.C.

References libMesh::ReferenceCounter::_enable_print_counter.

{
  _enable_print_counter = false;
  return;
}

enable_print_counter_info() [1/2]

void libMesh::ReferenceCounter::enable_print_counter_info ( )

staticinherited

Methods to enable/disable the reference counter output from print_info()

Definition at line 94 of file reference_counter.C.

References libMesh::ReferenceCounter::_enable_print_counter.

{
  _enable_print_counter = true;
  return;
}

enable_print_counter_info() [2/2]

void libMesh::ReferenceCounter::enable_print_counter_info ( )

staticinherited

Methods to enable/disable the reference counter output from print_info()

Definition at line 94 of file reference_counter.C.

References libMesh::ReferenceCounter::_enable_print_counter.

{
  _enable_print_counter = true;
  return;
}

enrich_basis_from_rhs_terms()

void libMesh::RBConstruction::enrich_basis_from_rhs_terms

(

const bool

resize_rb_eval_data = true

)

This function computes one basis function for each rhs term.

This is useful in some cases since we can avoid doing a full greedy if we know that we do not have any "left-hand side" parameters, for example.

Definition at line 1331 of file rb_construction.C.

References assert_convergence, check_convergence(), enrich_RB_space(), extra_linear_solver, libMesh::System::get_equation_systems(), get_Fq(), libMesh::LinearImplicitSystem::get_linear_solver(), libMesh::System::get_mesh(), libMesh::RBThetaExpansion::get_n_F_terms(), get_Nmax(), get_rb_evaluation(), get_rb_theta_expansion(), libMesh::RBParametrized::initialize_parameters(), libMesh::NumericVector< T >::l2_norm(), libMesh::ImplicitSystem::matrix, libMesh::System::name(), libMesh::out, post_process_truth_solution(), libMesh::NumericVector< T >::print_matlab(), libMesh::RBEvaluation::resize_data_structures(), libMesh::ExplicitSystem::rhs, solve_for_matrix_and_rhs(), truth_assembly(), update_system(), and libMesh::ExodusII_IO::write_equation_systems().

{
  LOG_SCOPE("enrich_basis_from_rhs_terms()", "RBConstruction");

  // initialize rb_eval's parameters
  get_rb_evaluation().initialize_parameters(*this);

  // possibly resize data structures according to Nmax
  if (resize_rb_eval_data)
    {
      get_rb_evaluation().resize_data_structures(get_Nmax());
    }

  libMesh::out << std::endl << "---- Enriching basis from rhs terms ----" << std::endl;

  truth_assembly();

  for (unsigned int q_f=0; q_f<get_rb_theta_expansion().get_n_F_terms(); q_f++)
    {
      libMesh::out << std::endl << "Performing truth solve with rhs from rhs term " << q_f << std::endl;

      *rhs = *get_Fq(q_f);

      if (rhs->l2_norm() == 0)
      {
        // Skip enrichment if the rhs is zero
        continue;
      }

      // truth_assembly assembles into matrix and rhs, so use those for the solve
      if (extra_linear_solver)
        {
          // If extra_linear_solver has been initialized, then we use it for the
          // truth solves.
          solve_for_matrix_and_rhs(*extra_linear_solver, *matrix, *rhs);

          if (assert_convergence)
            check_convergence(*extra_linear_solver);
        }
      else
        {
          solve_for_matrix_and_rhs(*get_linear_solver(), *matrix, *rhs);

          if (assert_convergence)
            check_convergence(*get_linear_solver());
        }

      // Debugging: enable this code to print the rhs that was used in
      // the most recent truth solve to a uniquely-named file.
#if 0
        {
          char temp_file[] = "truth_rhs_XXXXXX.dat";
          int fd = mkstemps(temp_file, 4);
          if (fd != -1)
            {
              libMesh::out << "Writing truth system rhs to file: " << temp_file << std::endl;
              rhs->print_matlab(std::string(temp_file));
            }
        }
#endif // 0

      // Debugging: enable this code to print the most recent truth
      // solution to a uniquely-named file.
#ifdef LIBMESH_HAVE_EXODUS_API
#if 0
        {
          // Note: mkstemps creates a file and returns an open file descriptor to it.
          // The filename is created from a template which must have 6 'X' characters followed
          // by a suffix having the specified length (in this case 4, for ".exo").
          char temp_file[] = "truth_XXXXXX.exo";
          int fd = mkstemps(temp_file, 4);
          if (fd != -1)
            {
              libMesh::out << "Writing truth solution to file: " << temp_file << std::endl;
              ExodusII_IO exo_io(this->get_mesh());
              std::set<std::string> system_names = {this->name()};
              exo_io.write_equation_systems(std::string(temp_file),
                                            this->get_equation_systems(), &system_names);
            }
        }
#endif // 0
#endif // LIBMESH_HAVE_EXODUS_API

      // Call user-defined post-processing routines on the truth solution.
      post_process_truth_solution();

      // Add orthogonal part of the snapshot to the RB space
      libMesh::out << "Enriching the RB space" << std::endl;
      enrich_RB_space();

      update_system();
    }
}

enrich_RB_space()

void libMesh::RBConstruction::enrich_RB_space ( )

protectedvirtual

Add a new basis function to the RB space.

This is called during train_reduced_basis.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 1719 of file rb_construction.C.

References _untransformed_basis_functions, _untransformed_solution, libMesh::RBEvaluation::basis_functions, libMesh::RBEvaluation::get_n_basis_functions(), get_non_dirichlet_inner_product_matrix_if_avail(), get_rb_evaluation(), libMesh::RBConstructionBase< LinearImplicitSystem >::inner_product_storage_vector, libMesh::System::solution, store_untransformed_basis, and libMesh::SparseMatrix< T >::vector_mult().

Referenced by enrich_basis_from_rhs_terms(), and train_reduced_basis_with_greedy().

{
  LOG_SCOPE("enrich_RB_space()", "RBConstruction");

  auto new_bf = solution->clone();

  std::unique_ptr<NumericVector<Number>> new_untransformed_bf;
  if (store_untransformed_basis)
    {
      new_untransformed_bf = _untransformed_solution->clone();
      libmesh_assert_equal_to(_untransformed_basis_functions.size(), get_rb_evaluation().get_n_basis_functions());
    }

  for (unsigned int index=0; index<get_rb_evaluation().get_n_basis_functions(); index++)
    {
      get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*inner_product_storage_vector, *new_bf);

      Number scalar =
        inner_product_storage_vector->dot(get_rb_evaluation().get_basis_function(index));
      new_bf->add(-scalar, get_rb_evaluation().get_basis_function(index));

      if (store_untransformed_basis)
        {
          new_untransformed_bf->add(-scalar, *_untransformed_basis_functions[index]);
        }
    }

  // Normalize new_bf
  get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*inner_product_storage_vector, *new_bf);
  Number new_bf_norm = std::sqrt( inner_product_storage_vector->dot(*new_bf) );

  if (new_bf_norm == 0.)
    {
      new_bf->zero(); // avoid potential nan's

      if (store_untransformed_basis)
        {
          new_untransformed_bf->zero();
        }
    }
  else
    {
      new_bf->scale(1./new_bf_norm);

      if (store_untransformed_basis)
        {
          new_untransformed_bf->scale(1./new_bf_norm);
        }
    }

  // load the new basis function into the basis_functions vector.
  get_rb_evaluation().basis_functions.emplace_back( std::move(new_bf) );

  if (store_untransformed_basis)
    {
      _untransformed_basis_functions.emplace_back( std::move(new_untransformed_bf) );
    }
}

final_linear_residual()

Real libMesh::LinearImplicitSystem::final_linear_residual ( ) const

inlineinherited

Returns

The final residual for the linear system solve.

Definition at line 160 of file linear_implicit_system.h.

References libMesh::LinearImplicitSystem::_final_linear_residual.

Referenced by compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), main(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

{ return _final_linear_residual; }

forward_qoi_parameter_sensitivity()

void libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity ( const QoISet &  qoi_indices, const ParameterVector &  parameters, SensitivityData &  sensitivities  )

overridevirtualinherited

Solves for the derivative of each of the system's quantities of interest q in qoi[qoi_indices] with respect to each parameter in parameters, placing the result for qoi i and parameter j into sensitivities[i][j].

Uses the forward sensitivity method.

Currently uses finite differenced derivatives (partial q / partial p) and (partial R / partial p).

Reimplemented from libMesh::System.

Definition at line 627 of file implicit_system.C.

References libMesh::SensitivityData::allocate_data(), libMesh::ExplicitSystem::assemble_qoi(), libMesh::ExplicitSystem::assemble_qoi_derivative(), libMesh::ImplicitSystem::assembly(), libMesh::NumericVector< T >::close(), libMesh::SparseMatrix< T >::close(), libMesh::NumericVector< T >::dot(), libMesh::System::get_adjoint_rhs(), libMesh::System::get_qoi_values(), libMesh::System::get_sensitivity_solution(), libMesh::QoISet::has_index(), libMesh::make_range(), libMesh::ImplicitSystem::matrix, libMesh::System::n_qois(), libMesh::Real, libMesh::ExplicitSystem::rhs, libMesh::ImplicitSystem::sensitivity_solve(), libMesh::ParameterVector::size(), and libMesh::TOLERANCE.

Referenced by main().

{
  ParameterVector & parameters_vec =
    const_cast<ParameterVector &>(parameters_in);

  const unsigned int Np = cast_int<unsigned int>
    (parameters_vec.size());
  const unsigned int Nq = this->n_qois();

  // An introduction to the problem:
  //
  // Residual R(u(p),p) = 0
  // partial R / partial u = J = system matrix
  //
  // This implies that:
  // d/dp(R) = 0
  // (partial R / partial p) +
  // (partial R / partial u) * (partial u / partial p) = 0

  // We first solve for (partial u / partial p) for each parameter:
  // J * (partial u / partial p) = - (partial R / partial p)

  this->sensitivity_solve(parameters_vec);

  // Get ready to fill in sensitivities:
  sensitivities.allocate_data(qoi_indices, *this, parameters_vec);

  // We use the identity:
  // dq/dp = (partial q / partial p) + (partial q / partial u) *
  //         (partial u / partial p)

  // We get (partial q / partial u) from the user
  this->assemble_qoi_derivative(qoi_indices,
                                /* include_liftfunc = */ true,
                                /* apply_constraints = */ false);

  // We don't need these to be closed() in this function, but libMesh
  // standard practice is to have them closed() by the time the
  // function exits
  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      this->get_adjoint_rhs(i).close();

  for (unsigned int j=0; j != Np; ++j)
    {
      // We currently get partial derivatives via central differencing

      // (partial q / partial p) ~= (q(p+dp)-q(p-dp))/(2*dp)

      Number old_parameter = *parameters_vec[j];

      const Real delta_p =
        TOLERANCE * std::max(std::abs(old_parameter), 1e-3);

      *parameters_vec[j] = old_parameter - delta_p;
      this->assemble_qoi(qoi_indices);
      const std::vector<Number> qoi_minus = this->get_qoi_values();

      *parameters_vec[j] = old_parameter + delta_p;
      this->assemble_qoi(qoi_indices);
      const std::vector<Number> qoi_plus = this->get_qoi_values();

      std::vector<Number> partialq_partialp(Nq, 0);
      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          partialq_partialp[i] = (qoi_plus[i] - qoi_minus[i]) / (2.*delta_p);

      // Don't leave the parameter changed
      *parameters_vec[j] = old_parameter;

      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          sensitivities[i][j] = partialq_partialp[i] +
            this->get_adjoint_rhs(i).dot(this->get_sensitivity_solution(j));
    }

  // All parameters_vec have been reset.
  // We didn't cache the original rhs or matrix for memory reasons,
  // but we can restore them to a state consistent solution -
  // principle of least surprise.
  this->assembly(true, true);
  this->rhs->close();
  this->matrix->close();
  this->assemble_qoi(qoi_indices);
}

generate_training_parameters_deterministic()

std::pair< std::size_t, std::size_t > libMesh::RBConstructionBase< LinearImplicitSystem >::generate_training_parameters_deterministic ( const Parallel::Communicator &  communicator, const std::map< std::string, bool > &  log_param_scale, std::map< std::string, std::vector< RBParameter >> &  local_training_parameters_in, const unsigned int  n_global_training_samples_in, const RBParameters &  min_parameters, const RBParameters &  max_parameters, const bool  serial_training_set = false  )

staticinherited

Static helper function for generating a deterministic set of parameters.

Only works with 1 or 2 parameters (as defined by the lengths of min/max parameters vectors), otherwise throws an error. The parameter n_global_training_samples_in is the total number of parameters to generate, and they will be split across all the processors (unless serial_training_set=true) in the local_training_parameters_in map.

Returns

a pair of {first_local_index,last_local_index}

Definition at line 572 of file rb_construction_base.C.

{
  libmesh_assert_equal_to ( min_parameters.n_parameters(), max_parameters.n_parameters() );
  const unsigned int num_params = min_parameters.n_parameters();

  if (num_params == 0)
    return {0,0};

  if (num_params > 3)
    libmesh_not_implemented_msg("ERROR: Deterministic training sample generation "
                                "not implemented for more than three parameters.");

  // TODO - we don't support vector-data here yet. This would only apply in the case where
  //        min or max are vector-valued, and all the generated points need to stay within those ranges.
  //        But typically we expect that if we're calling this function, we only have 1 min and 1 max,
  //        so the generated values are single-valued as well. The .get_value() calls will throw an error
  //        if this is not the case.

  // Reinitialize training_parameters_in (but don't remove existing keys!)
  const auto &[n_local_training_samples, first_local_index] =
      calculate_n_local_samples_and_index(communicator, n_global_training_samples_in,
                                         serial_training_set);
  const auto last_local_index = first_local_index + n_local_training_samples;
  for (const auto & pr : min_parameters)
    local_training_parameters_in[pr.first] = std::vector<RBParameter>(n_local_training_samples);

  // n_training_samples_per_param has 3 entries, but entries after "num_params"
  // are unused so we just set their value to 1. We need to set it to 1 (rather
  // than 0) so that we don't skip the inner part of the triply-nested loop over
  // n_training_samples_per_param below.
  std::vector<unsigned int> n_training_samples_per_param(3);
  for (unsigned int param=0; param<3; param++)
    {
      if (param < num_params)
        {
          n_training_samples_per_param[param] =
            static_cast<unsigned int>( std::round(std::pow(static_cast<Real>(n_global_training_samples_in), 1./num_params)) );
        }
      else
        {
          n_training_samples_per_param[param] = 1;
        }
    }

  {
    // The current implementation assumes that we have the same number of
    // samples in each parameter, so we check that n_training_samples_in
    // is consistent with this assumption.
    unsigned int total_samples_check = 1;
    for (unsigned int n_samples : n_training_samples_per_param)
      {
        total_samples_check *= n_samples;
      }

    libmesh_error_msg_if(total_samples_check != n_global_training_samples_in,
                         "Error: Number of training samples = "
                         << n_global_training_samples_in
                         << " does not enable a uniform grid of samples with "
                         << num_params << " parameters. Try "
                         << total_samples_check << " samples instead?");
  }

  // First we make a list of training samples associated with each parameter,
  // then we take a tensor product to obtain the final set of training samples.
  std::vector<std::vector<Real>> training_samples_per_param(num_params);
  {
    unsigned int i = 0;
    for (const auto & pr : min_parameters)
      {
        const std::string & param_name = pr.first;
        const bool use_log_scaling = libmesh_map_find(log_param_scale, param_name);
        Real min_param = min_parameters.get_value(param_name);
        Real max_param = max_parameters.get_value(param_name);

        training_samples_per_param[i].resize(n_training_samples_per_param[i]);

        for (unsigned int j=0; j<n_training_samples_per_param[i]; j++)
          {
            // Generate log10 scaled training parameters
            if (use_log_scaling)
              {
                Real epsilon = 1.e-6; // Prevent rounding errors triggering asserts
                Real log_min   = std::log10(min_param + epsilon);
                Real log_range = std::log10( (max_param-epsilon) / (min_param+epsilon) );
                Real step_size = log_range /
                  std::max((unsigned int)1,(n_training_samples_per_param[i]-1));

                if (j<(n_training_samples_per_param[i]-1))
                  {
                    training_samples_per_param[i][j] = std::pow(10., log_min + j*step_size );
                  }
                else
                  {
                    // due to rounding error, the last parameter can be slightly
                    // bigger than max_parameters, hence snap back to the max
                    training_samples_per_param[i][j] = max_param;
                  }
              }
            else
              {
                // Generate linearly scaled training parameters
                Real step_size = (max_param - min_param) /
                  std::max((unsigned int)1,(n_training_samples_per_param[i]-1));
                training_samples_per_param[i][j] = j*step_size + min_param;
              }

          }
        i++;
      }
  }

  // Now load into training_samples_in
  {
    std::vector<unsigned int> indices(3);
    unsigned int index_count = 0;
    for (indices[0]=0; indices[0]<n_training_samples_per_param[0]; indices[0]++)
      {
        for (indices[1]=0; indices[1]<n_training_samples_per_param[1]; indices[1]++)
          {
            for (indices[2]=0; indices[2]<n_training_samples_per_param[2]; indices[2]++)
              {
                unsigned int param_count = 0;
                for (const auto & pr : min_parameters)
                  {
                    std::vector<RBParameter> & training_vector =
                      libmesh_map_find(local_training_parameters_in, pr.first);
                    if (first_local_index <= index_count && index_count < last_local_index)
                      training_vector[index_count - first_local_index] =
                        {training_samples_per_param[param_count][indices[param_count]]};

                    param_count++;
                  }
                index_count++;
              }
          }
      }
  }
  return {first_local_index, first_local_index+n_local_training_samples};
}

generate_training_parameters_random()

std::pair< std::size_t, std::size_t > libMesh::RBConstructionBase< LinearImplicitSystem >::generate_training_parameters_random ( const Parallel::Communicator &  communicator, const std::map< std::string, bool > &  log_param_scale, std::map< std::string, std::vector< RBParameter >> &  local_training_parameters_in, const unsigned int  n_global_training_samples_in, const RBParameters &  min_parameters, const RBParameters &  max_parameters, const int  training_parameters_random_seed = -1, const bool  serial_training_set = false  )

staticinherited

Static helper function for generating a randomized set of parameters.

The parameter n_global_training_samples_in is the total number of parameters to generate, and they will be split across all the processors (unless serial_training_set=true) in the local_training_parameters_in map.

Returns

a pair of {first_local_index,last_local_index}

Definition at line 472 of file rb_construction_base.C.

{
  const unsigned int num_params = min_parameters.n_parameters();
  libmesh_error_msg_if(num_params!=max_parameters.n_parameters(),
    "Number of parameters must be identical for min/max.");

  // Clear training_parameters_in
  local_training_parameters_in.clear();

  if (num_params == 0)
    return {0,0};

  if (training_parameters_random_seed < 0)
    {
      if (!serial_training_set)
        {
          // seed the random number generator with the system time
          // and the processor ID so that the seed is different
          // on different processors
          std::srand( static_cast<unsigned>( std::time(0)*(1+communicator.rank()) ));
        }
      else
        {
          // seed the random number generator with the system time
          // only so that the seed is the same on all processors
          //
          // Note that we broadcast the time on processor 0 to make
          // sure all processors agree.
          unsigned int current_time = static_cast<unsigned>( std::time(0) );
          communicator.broadcast(current_time, 0);
          std::srand(current_time);
        }
    }
  else
    {
      if (!serial_training_set)
        {
          // seed the random number generator with the provided value
          // and the processor ID so that the seed is different
          // on different processors
          std::srand( static_cast<unsigned>( training_parameters_random_seed*(1+communicator.rank()) ));
        }
      else
        {
          // seed the random number generator with the provided value
          // so that the seed is the same on all processors
          std::srand( static_cast<unsigned>( training_parameters_random_seed ));
        }
    }

  // TODO - we don't support vector-data here yet. This would only apply in the case where
  //        min or max are vector-valued, and all the generated points need to stay within those ranges.
  //        But typically we expect that if we're calling this function, we only have 1 min and 1 max,
  //        so the generated values are single-valued as well. The .get_value() calls will throw an error
  //        if this is not the case.

  // initialize training_parameters_in
  const auto & [n_local_training_samples, first_local_index] =
      calculate_n_local_samples_and_index(communicator, n_global_training_samples_in,
                                          serial_training_set);
  for (const auto & pr : min_parameters)
    local_training_parameters_in[pr.first] = std::vector<RBParameter>(n_local_training_samples);

  // finally, set the values
  for (auto & [param_name, sample_vector] : local_training_parameters_in)
    {
      for (auto i : make_range(n_local_training_samples))
        {
          Real random_number = static_cast<Real>(std::rand()) / RAND_MAX; // in range [0,1]

          // Generate log10 scaled training parameters
          if (libmesh_map_find(log_param_scale, param_name))
            {
              Real log_min   = std::log10(min_parameters.get_value(param_name));
              Real log_range = std::log10(max_parameters.get_value(param_name) / min_parameters.get_value(param_name));

              sample_vector[i] = {std::pow(Real(10.), log_min + random_number*log_range )};
            }
          // Generate linearly scaled training parameters
          else
            {
              sample_vector[i] = {
                  random_number * (max_parameters.get_value(param_name) -
                                   min_parameters.get_value(param_name)) +
                  min_parameters.get_value(param_name)};
            }
        }
    }
  return {first_local_index, first_local_index+n_local_training_samples};
}

get_abs_training_tolerance()

Real libMesh::RBConstruction::get_abs_training_tolerance ( ) const

inline

Definition at line 227 of file rb_construction.h.

References abs_training_tolerance.

Referenced by print_info().

{ return abs_training_tolerance; }

get_adjoint_rhs() [1/2]

NumericVector< Number > & libMesh::System::get_adjoint_rhs ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's adjoint rhs vectors, by default the one corresponding to the first qoi. This what the user's QoI derivative code should assemble when setting up an adjoint problem

Definition at line 1294 of file system.C.

References libMesh::System::get_vector().

Referenced by libMesh::ImplicitSystem::adjoint_solve(), libMesh::FEMSystem::assemble_qoi_derivative(), libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity(), libMesh::ImplicitSystem::qoi_parameter_hessian(), libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product(), and libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve().

{
  std::ostringstream adjoint_rhs_name;
  adjoint_rhs_name << "adjoint_rhs" << i;

  return this->get_vector(adjoint_rhs_name.str());
}

get_adjoint_rhs() [2/2]

const NumericVector< Number > & libMesh::System::get_adjoint_rhs ( unsigned int  i = 0 ) const

inherited

Returns

A reference to one of the system's adjoint rhs vectors, by default the one corresponding to the first qoi.

Definition at line 1304 of file system.C.

References libMesh::System::get_vector().

{
  std::ostringstream adjoint_rhs_name;
  adjoint_rhs_name << "adjoint_rhs" << i;

  return this->get_vector(adjoint_rhs_name.str());
}

get_adjoint_solution() [1/2]

NumericVector< Number > & libMesh::System::get_adjoint_solution ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's adjoint solution vectors, by default the one corresponding to the first qoi.

Definition at line 1232 of file system.C.

References libMesh::System::get_vector().

Referenced by libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::AdaptiveTimeSolver::adjoint_advance_timestep(), libMesh::UnsteadySolver::adjoint_advance_timestep(), libMesh::TwostepTimeSolver::adjoint_solve(), libMesh::ImplicitSystem::adjoint_solve(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::AdjointResidualErrorEstimator::estimate_error(), HeatSystem::init_context(), libMesh::Euler2Solver::integrate_adjoint_refinement_error_estimate(), libMesh::EulerSolver::integrate_adjoint_refinement_error_estimate(), main(), libMesh::ImplicitSystem::qoi_parameter_hessian(), libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product(), libMesh::FileSolutionHistory::retrieve(), and libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve().

{
  std::ostringstream adjoint_name;
  adjoint_name << "adjoint_solution" << i;

  return this->get_vector(adjoint_name.str());
}

get_adjoint_solution() [2/2]

const NumericVector< Number > & libMesh::System::get_adjoint_solution ( unsigned int  i = 0 ) const

inherited

Returns

A reference to one of the system's adjoint solution vectors, by default the one corresponding to the first qoi.

Definition at line 1242 of file system.C.

References libMesh::System::get_vector().

{
  std::ostringstream adjoint_name;
  adjoint_name << "adjoint_solution" << i;

  return this->get_vector(adjoint_name.str());
}

get_all_matrices()

void libMesh::RBConstruction::get_all_matrices ( std::map< std::string, SparseMatrix< Number > *> &  all_matrices )

virtual

Get a map that stores pointers to all of the matrices.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 2455 of file rb_construction.C.

References get_Aq(), get_inner_product_matrix(), libMesh::RBThetaExpansion::get_n_A_terms(), get_non_dirichlet_Aq(), get_non_dirichlet_inner_product_matrix(), get_rb_theta_expansion(), store_dirichlet_operators, and store_non_dirichlet_operators.

Referenced by libMesh::TransientRBConstruction::get_all_matrices().

{
  all_matrices.clear();

  if (store_dirichlet_operators)
    all_matrices["inner_product"] = get_inner_product_matrix();

  if (store_non_dirichlet_operators)
    all_matrices["non_dirichlet_inner_product"] = get_non_dirichlet_inner_product_matrix();

  for (unsigned int q_a=0; q_a<get_rb_theta_expansion().get_n_A_terms(); q_a++)
    {
      std::stringstream matrix_name;
      matrix_name << "A" << q_a;

      if (store_dirichlet_operators)
        all_matrices[matrix_name.str()] = get_Aq(q_a);

      if (store_non_dirichlet_operators)
        {
          matrix_name << "_non_dirichlet";
          all_matrices[matrix_name.str()] = get_non_dirichlet_Aq(q_a);
        }
    }
}

get_all_variable_numbers()

void libMesh::System::get_all_variable_numbers ( std::vector< unsigned int > &  all_variable_numbers ) const

inherited

Fills all_variable_numbers with all the variable numbers for the variables that have been added to this system.

Definition at line 1403 of file system.C.

References libMesh::DofMap::get_all_variable_numbers(), and libMesh::System::get_dof_map().

Referenced by MeshFunctionTest::read_variable_info_from_output_data(), MeshfunctionDFEM::test_mesh_function_dfem(), MeshfunctionDFEM::test_mesh_function_dfem_grad(), and SystemsTest::testProjectCubeWithMeshFunction().

{
  this->get_dof_map().get_all_variable_numbers(all_variable_numbers);
}

get_all_vectors()

void libMesh::RBConstruction::get_all_vectors ( std::map< std::string, NumericVector< Number > *> &  all_vectors )

virtual

Get a map that stores pointers to all of the vectors.

Definition at line 2481 of file rb_construction.C.

References get_Fq(), libMesh::RBThetaExpansion::get_n_F_terms(), get_non_dirichlet_Fq(), get_output_vectors(), get_rb_theta_expansion(), store_dirichlet_operators, and store_non_dirichlet_operators.

{
  all_vectors.clear();

  get_output_vectors(all_vectors);

  for (unsigned int q_f=0; q_f<get_rb_theta_expansion().get_n_F_terms(); q_f++)
    {
      std::stringstream F_vector_name;
      F_vector_name << "F" << q_f;

      if (store_dirichlet_operators)
        all_vectors[F_vector_name.str()] = get_Fq(q_f);

      if (store_non_dirichlet_operators)
        {
          F_vector_name << "_non_dirichlet";
          all_vectors[F_vector_name.str()] = get_non_dirichlet_Fq(q_f);
        }
    }
}

get_Aq()

SparseMatrix< Number > * libMesh::RBConstruction::get_Aq

(

unsigned int

q

)

Get a pointer to Aq.

Definition at line 2371 of file rb_construction.C.

References Aq_vector, get_rb_theta_expansion(), and store_dirichlet_operators.

Referenced by add_scaled_Aq(), assemble_all_affine_operators(), get_all_matrices(), get_non_dirichlet_Aq_if_avail(), libMesh::TransientRBConstruction::truth_assembly(), truth_assembly(), and update_residual_terms().

{
  libmesh_error_msg_if(!store_dirichlet_operators,
                       "Error: Must have store_dirichlet_operators==true to access non_dirichlet_Aq.");

  libmesh_error_msg_if(q >= get_rb_theta_expansion().get_n_A_terms(),
                       "Error: We must have q < Q_a in get_Aq.");

  return Aq_vector[q].get();
}

get_closest_value()

Real libMesh::RBParametrized::get_closest_value ( Real  value, const std::vector< Real > &  list_of_values  )

staticinherited

Returns

The closest entry to value from list_of_values.

Definition at line 432 of file rb_parametrized.C.

References distance(), libMesh::Real, and value.

Referenced by libMesh::RBParametrized::is_value_in_list().

{
  libmesh_error_msg_if(list_of_values.empty(), "Error: list_of_values is empty.");

  Real min_distance = std::numeric_limits<Real>::max();
  Real closest_val = 0.;
  for (const auto & current_value : list_of_values)
    {
      Real distance = std::abs(value - current_value);
      if (distance < min_distance)
        {
          min_distance = distance;
          closest_val = current_value;
        }
    }

  return closest_val;
}

get_constraint_object()

System::Constraint & libMesh::System::get_constraint_object ( )

inherited

Return the user object for imposing constraints.

Definition at line 2019 of file system.C.

References libMesh::System::_constrain_system_object.

{
  libmesh_assert_msg(_constrain_system_object,"No constraint object available.");
  return *_constrain_system_object;
}

get_convergence_assertion_flag()

bool libMesh::RBConstruction::get_convergence_assertion_flag ( ) const

protected

Getter for the flag determining if convergence should be checked after each solve.

Definition at line 2745 of file rb_construction.C.

References assert_convergence.

{
  return assert_convergence;
}

get_current_training_parameter_index()

unsigned int libMesh::RBConstruction::get_current_training_parameter_index ( ) const

protected

Get/set the current training parameter index.

Definition at line 2765 of file rb_construction.C.

References _current_training_parameter_index.

Referenced by get_RB_error_bound().

{
  return _current_training_parameter_index;
}

get_delta_N()

unsigned int libMesh::RBConstruction::get_delta_N ( ) const

inline

Get delta_N, the number of basis functions we add to the RB space per iteration of the greedy algorithm.

For steady-state systems, this should be 1, but can be more than 1 for time-dependent systems.

Definition at line 469 of file rb_construction.h.

References delta_N.

Referenced by libMesh::TransientRBConstruction::add_IC_to_RB_space(), libMesh::TransientRBConstruction::enrich_RB_space(), libMesh::TransientRBConstruction::print_info(), libMesh::TransientRBConstruction::update_system(), libMesh::TransientRBConstruction::write_riesz_representors_to_files(), and write_riesz_representors_to_files().

{ return delta_N; }

get_deterministic_training_parameter_name()

const std::string& libMesh::RBConstructionBase< LinearImplicitSystem >::get_deterministic_training_parameter_name ( ) const

inherited

Get the name of the parameter that we will generate deterministic training parameters for.

get_discrete_parameter_values()

const std::map< std::string, std::vector< Real > > & libMesh::RBParametrized::get_discrete_parameter_values ( ) const

inherited

Get a const reference to the discrete parameter values.

Definition at line 359 of file rb_parametrized.C.

References libMesh::RBParametrized::_discrete_parameter_values, and libMesh::RBParametrized::parameters_initialized.

Referenced by libMesh::RBDataSerialization::add_parameter_ranges_to_builder(), libMesh::RBParametrized::check_if_valid_params(), libMesh::RBParametrized::get_n_discrete_params(), libMesh::RBParametrized::initialize_parameters(), libMesh::RBParametrized::print_discrete_parameter_values(), and libMesh::RBParametrized::write_discrete_parameter_values_to_file().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::get_discrete_parameter_values");

  return _discrete_parameter_values;
}

get_dof_map() [1/2]

const DofMap & libMesh::System::get_dof_map ( ) const

inlineinherited

Returns

A constant reference to this system's _dof_map.

Definition at line 2417 of file system.h.

References libMesh::System::_dof_map.

Referenced by libMesh::__libmesh_petsc_diff_solver_jacobian(), libMesh::__libmesh_petsc_diff_solver_residual(), libMesh::ExactSolution::_compute_error(), libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::DifferentiableSystem::add_dot_var_dirichlet_bcs(), libMesh::System::add_matrix(), libMesh::HPCoarsenTest::add_projection(), add_scaled_matrix_and_vector(), libMesh::System::add_variable(), libMesh::System::add_variable_array(), libMesh::System::add_variables(), libMesh::AdaptiveTimeSolver::adjoint_advance_timestep(), libMesh::UnsteadySolver::adjoint_advance_timestep(), libMesh::ImplicitSystem::adjoint_solve(), libMesh::NewmarkSolver::advance_timestep(), libMesh::AdaptiveTimeSolver::advance_timestep(), libMesh::UnsteadySolver::advance_timestep(), libMesh::EquationSystems::allgather(), libMesh::TransientRBConstruction::allocate_data_structures(), allocate_data_structures(), alternative_fe_assembly(), assemble(), LinearElasticity::assemble(), assemble_1D(), AssembleOptimization::assemble_A_and_F(), libMesh::ClawSystem::assemble_advection_matrices(), assemble_and_solve(), libMesh::ClawSystem::assemble_avg_coupling_matrices(), assemble_biharmonic(), libMesh::ClawSystem::assemble_boundary_condition_matrices(), assemble_divgrad(), assemble_elasticity(), assemble_ellipticdg(), assemble_func(), assemble_graddiv(), assemble_helmholtz(), libMesh::ClawSystem::assemble_jump_coupling_matrix(), assemble_laplace(), assemble_mass(), libMesh::ClawSystem::assemble_mass_matrix(), assemble_matrices(), assemble_matrix_and_rhs(), assemble_poisson(), assemble_SchroedingerEquation(), assemble_shell(), assemble_stokes(), assemble_temperature_jump(), assemble_wave(), libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::PetscDMWrapper::build_sf(), libMesh::System::calculate_norm(), compute_jacobian(), compute_residual(), compute_stresses(), LinearElasticityWithContact::compute_stresses(), LinearElasticity::compute_stresses(), LargeDeformationElasticity::compute_stresses(), libMesh::Problem_Interface::computeF(), libMesh::Problem_Interface::computeJacobian(), libMesh::Problem_Interface::computePreconditioner(), MyConstraint::constrain(), libMesh::VariationalSmootherConstraint::constrain_node_to_line(), libMesh::VariationalSmootherConstraint::constrain_node_to_plane(), libMesh::Nemesis_IO::copy_elemental_solution(), libMesh::ExodusII_IO::copy_elemental_solution(), libMesh::Nemesis_IO::copy_nodal_solution(), libMesh::Nemesis_IO::copy_scalar_solution(), libMesh::ExodusII_IO::copy_scalar_solution(), libMesh::CondensedEigenSystem::copy_sub_to_super(), libMesh::CondensedEigenSystem::copy_super_to_sub(), libMesh::System::create_static_condensation(), libMesh::ImplicitSystem::create_static_condensation_system_matrix(), create_wrapped_function(), DMCreateDomainDecomposition_libMesh(), DMCreateFieldDecomposition_libMesh(), DMlibMeshFunction(), DMlibMeshJacobian(), DMlibMeshSetSystem_libMesh(), libMesh::JumpErrorEstimator::estimate_error(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::ExactErrorEstimator::estimate_error(), fe_assembly(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::SubFunctor::find_dofs_to_send(), libMesh::VariationalSmootherConstraint::fix_node(), form_functionA(), form_functionB(), form_matrixA(), libMesh::System::get_all_variable_numbers(), libMesh::CondensedEigenSystem::get_eigenpair(), libMesh::System::get_info(), libMesh::System::has_static_condensation(), libMesh::System::has_variable(), libMesh::System::identify_variable_groups(), libMesh::SystemSubsetBySubdomain::init(), libMesh::SecondOrderUnsteadySolver::init_data(), libMesh::UnsteadySolver::init_data(), HeatSystem::init_data(), SimpleRBConstruction::init_data(), libMesh::ClawSystem::init_data(), LaplaceSystem::init_dirichlet_bcs(), NonManifoldCouplingTestBase::init_es(), libMesh::EigenSystem::init_matrices(), libMesh::System::init_matrices(), libMesh::PetscDMWrapper::init_petscdm(), libMesh::CondensedEigenSystem::initialize_condensed_dofs(), libMesh::OptimizationSystem::initialize_equality_constraints_storage(), libMesh::OptimizationSystem::initialize_inequality_constraints_storage(), LaplaceYoung::jacobian(), LargeDeformationElasticity::jacobian(), libMesh::System::late_matrix_init(), libMesh::libmesh_petsc_snes_fd_residual(), libMesh::libmesh_petsc_snes_jacobian(), libMesh::libmesh_petsc_snes_mffd_residual(), libMesh::libmesh_petsc_snes_residual(), libMesh::libmesh_petsc_snes_residual_helper(), libMesh::System::local_dof_indices(), AssembleOptimization::lower_and_upper_bounds(), main(), libMesh::DofMap::max_constraint_error(), LinearElasticityWithContact::move_mesh(), libMesh::System::n_components(), libMesh::System::n_variable_groups(), libMesh::System::n_vars(), libMesh::DGFEMContext::neighbor_side_fe_reinit(), libMesh::UnsteadySolver::old_nonlinear_solution(), libMesh::SecondOrderUnsteadySolver::old_solution_accel(), libMesh::SecondOrderUnsteadySolver::old_solution_rate(), libMesh::WeightedPatchRecoveryErrorEstimator::EstimateError::operator()(), libMesh::SmoothnessEstimator::EstimateSmoothness::operator()(), libMesh::PatchRecoveryErrorEstimator::EstimateError::operator()(), libMesh::RBSCMConstruction::perform_SCM_greedy(), periodic_bc_test_poisson(), libMesh::petsc_auto_fieldsplit(), libMesh::ErrorVector::plot_error(), libMesh::System::point_gradient(), libMesh::System::point_hessian(), libMesh::System::point_value(), libMesh::FEMContext::pre_fe_reinit(), libMesh::InterMeshProjection::project_system_vectors(), libMesh::System::re_update(), libMesh::System::read_parallel_data(), libMesh::System::read_SCALAR_dofs(), MeshFunctionTest::read_variable_info_from_output_data(), libMesh::SecondOrderUnsteadySolver::reinit(), libMesh::UnsteadySolver::reinit(), libMesh::System::reinit(), libMesh::System::reinit_constraints(), libMesh::EquationSystems::reinit_solutions(), LaplaceYoung::residual(), LargeDeformationElasticity::residual(), LinearElasticityWithContact::residual_and_jacobian(), libMesh::UnsteadySolver::retrieve_timestep(), OverlappingAlgebraicGhostingTest::run_ghosting_test(), OverlappingCouplingGhostingTest::run_sparsity_pattern_test(), libMesh::HPCoarsenTest::select_refinement(), libMesh::ImplicitSystem::sensitivity_solve(), set_context_solution_vec(), libMesh::PetscPreconditioner< T >::set_hypre_ads_data(), libMesh::PetscPreconditioner< T >::set_hypre_ams_data(), libMesh::PetscDMWrapper::set_point_range_in_section(), set_system_parameters(), SlitMeshRefinedSystemTest::setUp(), FETestBase< order, family, elem_type, N_x >::setUp(), SolidSystem::side_time_derivative(), libMesh::NewtonSolver::solve(), libMesh::PetscDiffSolver::solve(), libMesh::EigenSystem::solve(), solve_for_matrix_and_rhs(), SystemsTest::test100KVariables(), ConstraintOperatorTest::test1DCoarseningNewNodes(), ConstraintOperatorTest::test1DCoarseningOperator(), MeshFunctionTest::test_bad_gradient_var_with_out_of_mesh_value(), MeshFunctionTest::test_bad_hessian_var_with_out_of_mesh_value(), MeshfunctionDFEM::test_mesh_function_dfem(), MeshfunctionDFEM::test_mesh_function_dfem_grad(), MeshFunctionTest::test_p_level(), ProjectSolutionTest::test_partial_project_solution(), MeshFunctionTest::test_subdomain_id_sets(), SystemsTest::testBlockRestrictedVarNDofs(), DofMapTest::testConstraintLoopDetection(), DefaultCouplingTest::testCoupling(), PointNeighborCouplingTest::testCoupling(), EquationSystemsTest::testDisableDefaultGhosting(), SystemsTest::testDofCouplingWithVarGroups(), DofMapTest::testDofOwner(), MeshInputTest::testDynaReadPatch(), MeshInputTest::testExodusWriteElementDataFromDiscontinuousNodalData(), MeshAssignTest::testMeshMoveAssign(), PeriodicBCTest::testPeriodicBC(), SystemsTest::testPostInitAddVectorTypeChange(), SystemsTest::testProjectCubeWithMeshFunction(), SystemsTest::testProjectMatrix1D(), SystemsTest::testProjectMatrix2D(), SystemsTest::testProjectMatrix3D(), InfFERadialTest::testRefinement(), EquationSystemsTest::testSelectivePRefine(), BoundaryInfoTest::testShellFaceConstraints(), DisjointNeighborTest::testTempJump(), DisjointNeighborTest::testTempJumpRefine(), libMesh::MeshFunctionSolutionTransfer::transfer(), libMesh::BoundaryVolumeSolutionTransfer::transfer_boundary_volume(), libMesh::UnsteadySolver::update(), update_current_local_solution(), libMesh::System::variable(), libMesh::System::variable_group(), libMesh::System::variable_name(), libMesh::System::variable_number(), libMesh::System::variable_scalar_number(), libMesh::System::variable_type(), NonManifoldGhostingFunctorTest::verify_send_list_entries_helper(), libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve(), libMesh::ImplicitSystem::weighted_sensitivity_solve(), libMesh::Nemesis_IO_Helper::write_nodal_solution(), libMesh::System::write_parallel_data(), libMesh::EnsightIO::write_scalar_ascii(), libMesh::System::write_SCALAR_dofs(), libMesh::EnsightIO::write_vector_ascii(), and zero_constrained_dofs_on_vector().

{
  return *_dof_map;
}

get_dof_map() [2/2]

DofMap & libMesh::System::get_dof_map ( )

inlineinherited

Returns

A writable reference to this system's _dof_map.

Definition at line 2425 of file system.h.

References libMesh::System::_dof_map.

{
  return *_dof_map;
}

get_equation_systems() [1/2]

const EquationSystems& libMesh::System::get_equation_systems ( ) const

inlineinherited

Returns

A constant reference to this system's parent EquationSystems object.

Definition at line 767 of file system.h.

References libMesh::System::_equation_systems.

Referenced by libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::RBSCMConstruction::add_scaled_symm_Aq(), libMesh::NewmarkSystem::clear(), libMesh::FrequencySystem::clear_all(), compute_jacobian(), compute_residual(), LinearElasticityWithContact::compute_stresses(), SolidSystem::element_time_derivative(), enrich_basis_from_rhs_terms(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::AdjointResidualErrorEstimator::estimate_error(), libMesh::ExactErrorEstimator::find_squared_element_error(), libMesh::ImplicitSystem::get_linear_solve_parameters(), SolidSystem::init_data(), HeatSystem::init_data(), libMesh::FrequencySystem::init_data(), LaplaceYoung::jacobian(), libMesh::RBSCMConstruction::load_matrix_B(), LinearElasticityWithContact::move_mesh(), libMesh::FrequencySystem::n_frequencies(), libMesh::RBSCMConstruction::perform_SCM_greedy(), libMesh::System::point_gradient(), libMesh::System::point_value(), libMesh::InterMeshProjection::project_system_vectors(), libMesh::StaticCondensationDofMap::reinit(), LaplaceYoung::residual(), LinearElasticityWithContact::residual_and_jacobian(), libMesh::FileHistoryData::retrieve_adjoint_solution(), libMesh::FileHistoryData::retrieve_primal_solution(), libMesh::FileHistoryData::rewrite_stored_solution(), SolidSystem::save_initial_mesh(), libMesh::FrequencySystem::set_current_frequency(), libMesh::FrequencySystem::set_frequencies(), libMesh::FrequencySystem::set_frequencies_by_range(), libMesh::FrequencySystem::set_frequencies_by_steps(), libMesh::PetscPreconditioner< T >::set_hypre_ads_data(), libMesh::PetscPreconditioner< T >::set_hypre_ams_data(), libMesh::NewmarkSystem::set_newmark_parameters(), libMesh::NonlinearImplicitSystem::set_solver_parameters(), SolidSystem::side_time_derivative(), libMesh::EigenSystem::solve(), libMesh::CondensedEigenSystem::solve(), libMesh::FrequencySystem::solve(), libMesh::ClawSystem::solve_conservation_law(), solve_for_matrix_and_rhs(), libMesh::EigenSystem::solve_helper(), libMesh::FileHistoryData::store_adjoint_solution(), libMesh::FileHistoryData::store_initial_solution(), libMesh::FileHistoryData::store_primal_solution(), MeshFunctionTest::test_p_level(), MeshFunctionTest::test_subdomain_id_sets(), MeshAssignTest::testMeshMoveAssign(), libMesh::MeshFunctionSolutionTransfer::transfer(), libMesh::DirectSolutionTransfer::transfer(), libMesh::MeshfreeSolutionTransfer::transfer(), libMesh::DTKSolutionTransfer::transfer(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), and libMesh::WrappedFunction< Output >::WrappedFunction().

{ return _equation_systems; }

get_equation_systems() [2/2]

EquationSystems& libMesh::System::get_equation_systems ( )

inlineinherited

Returns

A reference to this system's parent EquationSystems object.

Definition at line 772 of file system.h.

References libMesh::System::_equation_systems.

{ return _equation_systems; }

get_evaluated_thetas()

const std::vector< Number > & libMesh::RBConstruction::get_evaluated_thetas ( unsigned int  training_parameter_index ) const

protected

Return the evaluated theta functions at the given training parameter index.

Definition at line 2776 of file rb_construction.C.

References _evaluated_thetas, libMesh::RBConstructionBase< LinearImplicitSystem >::get_first_local_training_index(), and libMesh::libmesh_assert().

Referenced by get_RB_error_bound().

{
  const numeric_index_type first_index = get_first_local_training_index();
  libmesh_assert(training_parameter_index >= first_index);

  const numeric_index_type local_index = training_parameter_index - first_index;
  libmesh_assert(local_index < _evaluated_thetas.size());

  return _evaluated_thetas[local_index];
}

get_first_local_training_index()

numeric_index_type libMesh::RBConstructionBase< LinearImplicitSystem >::get_first_local_training_index ( ) const

inherited

Get the first local index of the training parameters.

Definition at line 191 of file rb_construction_base.C.

Referenced by compute_max_error_bound(), get_evaluated_thetas(), and preevaluate_thetas().

{
  libmesh_error_msg_if(!_training_parameters_initialized,
                       "Error: training parameters must first be initialized.");

  // First we check if there are no parameters here, and in that case we
  // return 0 for a serial training set and comm().rank() for a parallel
  // training set. This is consistent with get_n_training_samples(), and
  // avoids accessing training_parameters.begin() when training_parameters
  // is empty.
  if (_training_parameters.empty())
    {
      if (serial_training_set)
        return 0;
      else
        return this->comm().rank();
    }

  return _first_local_index;
}

get_Fq()

NumericVector< Number > * libMesh::RBConstruction::get_Fq

(

unsigned int

q

)

Get a pointer to Fq.

Definition at line 2403 of file rb_construction.C.

References Fq_vector, get_rb_theta_expansion(), and store_dirichlet_operators.

Referenced by assemble_all_affine_vectors(), compute_Fq_representor_innerprods(), enrich_basis_from_rhs_terms(), get_all_vectors(), get_non_dirichlet_Fq_if_avail(), libMesh::TransientRBConstruction::truth_assembly(), and truth_assembly().

{
  libmesh_error_msg_if(!store_dirichlet_operators,
                       "Error: Must have store_dirichlet_operators==true to access non_dirichlet_Fq.");

  libmesh_error_msg_if(q >= get_rb_theta_expansion().get_n_F_terms(),
                       "Error: We must have q < Q_f in get_Fq.");

  return Fq_vector[q].get();
}

get_global_max_error_pair()

void libMesh::RBConstructionBase< LinearImplicitSystem >::get_global_max_error_pair ( const Parallel::Communicator &  communicator, std::pair< numeric_index_type, Real > &  error_pair  )

staticprotectedinherited

Static function to return the error pair (index,error) that is corresponds to the largest error on all processors.

Definition at line 134 of file rb_construction_base.C.

Referenced by compute_max_error_bound().

{
  // Set error_pair.second to the maximum global value and also
  // find which processor contains the maximum value
  unsigned int proc_ID_index;
  communicator.maxloc(error_pair.second, proc_ID_index);

  // Then broadcast error_pair.first from proc_ID_index
  communicator.broadcast(error_pair.first, proc_ID_index);
}

get_greedy_parameter()

const RBParameters & libMesh::RBConstruction::get_greedy_parameter

(

unsigned int

i

)

Return the parameters chosen during the i^th step of the Greedy algorithm.

Definition at line 1607 of file rb_construction.C.

References get_rb_evaluation(), and libMesh::RBEvaluation::greedy_param_list.

{
  libmesh_error_msg_if(i >= get_rb_evaluation().greedy_param_list.size(),
                       "Error: Argument in RBConstruction::get_greedy_parameter is too large.");

  return get_rb_evaluation().greedy_param_list[i];
}

get_info() [1/3]

std::string libMesh::ReferenceCounter::get_info ( )

staticinherited

Gets a string containing the reference information.

Definition at line 47 of file reference_counter.C.

References libMesh::ReferenceCounter::_counts, and libMesh::Quality::name().

Referenced by libMesh::ReferenceCounter::print_info().

{
#if defined(LIBMESH_ENABLE_REFERENCE_COUNTING) && defined(DEBUG)

  std::ostringstream oss;

  oss << '\n'
      << " ---------------------------------------------------------------------------- \n"
      << "| Reference count information                                                |\n"
      << " ---------------------------------------------------------------------------- \n";

  for (const auto & [name, cd] : _counts)
    oss << "| " << name << " reference count information:\n"
        << "|  Creations:    " << cd.first    << '\n'
        << "|  Destructions: " << cd.second << '\n';

  oss << " ---------------------------------------------------------------------------- \n";

  return oss.str();

#else

  return "";

#endif
}

get_info() [2/3]

std::string libMesh::ReferenceCounter::get_info ( )

staticinherited

Gets a string containing the reference information.

Definition at line 47 of file reference_counter.C.

References libMesh::ReferenceCounter::_counts, and libMesh::Quality::name().

Referenced by libMesh::ReferenceCounter::print_info().

{
#if defined(LIBMESH_ENABLE_REFERENCE_COUNTING) && defined(DEBUG)

  std::ostringstream oss;

  oss << '\n'
      << " ---------------------------------------------------------------------------- \n"
      << "| Reference count information                                                |\n"
      << " ---------------------------------------------------------------------------- \n";

  for (const auto & [name, cd] : _counts)
    oss << "| " << name << " reference count information:\n"
        << "|  Creations:    " << cd.first    << '\n'
        << "|  Destructions: " << cd.second << '\n';

  oss << " ---------------------------------------------------------------------------- \n";

  return oss.str();

#else

  return "";

#endif
}

get_info() [3/3]

std::string libMesh::System::get_info ( ) const

inherited

Returns

A string containing information about the system.

Definition at line 1822 of file system.C.

References libMesh::ParallelObject::comm(), libMesh::FEType::family, libMesh::System::get_dof_map(), libMesh::DofMap::get_info(), libMesh::DofMap::get_static_condensation(), libMesh::System::has_static_condensation(), libMesh::FEType::inf_map, libMesh::System::is_initialized(), libMesh::make_range(), TIMPI::Communicator::max(), libMesh::System::n_constrained_dofs(), libMesh::DofMapBase::n_dofs(), libMesh::System::n_dofs(), libMesh::System::n_local_constrained_dofs(), libMesh::System::n_local_dofs(), libMesh::System::n_matrices(), libMesh::System::n_variable_groups(), libMesh::VariableGroup::n_variables(), libMesh::System::n_vectors(), libMesh::VariableGroup::name(), libMesh::System::name(), libMesh::System::number(), libMesh::FEType::order, libMesh::FEType::radial_family, libMesh::FEType::radial_order, libMesh::System::system_type(), libMesh::Variable::type(), libMesh::DofMap::variable_group(), and libMesh::System::variable_group().

Referenced by SystemsTest::testUninitializedInfo().

{
  std::ostringstream oss;


  const std::string & sys_name = this->name();

  oss << "   System #"  << this->number() << ", \"" << sys_name << "\"\n"
      << "    Type \""  << this->system_type() << "\"\n"
      << "    Variables=";

  for (auto vg : make_range(this->n_variable_groups()))
    {
      const VariableGroup & vg_description (this->variable_group(vg));

      if (vg_description.n_variables() > 1) oss << "{ ";
      for (auto vn : make_range(vg_description.n_variables()))
        oss << "\"" << vg_description.name(vn) << "\" ";
      if (vg_description.n_variables() > 1) oss << "} ";
    }

  oss << '\n';

  oss << "    Finite Element Types=";
#ifndef LIBMESH_ENABLE_INFINITE_ELEMENTS
  for (auto vg : make_range(this->n_variable_groups()))
    oss << "\""
        << Utility::enum_to_string<FEFamily>(this->get_dof_map().variable_group(vg).type().family)
        << "\" ";
#else
  for (auto vg : make_range(this->n_variable_groups()))
    {
      oss << "\""
          << Utility::enum_to_string<FEFamily>(this->get_dof_map().variable_group(vg).type().family)
          << "\", \""
          << Utility::enum_to_string<FEFamily>(this->get_dof_map().variable_group(vg).type().radial_family)
          << "\" ";
    }

  oss << '\n' << "    Infinite Element Mapping=";
  for (auto vg : make_range(this->n_variable_groups()))
    oss << "\""
        << Utility::enum_to_string<InfMapType>(this->get_dof_map().variable_group(vg).type().inf_map)
        << "\" ";
#endif

  oss << '\n';

  oss << "    Approximation Orders=";
  for (auto vg : make_range(this->n_variable_groups()))
    {
#ifndef LIBMESH_ENABLE_INFINITE_ELEMENTS
      oss << "\""
          << Utility::enum_to_string<Order>(this->get_dof_map().variable_group(vg).type().order)
          << "\" ";
#else
      oss << "\""
          << Utility::enum_to_string<Order>(this->get_dof_map().variable_group(vg).type().order)
          << "\", \""
          << Utility::enum_to_string<Order>(this->get_dof_map().variable_group(vg).type().radial_order)
          << "\" ";
#endif
    }

  oss << '\n';

  if (this->is_initialized())
    {
      oss << "    n_dofs()="             << this->n_dofs()             << '\n';
      dof_id_type local_dofs = this->n_local_dofs();
      oss << "    n_local_dofs()="       << local_dofs                 << '\n';
      this->comm().max(local_dofs);
      oss << "    max(n_local_dofs())="  << local_dofs                 << '\n';
#ifdef LIBMESH_ENABLE_CONSTRAINTS
      if (this->n_constrained_dofs())
        {
          oss << "    n_constrained_dofs()=" << this->n_constrained_dofs() << '\n';
          oss << "    n_local_constrained_dofs()=" << this->n_local_constrained_dofs() << '\n';
          dof_id_type local_unconstrained_dofs = this->n_local_dofs() - this->n_local_constrained_dofs();
          this->comm().max(local_unconstrained_dofs);
          oss << "    max(local unconstrained dofs)=" << local_unconstrained_dofs << '\n';
        }
#endif
      if (this->has_static_condensation())
        oss << "    n uncondensed dofs="
            << this->get_dof_map().get_static_condensation().n_dofs() << '\n';
    }
  else
    oss << "    (still uninitialized)\n";

  oss << "    " << "n_vectors()="  << this->n_vectors()  << '\n';
  oss << "    " << "n_matrices()="  << this->n_matrices()  << '\n';
  //   oss << "    " << "n_additional_matrices()=" << this->n_additional_matrices() << '\n';

  oss << this->get_dof_map().get_info();

  return oss.str();
}

get_inner_product_assembly()

ElemAssembly & libMesh::RBConstruction::get_inner_product_assembly

(

)

Returns

A reference to the inner product assembly object

Definition at line 427 of file rb_construction.C.

References inner_product_assembly, and use_energy_inner_product.

{
  libmesh_error_msg_if(use_energy_inner_product,
                       "Error: inner_product_assembly not available since we're using energy inner-product");

  libmesh_error_msg_if(!inner_product_assembly,
                       "Error: inner_product_assembly hasn't been initialized yet");

  return *inner_product_assembly;
}

get_inner_product_matrix() [1/2]

SparseMatrix< Number > * libMesh::RBConstruction::get_inner_product_matrix

(

)

Get a pointer to inner_product_matrix.

Accessing via this function, rather than directly through the class member allows us to do error checking (e.g. inner_product_matrix is not defined in low-memory mode).

Definition at line 2321 of file rb_construction.C.

References inner_product_matrix, and store_dirichlet_operators.

Referenced by get_all_matrices(), get_non_dirichlet_inner_product_matrix_if_avail(), and libMesh::RBSCMConstruction::load_matrix_B().

{
  libmesh_error_msg_if(!store_dirichlet_operators,
                       "Error: Must have store_dirichlet_operators==true to access inner_product_matrix.");
  return inner_product_matrix.get();
}

get_inner_product_matrix() [2/2]

const SparseMatrix< Number > * libMesh::RBConstruction::get_inner_product_matrix

(

)

const

Definition at line 2328 of file rb_construction.C.

References inner_product_matrix, and store_dirichlet_operators.

{
  libmesh_error_msg_if(!store_dirichlet_operators,
                       "Error: Must have store_dirichlet_operators==true to access inner_product_matrix.");
  return inner_product_matrix.get();
}

get_last_local_training_index()

numeric_index_type libMesh::RBConstructionBase< LinearImplicitSystem >::get_last_local_training_index ( ) const

inherited

Get the last local index of the training parameters.

Definition at line 213 of file rb_construction_base.C.

Referenced by compute_max_error_bound().

{
  libmesh_error_msg_if(!_training_parameters_initialized,
                       "Error: training parameters must first be initialized.");

  if (_training_parameters.empty())
    return 0;

  return _first_local_index + _n_local_training_samples;
}

get_linear_solve_parameters()

std::pair< unsigned int, Real > libMesh::ImplicitSystem::get_linear_solve_parameters ( ) const

virtualinherited

Returns

An integer corresponding to the upper iteration count limit and a Real corresponding to the convergence tolerance to be used in linear adjoint and/or sensitivity solves

Reimplemented in libMesh::NonlinearImplicitSystem, and libMesh::DifferentiableSystem.

Definition at line 1237 of file implicit_system.C.

References libMesh::Parameters::get(), libMesh::System::get_equation_systems(), libMesh::Parameters::have_parameter(), libMesh::EquationSystems::parameters, libMesh::System::parameters, and libMesh::Real.

Referenced by libMesh::ImplicitSystem::adjoint_solve(), libMesh::ImplicitSystem::sensitivity_solve(), libMesh::NonlinearImplicitSystem::set_solver_parameters(), libMesh::FrequencySystem::solve(), libMesh::LinearImplicitSystem::solve(), libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve(), and libMesh::ImplicitSystem::weighted_sensitivity_solve().

{
  return std::make_pair(
      parameters.have_parameter<unsigned int>("linear solver maximum iterations")
          ? parameters.get<unsigned int>("linear solver maximum iterations")
          : this->get_equation_systems().parameters.get<unsigned int>(
                "linear solver maximum iterations"),
      parameters.have_parameter<Real>("linear solver tolerance")
          ? parameters.get<Real>("linear solver tolerance")
          : this->get_equation_systems().parameters.get<Real>("linear solver tolerance"));
}

get_linear_solver()

LinearSolver< Number > * libMesh::LinearImplicitSystem::get_linear_solver ( ) const

overridevirtualinherited

Returns

A pointer to a linear solver appropriate for use in adjoint and/or sensitivity solves

Reimplemented from libMesh::ImplicitSystem.

Definition at line 360 of file linear_implicit_system.C.

References libMesh::ImplicitSystem::linear_solver.

Referenced by compute_output_dual_innerprods(), enrich_basis_from_rhs_terms(), initialize_rb_construction(), main(), SystemsTest::testDofCouplingWithVarGroups(), libMesh::TransientRBConstruction::truth_solve(), and truth_solve().

{
  return linear_solver.get();
}

get_local_n_training_samples()

numeric_index_type libMesh::RBConstructionBase< LinearImplicitSystem >::get_local_n_training_samples ( ) const

inherited

Get the total number of training samples local to this processor.

Definition at line 175 of file rb_construction_base.C.

Referenced by compute_max_error_bound(), and preevaluate_thetas().

{
  libmesh_error_msg_if(!_training_parameters_initialized,
                       "Error: training parameters must first be initialized.");

  // First we check if there are no parameters here, and in that case we
  // return 1 for both serial and parallel training sets. This is consistent
  // with get_n_training_samples(), and avoids accessing
  // training_parameters.begin() when training_parameters is empty.
  if (_training_parameters.empty())
    return 1;

  return _n_local_training_samples;
}

get_matrix() [1/2]

const SparseMatrix< Number > & libMesh::System::get_matrix ( std::string_view  mat_name ) const

inherited

Returns

A const reference to this system's matrix named mat_name.

Definition at line 1111 of file system.C.

References libMesh::System::_matrices.

Referenced by add_M_C_K_helmholtz(), assemble(), assemble_helmholtz(), assemble_poisson(), libMesh::NewmarkSystem::compute_matrix(), libMesh::ImplicitSystem::get_system_matrix(), main(), libMesh::EigenTimeSolver::solve(), and libMesh::NewmarkSystem::update_rhs().

{
  return *libmesh_map_find(_matrices, mat_name);
}

get_matrix() [2/2]

SparseMatrix< Number > & libMesh::System::get_matrix ( std::string_view  mat_name )

inherited

Returns

A writable reference to this system's matrix named mat_name.

Definition at line 1118 of file system.C.

References libMesh::System::_matrices.

{
  return *libmesh_map_find(_matrices, mat_name);
}

get_matrix_for_output_dual_solves()

SparseMatrix< Number > & libMesh::RBConstruction::get_matrix_for_output_dual_solves ( )

protectedvirtual

Return the matrix for the output residual dual norm solves.

By default we use the inner product matrix for steady state problems.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 2086 of file rb_construction.C.

References inner_product_matrix.

Referenced by compute_output_dual_innerprods().

{
  return *inner_product_matrix;
}

get_mesh() [1/2]

const MeshBase & libMesh::System::get_mesh ( ) const

inlineinherited

Returns

A constant reference to this systems's _mesh.

Definition at line 2401 of file system.h.

References libMesh::System::_mesh.

Referenced by libMesh::ExactSolution::_compute_error(), LinearElasticityWithContact::add_contact_edge_elements(), libMesh::PetscDMWrapper::add_dofs_to_section(), libMesh::System::add_matrix(), libMesh::HPCoarsenTest::add_projection(), add_scaled_matrix_and_vector(), libMesh::DofMap::add_variables(), AssembleOptimization::assemble_A_and_F(), libMesh::ClawSystem::assemble_advection_matrices(), libMesh::ClawSystem::assemble_avg_coupling_matrices(), libMesh::ClawSystem::assemble_boundary_condition_matrices(), libMesh::ClawSystem::assemble_jump_coupling_matrix(), libMesh::ClawSystem::assemble_mass_matrix(), libMesh::FEMSystem::assemble_qoi(), libMesh::FEMSystem::assemble_qoi_derivative(), libMesh::FEMSystem::assembly(), libMesh::VariationalSmootherSystem::assembly(), AssemblyA0::boundary_assembly(), AssemblyA1::boundary_assembly(), AssemblyF0::boundary_assembly(), AssemblyF1::boundary_assembly(), AssemblyA2::boundary_assembly(), AssemblyF2::boundary_assembly(), libMesh::System::calculate_norm(), compute_jacobian(), libMesh::VariationalSmootherSystem::compute_mesh_quality_info(), compute_residual(), LinearElasticityWithContact::compute_stresses(), libMesh::VariationalSmootherConstraint::constrain(), libMesh::VariationalSmootherConstraint::constrain_node_to_line(), libMesh::VariationalSmootherConstraint::constrain_node_to_plane(), libMesh::System::create_static_condensation(), libMesh::ImplicitSystem::create_static_condensation_system_matrix(), libMesh::RBEIMEvaluation::distribute_bfs(), DMCreateDomainDecomposition_libMesh(), DMCreateFieldDecomposition_libMesh(), DMlibMeshSetSystem_libMesh(), SolidSystem::element_time_derivative(), HeatSystem::element_time_derivative(), enrich_basis_from_rhs_terms(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_interiors(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_sides(), libMesh::PatchRecoveryErrorEstimator::estimate_error(), libMesh::WeightedPatchRecoveryErrorEstimator::estimate_error(), libMesh::JumpErrorEstimator::estimate_error(), libMesh::AdjointResidualErrorEstimator::estimate_error(), libMesh::ExactErrorEstimator::estimate_error(), libMesh::SmoothnessEstimator::estimate_smoothness(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::SubFunctor::find_dofs_to_send(), libMesh::VariationalSmootherConstraint::fix_node(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::GenericProjector(), LinearElasticityWithContact::get_least_and_max_gap_function(), libMesh::SystemSubsetBySubdomain::init(), libMesh::RBEIMConstruction::init_context(), SolidSystem::init_data(), libMesh::VariationalSmootherSystem::init_data(), libMesh::System::init_data(), libMesh::System::init_matrices(), libMesh::PetscDMWrapper::init_petscdm(), LinearElasticityWithContact::initialize_contact_load_paths(), libMesh::RBEIMConstruction::initialize_qp_data(), libMesh::System::local_dof_indices(), libMesh::DofMap::max_constraint_error(), libMesh::FEMSystem::mesh_position_get(), libMesh::FEMSystem::mesh_position_set(), LinearElasticityWithContact::move_mesh(), libMesh::System::n_components(), libMesh::RBEIMEvaluation::node_distribute_bfs(), libMesh::WeightedPatchRecoveryErrorEstimator::EstimateError::operator()(), libMesh::SmoothnessEstimator::EstimateSmoothness::operator()(), libMesh::PatchRecoveryErrorEstimator::EstimateError::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectVertices::operator()(), libMesh::petsc_auto_fieldsplit(), libMesh::System::point_gradient(), libMesh::System::point_hessian(), libMesh::System::point_value(), libMesh::FEMSystem::postprocess(), libMesh::RBParametrizedFunction::preevaluate_parametrized_function_on_mesh(), libMesh::RBParametrizedFunction::preevaluate_parametrized_function_on_mesh_sides(), libMesh::VariationalSmootherSystem::prepare_for_smoothing(), libMesh::System::read_header(), libMesh::RBEvaluation::read_in_vectors_from_multiple_files(), libMesh::System::read_parallel_data(), libMesh::System::read_serialized_vector(), libMesh::System::read_serialized_vectors(), libMesh::System::reinit(), LinearElasticityWithContact::residual_and_jacobian(), OverlappingAlgebraicGhostingTest::run_ghosting_test(), OverlappingCouplingGhostingTest::run_sparsity_pattern_test(), SolidSystem::save_initial_mesh(), libMesh::HPSingularity::select_refinement(), libMesh::HPCoarsenTest::select_refinement(), libMesh::PetscPreconditioner< T >::set_hypre_ads_data(), libMesh::PetscPreconditioner< T >::set_hypre_ams_data(), libMesh::PetscPreconditioner< T >::set_petsc_aux_data(), libMesh::PetscDMWrapper::set_point_range_in_section(), libMesh::RBEIMEvaluation::side_distribute_bfs(), SolidSystem::side_time_derivative(), libMesh::PetscDiffSolver::solve(), libMesh::ClawSystem::solve_conservation_law(), MeshAssignTest::testMeshMoveAssign(), libMesh::MeshFunctionSolutionTransfer::transfer(), libMesh::BoundaryVolumeSolutionTransfer::transfer(), libMesh::BoundaryVolumeSolutionTransfer::transfer_boundary_volume(), libMesh::BoundaryVolumeSolutionTransfer::transfer_volume_boundary(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::System::write_header(), libMesh::RBEvaluation::write_out_vectors(), libMesh::System::write_parallel_data(), libMesh::System::write_serialized_vector(), libMesh::System::write_serialized_vectors(), and libMesh::System::zero_variable().

{
  return _mesh;
}

get_mesh() [2/2]

MeshBase & libMesh::System::get_mesh ( )

inlineinherited

Returns

A reference to this systems's _mesh.

Definition at line 2409 of file system.h.

References libMesh::System::_mesh.

{
  return _mesh;
}

get_n_continuous_params()

unsigned int libMesh::RBParametrized::get_n_continuous_params ( ) const

inherited

Get the number of continuous parameters.

Definition at line 112 of file rb_parametrized.C.

References libMesh::RBParametrized::get_n_discrete_params(), libMesh::RBParametrized::get_n_params(), libMesh::libmesh_assert(), and libMesh::RBParametrized::parameters_initialized.

Referenced by libMesh::RBDataSerialization::add_parameter_ranges_to_builder(), and libMesh::RBParametrized::write_parameter_ranges_to_file().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::get_n_continuous_params");

  libmesh_assert(get_n_params() >= get_n_discrete_params());

  return static_cast<unsigned int>(get_n_params() - get_n_discrete_params());
}

get_n_discrete_params()

unsigned int libMesh::RBParametrized::get_n_discrete_params ( ) const

inherited

Get the number of discrete parameters.

Definition at line 121 of file rb_parametrized.C.

References libMesh::RBParametrized::get_discrete_parameter_values(), and libMesh::RBParametrized::parameters_initialized.

Referenced by libMesh::RBDataSerialization::add_parameter_ranges_to_builder(), libMesh::RBParametrized::get_n_continuous_params(), and libMesh::RBParametrized::write_discrete_parameter_values_to_file().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::get_n_discrete_params");

  return cast_int<unsigned int>
    (get_discrete_parameter_values().size());
}

get_n_params()

unsigned int libMesh::RBParametrized::get_n_params ( ) const

inherited

Get the number of parameters.

Definition at line 103 of file rb_parametrized.C.

References libMesh::RBParameters::n_parameters(), libMesh::RBParametrized::parameters_initialized, libMesh::RBParametrized::parameters_max, and libMesh::RBParametrized::parameters_min.

Referenced by libMesh::RBParametrized::check_if_valid_params(), libMesh::RBEIMConstruction::compute_max_eim_error(), compute_max_error_bound(), libMesh::RBParametrized::get_n_continuous_params(), libMesh::RBSCMConstruction::print_info(), libMesh::RBEIMConstruction::print_info(), print_info(), libMesh::RBEIMEvaluation::set_eim_error_indicator_active(), and train_reduced_basis_with_POD().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::get_n_params");

  libmesh_assert_equal_to ( parameters_min.n_parameters(), parameters_max.n_parameters() );

  return parameters_min.n_parameters();
}

get_n_training_samples()

numeric_index_type libMesh::RBConstructionBase< LinearImplicitSystem >::get_n_training_samples ( ) const

inherited

Get the number of global training samples.

Definition at line 153 of file rb_construction_base.C.

Referenced by print_info(), and train_reduced_basis_with_POD().

{
  libmesh_error_msg_if(!_training_parameters_initialized,
                       "Error: training parameters must first be initialized.");

  // First we check if there are no parameters here, and in that case we
  // return 1 since a single training sample is sufficient to generate an
  // RB approximation if there are no parameters. Note that in parallel,
  // and when we don't have a serial training set, set return comm().size()
  // so that each processor is assigned a single (empty) training sample.
  if (_training_parameters.empty())
    {
      if (serial_training_set)
        return 1;
      else
        return this->comm().size();
    }

  return _n_global_training_samples;
}

get_Nmax()

unsigned int libMesh::RBConstruction::get_Nmax ( ) const

inline

Get/set Nmax, the maximum number of RB functions we are willing to compute.

Definition at line 253 of file rb_construction.h.

References Nmax.

Referenced by enrich_basis_from_rhs_terms(), libMesh::TransientRBConstruction::enrich_RB_space(), greedy_termination_test(), train_reduced_basis_with_greedy(), and train_reduced_basis_with_POD().

{ return Nmax; }

get_non_dirichlet_Aq()

SparseMatrix< Number > * libMesh::RBConstruction::get_non_dirichlet_Aq

(

unsigned int

q

)

Get a pointer to non_dirichlet_Aq.

Definition at line 2382 of file rb_construction.C.

References get_rb_theta_expansion(), non_dirichlet_Aq_vector, and store_non_dirichlet_operators.

Referenced by assemble_all_affine_operators(), get_all_matrices(), and get_non_dirichlet_Aq_if_avail().

{
  libmesh_error_msg_if(!store_non_dirichlet_operators,
                       "Error: Must have store_non_dirichlet_operators==true to access non_dirichlet_Aq.");

  libmesh_error_msg_if(q >= get_rb_theta_expansion().get_n_A_terms(),
                       "Error: We must have q < Q_a in get_Aq.");

  return non_dirichlet_Aq_vector[q].get();
}

get_non_dirichlet_Aq_if_avail()

SparseMatrix< Number > * libMesh::RBConstruction::get_non_dirichlet_Aq_if_avail

(

unsigned int

q

)

Get a pointer to non_dirichlet_Aq if it's available, otherwise get Aq.

Definition at line 2393 of file rb_construction.C.

References get_Aq(), get_non_dirichlet_Aq(), and store_non_dirichlet_operators.

Referenced by update_RB_system_matrices().

{
  if (store_non_dirichlet_operators)
    {
      return get_non_dirichlet_Aq(q);
    }

  return get_Aq(q);
}

get_non_dirichlet_Fq()

NumericVector< Number > * libMesh::RBConstruction::get_non_dirichlet_Fq

(

unsigned int

q

)

Get a pointer to non-Dirichlet Fq.

Definition at line 2414 of file rb_construction.C.

References get_rb_theta_expansion(), non_dirichlet_Fq_vector, and store_non_dirichlet_operators.

Referenced by assemble_all_affine_vectors(), get_all_vectors(), and get_non_dirichlet_Fq_if_avail().

{
  libmesh_error_msg_if(!store_non_dirichlet_operators,
                       "Error: Must have store_non_dirichlet_operators==true to access non_dirichlet_Fq.");

  libmesh_error_msg_if(q >= get_rb_theta_expansion().get_n_F_terms(),
                       "Error: We must have q < Q_f in get_Fq.");

  return non_dirichlet_Fq_vector[q].get();
}

get_non_dirichlet_Fq_if_avail()

NumericVector< Number > * libMesh::RBConstruction::get_non_dirichlet_Fq_if_avail

(

unsigned int

q

)

Get a pointer to non_dirichlet_Fq if it's available, otherwise get Fq.

Definition at line 2425 of file rb_construction.C.

References get_Fq(), get_non_dirichlet_Fq(), and store_non_dirichlet_operators.

Referenced by update_RB_system_matrices().

{
  if (store_non_dirichlet_operators)
    {
      return get_non_dirichlet_Fq(q);
    }

  return get_Fq(q);
}

get_non_dirichlet_inner_product_matrix() [1/2]

SparseMatrix< Number > * libMesh::RBConstruction::get_non_dirichlet_inner_product_matrix

(

)

Get the non-Dirichlet (or more generally no-constraints) version of the inner-product matrix.

This is useful for performing multiplications on vectors that already have constraints enforced.

Definition at line 2335 of file rb_construction.C.

References non_dirichlet_inner_product_matrix, and store_non_dirichlet_operators.

Referenced by get_all_matrices(), and get_non_dirichlet_inner_product_matrix_if_avail().

{
  libmesh_error_msg_if(!store_non_dirichlet_operators,
                       "Error: Must have store_non_dirichlet_operators==true to access non_dirichlet_inner_product_matrix.");

  return non_dirichlet_inner_product_matrix.get();
}

get_non_dirichlet_inner_product_matrix() [2/2]

const SparseMatrix< Number > * libMesh::RBConstruction::get_non_dirichlet_inner_product_matrix

(

)

const

Definition at line 2343 of file rb_construction.C.

References non_dirichlet_inner_product_matrix, and store_non_dirichlet_operators.

{
  libmesh_error_msg_if(!store_non_dirichlet_operators,
                       "Error: Must have store_non_dirichlet_operators==true to access non_dirichlet_inner_product_matrix.");

  return non_dirichlet_inner_product_matrix.get();
}

get_non_dirichlet_inner_product_matrix_if_avail() [1/2]

SparseMatrix< Number > * libMesh::RBConstruction::get_non_dirichlet_inner_product_matrix_if_avail

(

)

Get the non-Dirichlet inner-product matrix if it's available, otherwise get the inner-product matrix with constraints.

Definition at line 2351 of file rb_construction.C.

References get_inner_product_matrix(), get_non_dirichlet_inner_product_matrix(), and store_non_dirichlet_operators.

Referenced by libMesh::TransientRBConstruction::add_IC_to_RB_space(), compute_Fq_representor_innerprods(), compute_residual_dual_norm_slow(), libMesh::TransientRBConstruction::enrich_RB_space(), enrich_RB_space(), print_basis_function_orthogonality(), libMesh::TransientRBConstruction::set_error_temporal_data(), train_reduced_basis_with_POD(), truth_solve(), update_RB_system_matrices(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

{
  if (store_non_dirichlet_operators)
    {
      return get_non_dirichlet_inner_product_matrix();
    }

  return get_inner_product_matrix();
}

get_non_dirichlet_inner_product_matrix_if_avail() [2/2]

const SparseMatrix< Number > * libMesh::RBConstruction::get_non_dirichlet_inner_product_matrix_if_avail

(

)

const

Definition at line 2361 of file rb_construction.C.

References get_inner_product_matrix(), get_non_dirichlet_inner_product_matrix(), and store_non_dirichlet_operators.

{
  if (store_non_dirichlet_operators)
    {
      return get_non_dirichlet_inner_product_matrix();
    }

  return get_inner_product_matrix();
}

get_non_dirichlet_output_vector()

NumericVector< Number > * libMesh::RBConstruction::get_non_dirichlet_output_vector	(	unsigned int	n,
		unsigned int	q_l
	)

Get a pointer to non-Dirichlet output vector.

Definition at line 2445 of file rb_construction.C.

References get_rb_theta_expansion(), and non_dirichlet_outputs_vector.

Referenced by assemble_all_output_vectors(), and get_output_vectors().

{
  libmesh_error_msg_if((n >= get_rb_theta_expansion().get_n_outputs()) ||
                       (q_l >= get_rb_theta_expansion().get_n_output_terms(n)),
                       "Error: We must have n < n_outputs and "
                       "q_l < get_rb_theta_expansion().get_n_output_terms(n) in get_non_dirichlet_output_vector.");

  return non_dirichlet_outputs_vector[n][q_l].get();
}

get_normalize_rb_bound_in_greedy()

bool libMesh::RBConstruction::get_normalize_rb_bound_in_greedy ( ) const

inline

Definition at line 234 of file rb_construction.h.

References normalize_rb_bound_in_greedy.

Referenced by print_info().

{ return normalize_rb_bound_in_greedy; }

get_output_vector()

NumericVector< Number > * libMesh::RBConstruction::get_output_vector	(	unsigned int	n,
		unsigned int	q_l
	)

Get a pointer to the n^th output.

Definition at line 2435 of file rb_construction.C.

References get_rb_theta_expansion(), and outputs_vector.

Referenced by assemble_all_output_vectors(), compute_output_dual_innerprods(), get_output_vectors(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), and update_RB_system_matrices().

{
  libmesh_error_msg_if((n >= get_rb_theta_expansion().get_n_outputs()) ||
                       (q_l >= get_rb_theta_expansion().get_n_output_terms(n)),
                       "Error: We must have n < n_outputs and "
                       "q_l < get_rb_theta_expansion().get_n_output_terms(n) in get_output_vector.");

  return outputs_vector[n][q_l].get();
}

get_output_vectors()

void libMesh::RBConstruction::get_output_vectors ( std::map< std::string, NumericVector< Number > *> &  all_vectors )

virtual

Get a map that stores pointers to all of the vectors.

Definition at line 2503 of file rb_construction.C.

References libMesh::RBThetaExpansion::get_n_output_terms(), libMesh::RBThetaExpansion::get_n_outputs(), get_non_dirichlet_output_vector(), get_output_vector(), get_rb_theta_expansion(), store_dirichlet_operators, and store_non_dirichlet_operators.

Referenced by get_all_vectors().

{
  output_vectors.clear();

  for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
    for (unsigned int q_l=0; q_l<get_rb_theta_expansion().get_n_output_terms(n); q_l++)
      {
        std::stringstream output_name;
        if (store_dirichlet_operators)
          {
            output_name << "output_" << n << "_"<< q_l;
            output_vectors[output_name.str()] = get_output_vector(n,q_l);
          }

        if (store_non_dirichlet_operators)
          {
            output_name << "_non_dirichlet";
            output_vectors[output_name.str()] = get_non_dirichlet_output_vector(n,q_l);
          }
      }
}

get_parameter_max()

Real libMesh::RBParametrized::get_parameter_max ( const std::string &  param_name ) const

inherited

Get maximum allowable value of parameter param_name.

Definition at line 171 of file rb_parametrized.C.

References libMesh::RBParameters::get_value(), libMesh::RBParametrized::parameters_initialized, and libMesh::RBParametrized::parameters_max.

Referenced by libMesh::RBParametrized::check_if_valid_params(), libMesh::RBSCMConstruction::print_info(), libMesh::RBEIMConstruction::print_info(), and print_info().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::get_parameter_max");

  return parameters_max.get_value(param_name);
}

get_parameter_min()

Real libMesh::RBParametrized::get_parameter_min ( const std::string &  param_name ) const

inherited

Get minimum allowable value of parameter param_name.

Definition at line 164 of file rb_parametrized.C.

References libMesh::RBParameters::get_value(), libMesh::RBParametrized::parameters_initialized, and libMesh::RBParametrized::parameters_min.

Referenced by libMesh::RBParametrized::check_if_valid_params(), libMesh::RBSCMConstruction::print_info(), libMesh::RBEIMConstruction::print_info(), and print_info().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::get_parameter_min");

  return parameters_min.get_value(param_name);
}

get_parameters()

const RBParameters & libMesh::RBParametrized::get_parameters ( ) const

inherited

Get the current parameters.

Definition at line 143 of file rb_parametrized.C.

References libMesh::RBParametrized::parameters, and libMesh::RBParametrized::parameters_initialized.

Referenced by libMesh::TransientRBConstruction::add_scaled_mass_matrix(), libMesh::TransientRBEvaluation::cache_online_residual_terms(), libMesh::RBEvaluation::compute_residual_dual_norm(), libMesh::RBSCMConstruction::compute_SCM_bounds_on_training_set(), libMesh::RBSCMConstruction::enrich_C_J(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_interiors(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_nodes(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_sides(), libMesh::RBEvaluation::eval_output_dual_norm(), libMesh::RBSCMConstruction::evaluate_stability_constant(), get_RB_error_bound(), libMesh::RBSCMEvaluation::get_SCM_LB(), libMesh::RBSCMEvaluation::get_SCM_UB(), SimpleRBEvaluation::get_stability_lower_bound(), greedy_termination_test(), libMesh::RBEIMConstruction::initialize_parametrized_functions_in_training_set(), libMesh::RBSCMEvaluation::legacy_read_offline_data_from_files(), libMesh::TransientRBConstruction::mass_matrix_scaled_matvec(), preevaluate_thetas(), libMesh::RBSCMConstruction::print_info(), libMesh::RBEIMConstruction::print_info(), print_info(), libMesh::RBParametrized::print_parameters(), libMesh::RBSCMConstruction::process_parameters_file(), libMesh::TransientRBEvaluation::rb_solve(), libMesh::RBEvaluation::rb_solve(), libMesh::RBSCMEvaluation::save_current_parameters(), libMesh::RBEIMConstruction::train_eim_approximation_with_greedy(), libMesh::RBEIMConstruction::train_eim_approximation_with_POD(), libMesh::TransientRBConstruction::truth_assembly(), truth_assembly(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::TransientRBEvaluation::uncached_compute_residual_dual_norm(), and update_greedy_param_list().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::get_parameters");

  return parameters;
}

get_parameters_max()

const RBParameters & libMesh::RBParametrized::get_parameters_max ( ) const

inherited

Get an RBParameters object that specifies the maximum allowable value for each parameter.

Definition at line 157 of file rb_parametrized.C.

References libMesh::RBParametrized::parameters_initialized, and libMesh::RBParametrized::parameters_max.

Referenced by libMesh::RBDataSerialization::add_parameter_ranges_to_builder(), libMesh::RBParametrized::initialize_parameters(), libMesh::RBSCMConstruction::process_parameters_file(), libMesh::RBEIMConstruction::set_rb_construction_parameters(), set_rb_construction_parameters(), and libMesh::RBParametrized::write_parameter_ranges_to_file().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::get_parameters_max");

  return parameters_max;
}

get_parameters_min()

const RBParameters & libMesh::RBParametrized::get_parameters_min ( ) const

inherited

Get an RBParameters object that specifies the minimum allowable value for each parameter.

Definition at line 150 of file rb_parametrized.C.

References libMesh::RBParametrized::parameters_initialized, and libMesh::RBParametrized::parameters_min.

Referenced by libMesh::RBDataSerialization::add_parameter_ranges_to_builder(), libMesh::RBParametrized::initialize_parameters(), libMesh::RBSCMConstruction::process_parameters_file(), libMesh::RBEIMConstruction::set_rb_construction_parameters(), set_rb_construction_parameters(), and libMesh::RBParametrized::write_parameter_ranges_to_file().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::get_parameters_min");

  return parameters_min;
}

get_params_from_training_set()

RBParameters libMesh::RBConstructionBase< LinearImplicitSystem >::get_params_from_training_set ( unsigned int  global_index )

protectedinherited

Return the RBParameters in index global_index of the global training set.

Why do we use an index here? RBParameters supports loading the full sample set. This seems probably unnecessary now to load individually. Maybe it's a memory issue?

Definition at line 231 of file rb_construction_base.C.

{
  libmesh_error_msg_if(!_training_parameters_initialized,
                       "Error: training parameters must first be initialized.");

  // If the _training_parameters are empty, return an empty RBParameters.
  // Otherwise, create a new RBParameters object from the single sample requested.
  RBParameters params;
  if (!_training_parameters.empty())
    {
      libmesh_error_msg_if((global_index < this->get_first_local_training_index()) ||
                           (global_index >= this->get_last_local_training_index()),
                           "Error: index "
                           << global_index
                           << " must be within range: "
                           << this->get_first_local_training_index()
                           << " - "
                           << this->get_last_local_training_index());

      const numeric_index_type local_index = global_index - get_first_local_training_index();
      for (const auto & [param_name, sample_vector] : _training_parameters)
        params.set_value(param_name, sample_vector[local_index]);

      // Copy all extra values into the new RBParameters.
      // We assume that the samples may be indexed differently for extra parameters,
      // so we don't just copy the local_index value.
      const auto & mine = get_parameters();
      for (const auto & [key, extra_sample_vector] :
           as_range(mine.extra_begin(), mine.extra_end()))
        {
          for (const auto idx : index_range(extra_sample_vector))
            params.set_extra_value(key, idx, extra_sample_vector[idx]);
        }
    }

  return params;
}

get_preevaluate_thetas_flag()

bool libMesh::RBConstruction::get_preevaluate_thetas_flag

(

)

const

Get/set flag to pre-evaluate the theta functions.

Definition at line 2755 of file rb_construction.C.

References _preevaluate_thetas_flag.

Referenced by compute_max_error_bound(), get_RB_error_bound(), and train_reduced_basis_with_greedy().

{
  return _preevaluate_thetas_flag;
}

get_project_with_constraints()

bool libMesh::System::get_project_with_constraints ( )

inlineinherited

Setter and getter functions for project_with_constraints boolean.

Definition at line 1837 of file system.h.

References libMesh::System::project_with_constraints.

Referenced by libMesh::AdjointRefinementEstimator::estimate_error().

  {
    return project_with_constraints;
  }

get_qoi_error_estimate_value()

Number libMesh::System::get_qoi_error_estimate_value ( unsigned int  qoi_index ) const

inherited

Definition at line 2206 of file system.C.

References libMesh::System::_qoi_error_estimates, and libMesh::libmesh_assert().

Referenced by libMesh::TwostepTimeSolver::integrate_adjoint_refinement_error_estimate(), and main().

{
  libmesh_assert(qoi_index < _qoi_error_estimates.size());
  return _qoi_error_estimates[qoi_index];
}

get_qoi_value()

Number libMesh::System::get_qoi_value ( unsigned int  qoi_index ) const

inherited

Definition at line 2179 of file system.C.

References libMesh::System::_qoi, and libMesh::libmesh_assert().

Referenced by libMesh::Euler2Solver::integrate_qoi_timestep(), libMesh::TwostepTimeSolver::integrate_qoi_timestep(), libMesh::EulerSolver::integrate_qoi_timestep(), main(), libMesh::ImplicitSystem::qoi_parameter_hessian(), and libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product().

{
  libmesh_assert(qoi_index < _qoi.size());
  return _qoi[qoi_index];
}

get_qoi_values()

std::vector< Number > libMesh::System::get_qoi_values ( ) const

inherited

Returns a copy of qoi, not a reference.

Definition at line 2186 of file system.C.

References libMesh::System::_qoi.

Referenced by libMesh::ImplicitSystem::adjoint_qoi_parameter_sensitivity(), libMesh::FEMSystem::assemble_qoi(), libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity(), libMesh::ImplicitSystem::qoi_parameter_hessian(), and libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product().

{
  return this->_qoi;
}

get_rb_assembly_expansion()

RBAssemblyExpansion & libMesh::RBConstruction::get_rb_assembly_expansion

(

)

Returns

A reference to the rb_assembly_expansion object

Definition at line 414 of file rb_construction.C.

References rb_assembly_expansion.

Referenced by libMesh::TransientRBConstruction::assemble_Mq_matrix(), libMesh::TransientRBConstruction::initialize_rb_construction(), and initialize_rb_construction().

{
  libmesh_error_msg_if(!rb_assembly_expansion, "Error: RBAssemblyExpansion object hasn't been initialized yet");

  return *rb_assembly_expansion;
}

get_RB_error_bound()

Real libMesh::RBConstruction::get_RB_error_bound ( )

protectedvirtual

Returns

The RB error bound for the current parameters.

Used in the Greedy algorithm to select the next parameter.

Definition at line 1784 of file rb_construction.C.

References abs_training_tolerance, get_current_training_parameter_index(), libMesh::RBEvaluation::get_error_bound_normalization(), get_evaluated_thetas(), libMesh::RBParametrized::get_parameters(), get_preevaluate_thetas_flag(), get_rb_evaluation(), normalize_rb_bound_in_greedy, libMesh::RBEvaluation::rb_solve(), libMesh::Real, and libMesh::RBParametrized::set_parameters().

Referenced by compute_max_error_bound().

{
  get_rb_evaluation().set_parameters( get_parameters() );

  Real error_bound = 0.;
  if (get_preevaluate_thetas_flag())
    {
      // Obtain the pre-evaluated theta functions from the current training parameter index
      const auto & evaluated_thetas = get_evaluated_thetas(get_current_training_parameter_index());
      error_bound = get_rb_evaluation().rb_solve(get_rb_evaluation().get_n_basis_functions(),
                                                 &evaluated_thetas);
    }
  else
    error_bound = get_rb_evaluation().rb_solve(get_rb_evaluation().get_n_basis_functions());


  if (normalize_rb_bound_in_greedy)
    {
      Real error_bound_normalization = get_rb_evaluation().get_error_bound_normalization();

      if ((error_bound < abs_training_tolerance) ||
          (error_bound_normalization < abs_training_tolerance))
        {
          // We don't want to normalize this error bound if the bound or the
          // normalization value are below the absolute tolerance. Hence do nothing
          // in this case.
        }
      else
        error_bound /= error_bound_normalization;
    }

  return error_bound;
}

get_rb_evaluation() [1/2]

RBEvaluation & libMesh::RBConstruction::get_rb_evaluation

(

)

Get a reference to the RBEvaluation object.

Definition at line 179 of file rb_construction.C.

References rb_eval.

Referenced by libMesh::TransientRBConstruction::add_IC_to_RB_space(), libMesh::TransientRBConstruction::assemble_affine_expansion(), compute_Fq_representor_innerprods(), compute_max_error_bound(), compute_output_dual_innerprods(), compute_residual_dual_norm_slow(), enrich_basis_from_rhs_terms(), libMesh::TransientRBConstruction::enrich_RB_space(), enrich_RB_space(), get_greedy_parameter(), get_RB_error_bound(), get_rb_theta_expansion(), greedy_termination_test(), load_basis_function(), libMesh::TransientRBConstruction::load_rb_solution(), load_rb_solution(), main(), preevaluate_thetas(), print_basis_function_orthogonality(), libMesh::TransientRBConstruction::process_parameters_file(), libMesh::TransientRBConstruction::read_riesz_representors_from_files(), read_riesz_representors_from_files(), recompute_all_residual_terms(), libMesh::TransientRBConstruction::set_error_temporal_data(), train_reduced_basis_with_greedy(), train_reduced_basis_with_POD(), update_greedy_param_list(), libMesh::TransientRBConstruction::update_RB_initial_condition_all_N(), libMesh::TransientRBConstruction::update_RB_system_matrices(), update_RB_system_matrices(), libMesh::TransientRBConstruction::update_residual_terms(), update_residual_terms(), libMesh::TransientRBConstruction::write_riesz_representors_to_files(), and write_riesz_representors_to_files().

{
  libmesh_error_msg_if(!rb_eval, "Error: RBEvaluation object hasn't been initialized yet");

  return *rb_eval;
}

get_rb_evaluation() [2/2]

const RBEvaluation & libMesh::RBConstruction::get_rb_evaluation

(

)

const

Definition at line 186 of file rb_construction.C.

References rb_eval.

{
  libmesh_error_msg_if(!rb_eval, "Error: RBEvaluation object hasn't been initialized yet");

  return *rb_eval;
}

get_rb_theta_expansion() [1/2]

RBThetaExpansion & libMesh::RBConstruction::get_rb_theta_expansion

(

)

Get a reference to the RBThetaExpansion object that that belongs to rb_eval.

Definition at line 198 of file rb_construction.C.

References get_rb_evaluation(), and libMesh::RBEvaluation::get_rb_theta_expansion().

Referenced by add_scaled_Aq(), libMesh::TransientRBConstruction::add_scaled_mass_matrix(), libMesh::TransientRBConstruction::allocate_data_structures(), allocate_data_structures(), libMesh::TransientRBConstruction::assemble_all_affine_operators(), assemble_all_affine_operators(), assemble_all_affine_vectors(), assemble_all_output_vectors(), assemble_Aq_matrix(), assemble_Fq_vector(), assemble_inner_product_matrix(), libMesh::TransientRBConstruction::assemble_Mq_matrix(), compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), enrich_basis_from_rhs_terms(), libMesh::TransientRBConstruction::get_all_matrices(), get_all_matrices(), get_all_vectors(), get_Aq(), get_Fq(), libMesh::TransientRBConstruction::get_M_q(), get_non_dirichlet_Aq(), get_non_dirichlet_Fq(), libMesh::TransientRBConstruction::get_non_dirichlet_M_q(), get_non_dirichlet_output_vector(), get_output_vector(), get_output_vectors(), libMesh::TransientRBConstruction::initialize_rb_construction(), initialize_rb_construction(), libMesh::TransientRBConstruction::mass_matrix_scaled_matvec(), libMesh::TransientRBConstruction::print_info(), print_info(), libMesh::TransientRBConstruction::truth_assembly(), truth_assembly(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::TransientRBConstruction::update_RB_system_matrices(), update_RB_system_matrices(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

{
  return get_rb_evaluation().get_rb_theta_expansion();
}

get_rb_theta_expansion() [2/2]

const RBThetaExpansion & libMesh::RBConstruction::get_rb_theta_expansion

(

)

const

Definition at line 203 of file rb_construction.C.

References get_rb_evaluation(), and libMesh::RBEvaluation::get_rb_theta_expansion().

{
  return get_rb_evaluation().get_rb_theta_expansion();
}

get_RB_training_type()

const std::string & libMesh::RBConstruction::get_RB_training_type

(

)

const

Definition at line 1698 of file rb_construction.C.

References RB_training_type.

Referenced by print_info(), libMesh::TransientRBConstruction::train_reduced_basis(), and train_reduced_basis().

{
  return RB_training_type;
}

get_rel_training_tolerance()

Real libMesh::RBConstruction::get_rel_training_tolerance ( ) const

inline

Definition at line 220 of file rb_construction.h.

References rel_training_tolerance.

Referenced by print_info().

{ return rel_training_tolerance; }

get_sensitivity_rhs() [1/2]

NumericVector< Number > & libMesh::System::get_sensitivity_rhs ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's sensitivity rhs vectors, by default the one corresponding to the first parameter. By default these vectors are built by the library, using finite differences, when assemble_residual_derivatives() is called.

When assembled, this vector should hold -(partial R / partial p_i)

Definition at line 1324 of file system.C.

References libMesh::System::get_vector().

Referenced by libMesh::ImplicitSystem::adjoint_qoi_parameter_sensitivity(), and libMesh::ImplicitSystem::sensitivity_solve().

{
  std::ostringstream sensitivity_rhs_name;
  sensitivity_rhs_name << "sensitivity_rhs" << i;

  return this->get_vector(sensitivity_rhs_name.str());
}

get_sensitivity_rhs() [2/2]

const NumericVector< Number > & libMesh::System::get_sensitivity_rhs ( unsigned int  i = 0 ) const

inherited

Returns

A reference to one of the system's sensitivity rhs vectors, by default the one corresponding to the first parameter.

Definition at line 1334 of file system.C.

References libMesh::System::get_vector().

{
  std::ostringstream sensitivity_rhs_name;
  sensitivity_rhs_name << "sensitivity_rhs" << i;

  return this->get_vector(sensitivity_rhs_name.str());
}

get_sensitivity_solution() [1/2]

NumericVector< Number > & libMesh::System::get_sensitivity_solution ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's solution sensitivity vectors, by default the one corresponding to the first parameter.

Definition at line 1179 of file system.C.

References libMesh::System::get_vector().

Referenced by libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity(), libMesh::ImplicitSystem::qoi_parameter_hessian(), and libMesh::ImplicitSystem::sensitivity_solve().

{
  std::ostringstream sensitivity_name;
  sensitivity_name << "sensitivity_solution" << i;

  return this->get_vector(sensitivity_name.str());
}

get_sensitivity_solution() [2/2]

const NumericVector< Number > & libMesh::System::get_sensitivity_solution ( unsigned int  i = 0 ) const

inherited

Returns

A reference to one of the system's solution sensitivity vectors, by default the one corresponding to the first parameter.

Definition at line 1189 of file system.C.

References libMesh::System::get_vector().

{
  std::ostringstream sensitivity_name;
  sensitivity_name << "sensitivity_solution" << i;

  return this->get_vector(sensitivity_name.str());
}

get_shell_matrix()

ShellMatrix<Number>* libMesh::LinearImplicitSystem::get_shell_matrix ( )

inlineinherited

Returns

A pointer to the currently attached shell matrix, if any, otherwise nullptr.

Definition at line 182 of file linear_implicit_system.h.

References libMesh::LinearImplicitSystem::_shell_matrix.

{ return _shell_matrix; }

get_static_condensation()

StaticCondensation & libMesh::ImplicitSystem::get_static_condensation ( )

inlineinherited

Returns

The static condensation system matrix

Definition at line 362 of file implicit_system.h.

References libMesh::ImplicitSystem::_sc_system_matrix, and libMesh::libmesh_assert().

Referenced by assemble_poisson().

{
  libmesh_assert(_sc_system_matrix);
  return *_sc_system_matrix;
}

get_system_matrix() [1/2]

const SparseMatrix< Number > & libMesh::ImplicitSystem::get_system_matrix ( ) const

inherited

Returns

A const reference to the system's primary matrix.

Definition at line 1251 of file implicit_system.C.

References libMesh::System::get_matrix(), libMesh::libmesh_assert(), and libMesh::ImplicitSystem::matrix.

Referenced by libMesh::HDGProblem::add_matrix(), assemble(), LinearElasticity::assemble(), assemble_1D(), assemble_biharmonic(), assemble_divgrad(), assemble_elasticity(), assemble_ellipticdg(), assemble_graddiv(), assemble_helmholtz(), assemble_laplace(), assemble_matrix_and_rhs(), assemble_poisson(), assemble_shell(), assemble_stokes(), assemble_wave(), assembly_with_dg_fem_context(), libMesh::HDGProblem::jacobian(), libMesh::libmesh_petsc_snes_residual(), main(), OverlappingCouplingGhostingTest::run_sparsity_pattern_test(), and libMesh::EigenTimeSolver::solve().

{
  libmesh_assert(matrix);
  libmesh_assert_equal_to(&get_matrix("System Matrix"), matrix);
  return *matrix;
}

get_system_matrix() [2/2]

SparseMatrix< Number > & libMesh::ImplicitSystem::get_system_matrix ( )

inherited

Returns

A reference to the system's primary matrix.

Definition at line 1260 of file implicit_system.C.

References libMesh::System::get_matrix(), libMesh::libmesh_assert(), and libMesh::ImplicitSystem::matrix.

{
  libmesh_assert(matrix);
  libmesh_assert_equal_to(&get_matrix("System Matrix"), matrix);
  return *matrix;
}

get_vector() [1/4]

const NumericVector< Number > & libMesh::System::get_vector ( std::string_view  vec_name ) const

inherited

Returns

A const reference to this system's additional vector named vec_name. Access is only granted when the vector is already properly initialized.

Definition at line 931 of file system.C.

References libMesh::System::_vectors.

Referenced by libMesh::UniformRefinementEstimator::_estimate_error(), add_M_C_K_helmholtz(), libMesh::AdaptiveTimeSolver::adjoint_advance_timestep(), libMesh::UnsteadySolver::adjoint_advance_timestep(), libMesh::NewmarkSolver::advance_timestep(), libMesh::AdaptiveTimeSolver::advance_timestep(), libMesh::UnsteadySolver::advance_timestep(), apply_initial(), assemble(), libMesh::System::compare(), libMesh::NewmarkSolver::compute_initial_accel(), libMesh::UnsteadySolver::du(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::System::get_adjoint_rhs(), libMesh::System::get_adjoint_solution(), libMesh::System::get_sensitivity_rhs(), libMesh::System::get_sensitivity_solution(), libMesh::System::get_weighted_sensitivity_adjoint_solution(), libMesh::System::get_weighted_sensitivity_solution(), libMesh::NewmarkSystem::initial_conditions(), AssembleOptimization::lower_and_upper_bounds(), main(), libMesh::NewmarkSolver::project_initial_accel(), libMesh::SecondOrderUnsteadySolver::project_initial_rate(), libMesh::InterMeshProjection::project_system_vectors(), libMesh::SecondOrderUnsteadySolver::reinit(), libMesh::UnsteadySolver::reinit(), libMesh::FileSolutionHistory::retrieve(), libMesh::UnsteadySolver::retrieve_timestep(), libMesh::MemoryHistoryData::retrieve_vectors(), libMesh::TwostepTimeSolver::solve(), libMesh::FrequencySystem::solve(), libMesh::UnsteadySolver::update(), libMesh::NewmarkSystem::update_rhs(), and libMesh::NewmarkSystem::update_u_v_a().

{
  return *(libmesh_map_find(_vectors, vec_name));
}

get_vector() [2/4]

NumericVector< Number > & libMesh::System::get_vector ( std::string_view  vec_name )

inherited

Returns

A writable reference to this system's additional vector named vec_name. Access is only granted when the vector is already properly initialized.

Definition at line 938 of file system.C.

References libMesh::System::_vectors.

{
  return *(libmesh_map_find(_vectors, vec_name));
}

get_vector() [3/4]

const NumericVector< Number > & libMesh::System::get_vector ( const unsigned int  vec_num ) const

inherited

Returns

A const reference to this system's additional vector number vec_num (where the vectors are counted starting with 0).

Definition at line 945 of file system.C.

References libMesh::System::_vectors, and libMesh::System::vectors_begin().

{
  // If we don't have that many vectors, throw an error
  libmesh_assert_less(vec_num, _vectors.size());

  // Otherwise return a reference to the vec_num'th vector
  auto it = vectors_begin();
  std::advance(it, vec_num);
  return *(it->second);
}

get_vector() [4/4]

NumericVector< Number > & libMesh::System::get_vector ( const unsigned int  vec_num )

inherited

Returns

A writable reference to this system's additional vector number vec_num (where the vectors are counted starting with 0).

Definition at line 958 of file system.C.

References libMesh::System::_vectors, and libMesh::System::vectors_begin().

{
  // If we don't have that many vectors, throw an error
  libmesh_assert_less(vec_num, _vectors.size());

  // Otherwise return a reference to the vec_num'th vector
  auto it = vectors_begin();
  std::advance(it, vec_num);
  return *(it->second);
}

get_weighted_sensitivity_adjoint_solution() [1/2]

NumericVector< Number > & libMesh::System::get_weighted_sensitivity_adjoint_solution ( unsigned int  i = 0 )

inherited

Returns

A reference to one of the system's weighted sensitivity adjoint solution vectors, by default the one corresponding to the first qoi.

Definition at line 1264 of file system.C.

References libMesh::System::get_vector().

Referenced by libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product(), and libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve().

{
  std::ostringstream adjoint_name;
  adjoint_name << "weighted_sensitivity_adjoint_solution" << i;

  return this->get_vector(adjoint_name.str());
}

get_weighted_sensitivity_adjoint_solution() [2/2]

const NumericVector< Number > & libMesh::System::get_weighted_sensitivity_adjoint_solution ( unsigned int  i = 0 ) const

inherited

Returns

A reference to one of the system's weighted sensitivity adjoint solution vectors, by default the one corresponding to the first qoi.

Definition at line 1274 of file system.C.

References libMesh::System::get_vector().

{
  std::ostringstream adjoint_name;
  adjoint_name << "weighted_sensitivity_adjoint_solution" << i;

  return this->get_vector(adjoint_name.str());
}

get_weighted_sensitivity_solution() [1/2]

NumericVector< Number > & libMesh::System::get_weighted_sensitivity_solution ( )

inherited

Returns

A reference to the solution of the last weighted sensitivity solve

Definition at line 1206 of file system.C.

References libMesh::System::get_vector().

Referenced by libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product(), and libMesh::ImplicitSystem::weighted_sensitivity_solve().

{
  return this->get_vector("weighted_sensitivity_solution");
}

get_weighted_sensitivity_solution() [2/2]

const NumericVector< Number > & libMesh::System::get_weighted_sensitivity_solution ( ) const

inherited

Returns

A reference to the solution of the last weighted sensitivity solve

Definition at line 1213 of file system.C.

References libMesh::System::get_vector().

{
  return this->get_vector("weighted_sensitivity_solution");
}

greedy_termination_test()

bool libMesh::RBConstruction::greedy_termination_test ( Real  abs_greedy_error, Real  initial_greedy_error, int  count  )

protectedvirtual

Function that indicates when to terminate the Greedy basis training.

Override in subclasses to specialize.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 1565 of file rb_construction.C.

References abs_training_tolerance, exit_on_repeated_greedy_parameters, libMesh::RBEvaluation::get_n_basis_functions(), get_Nmax(), libMesh::RBParametrized::get_parameters(), get_rb_evaluation(), libMesh::RBEvaluation::greedy_param_list, libMesh::out, libMesh::Real, and rel_training_tolerance.

Referenced by libMesh::TransientRBConstruction::greedy_termination_test(), and train_reduced_basis_with_greedy().

{
  if (abs_greedy_error < this->abs_training_tolerance)
    {
      libMesh::out << "Absolute error tolerance reached." << std::endl;
      return true;
    }

  Real rel_greedy_error = abs_greedy_error/initial_error;
  if (rel_greedy_error < this->rel_training_tolerance)
    {
      libMesh::out << "Relative error tolerance reached." << std::endl;
      return true;
    }

  RBEvaluation & rbe = get_rb_evaluation();

  if (rbe.get_n_basis_functions() >= this->get_Nmax())
    {
      libMesh::out << "Maximum number of basis functions reached: Nmax = "
                   << get_Nmax() << std::endl;
      return true;
    }

  if (exit_on_repeated_greedy_parameters)
    for (auto & plist : rbe.greedy_param_list)
      if (plist == get_parameters())
        {
          libMesh::out << "Exiting greedy because the same parameters were selected twice" << std::endl;
          return true;
        }

  return false;
}

has_constraint_object()

bool libMesh::System::has_constraint_object ( ) const

inherited

Returns

true if there is a user-defined constraint object attached to this object, false otherwise. Calling System:: get_constraint_object() when there is no user-defined constraint object attached leads to either undefined behavior (dereferencing a nullptr) or an assert (in dbg mode) so you should call this function first unless you are sure there is a user-defined constraint object attached.

Definition at line 2014 of file system.C.

References libMesh::System::_constrain_system_object.

{
  return _constrain_system_object != nullptr;
}

has_static_condensation()

bool libMesh::System::has_static_condensation ( ) const

inherited

Returns

Whether this system will be statically condensed

Definition at line 2664 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::has_static_condensation().

Referenced by libMesh::System::add_matrix(), assemble_poisson(), libMesh::LinearImplicitSystem::clear(), libMesh::ImplicitSystem::clear(), libMesh::NonlinearImplicitSystem::clear(), libMesh::System::get_info(), libMesh::ImplicitSystem::ImplicitSystem(), libMesh::LinearImplicitSystem::LinearImplicitSystem(), and libMesh::NonlinearImplicitSystem::NonlinearImplicitSystem().

{
  return this->get_dof_map().has_static_condensation();
}

has_variable()

bool libMesh::System::has_variable ( std::string_view  var ) const

inherited

Returns

true if a variable named var exists in this System

Definition at line 1393 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::has_variable().

Referenced by libMesh::GMVIO::copy_nodal_solution(), and main().

{
  return this->get_dof_map().has_variable(var);
}

have_matrix()

bool libMesh::System::have_matrix ( std::string_view  mat_name ) const

inlineinherited

Returns

true if this System has a matrix associated with the given name, false otherwise.

Definition at line 1933 of file system.h.

References libMesh::System::_matrices.

Referenced by libMesh::EigenTimeSolver::init().

{ return _matrices.count(mat_name); }

have_vector()

bool libMesh::System::have_vector ( std::string_view  vec_name ) const

inlineinherited

Returns

true if this System has a vector associated with the given name, false otherwise.

Definition at line 2491 of file system.h.

References libMesh::System::_vectors.

{
  return (_vectors.count(vec_name));
}

hide_output()

bool& libMesh::System::hide_output ( )

inlineinherited

Returns

A writable reference to a boolean that determines if this system can be written to file or not. If set to true, then EquationSystems::write will ignore this system.

Definition at line 1852 of file system.h.

References libMesh::System::_hide_output.

Referenced by libMesh::StaticCondensationDofMap::reinit(), libMesh::PetscPreconditioner< T >::set_hypre_ads_data(), and libMesh::PetscPreconditioner< T >::set_hypre_ams_data().

{ return _hide_output; }

identify_variable_groups() [1/2]

bool libMesh::System::identify_variable_groups ( ) const

inherited

Returns

true when VariableGroup structures should be automatically identified, false otherwise.

Definition at line 2679 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::identify_variable_groups().

{
  return this->get_dof_map().identify_variable_groups();
}

identify_variable_groups() [2/2]

void libMesh::System::identify_variable_groups ( const bool  ivg )

inherited

Toggle automatic VariableGroup identification.

Definition at line 2684 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::identify_variable_groups().

{
  this->get_dof_map().identify_variable_groups(ivg);
}

increment_constructor_count() [1/2]

void libMesh::ReferenceCounter::increment_constructor_count ( const std::string &  name )

inlineprotectednoexceptinherited

Increments the construction counter.

Should be called in the constructor of any derived class that will be reference counted.

Definition at line 183 of file reference_counter.h.

References libMesh::err, libMesh::BasicOStreamProxy< charT, traits >::get(), libMesh::Quality::name(), and libMesh::Threads::spin_mtx.

Referenced by libMesh::ReferenceCountedObject< RBParametrized >::ReferenceCountedObject().

{
  libmesh_try
  {
    Threads::spin_mutex::scoped_lock lock(Threads::spin_mtx);
    std::pair<unsigned int, unsigned int> & p = _counts[name];
    p.first++;
  }
  libmesh_catch (...)
  {
    auto stream = libMesh::err.get();
    stream->exceptions(stream->goodbit); // stream must not throw
    libMesh::err << "Encountered unrecoverable error while calling "
                 << "ReferenceCounter::increment_constructor_count() "
                 << "for a(n) " << name << " object." << std::endl;
    std::terminate();
  }
}

increment_constructor_count() [2/2]

void libMesh::ReferenceCounter::increment_constructor_count ( const std::string &  name )

inlineprotectednoexceptinherited

Increments the construction counter.

Should be called in the constructor of any derived class that will be reference counted.

Definition at line 183 of file reference_counter.h.

References libMesh::err, libMesh::BasicOStreamProxy< charT, traits >::get(), libMesh::Quality::name(), and libMesh::Threads::spin_mtx.

Referenced by libMesh::ReferenceCountedObject< RBParametrized >::ReferenceCountedObject().

{
  libmesh_try
  {
    Threads::spin_mutex::scoped_lock lock(Threads::spin_mtx);
    std::pair<unsigned int, unsigned int> & p = _counts[name];
    p.first++;
  }
  libmesh_catch (...)
  {
    auto stream = libMesh::err.get();
    stream->exceptions(stream->goodbit); // stream must not throw
    libMesh::err << "Encountered unrecoverable error while calling "
                 << "ReferenceCounter::increment_constructor_count() "
                 << "for a(n) " << name << " object." << std::endl;
    std::terminate();
  }
}

increment_destructor_count() [1/2]

void libMesh::ReferenceCounter::increment_destructor_count ( const std::string &  name )

inlineprotectednoexceptinherited

Increments the destruction counter.

Should be called in the destructor of any derived class that will be reference counted.

Definition at line 207 of file reference_counter.h.

References libMesh::err, libMesh::BasicOStreamProxy< charT, traits >::get(), libMesh::Quality::name(), and libMesh::Threads::spin_mtx.

Referenced by libMesh::ReferenceCountedObject< RBParametrized >::~ReferenceCountedObject().

{
  libmesh_try
  {
    Threads::spin_mutex::scoped_lock lock(Threads::spin_mtx);
    std::pair<unsigned int, unsigned int> & p = _counts[name];
    p.second++;
  }
  libmesh_catch (...)
  {
    auto stream = libMesh::err.get();
    stream->exceptions(stream->goodbit); // stream must not throw
    libMesh::err << "Encountered unrecoverable error while calling "
                 << "ReferenceCounter::increment_destructor_count() "
                 << "for a(n) " << name << " object." << std::endl;
    std::terminate();
  }
}

increment_destructor_count() [2/2]

void libMesh::ReferenceCounter::increment_destructor_count ( const std::string &  name )

inlineprotectednoexceptinherited

Increments the destruction counter.

Should be called in the destructor of any derived class that will be reference counted.

Definition at line 207 of file reference_counter.h.

References libMesh::err, libMesh::BasicOStreamProxy< charT, traits >::get(), libMesh::Quality::name(), and libMesh::Threads::spin_mtx.

Referenced by libMesh::ReferenceCountedObject< RBParametrized >::~ReferenceCountedObject().

{
  libmesh_try
  {
    Threads::spin_mutex::scoped_lock lock(Threads::spin_mtx);
    std::pair<unsigned int, unsigned int> & p = _counts[name];
    p.second++;
  }
  libmesh_catch (...)
  {
    auto stream = libMesh::err.get();
    stream->exceptions(stream->goodbit); // stream must not throw
    libMesh::err << "Encountered unrecoverable error while calling "
                 << "ReferenceCounter::increment_destructor_count() "
                 << "for a(n) " << name << " object." << std::endl;
    std::terminate();
  }
}

init()

void libMesh::System::init ( )

inherited

Initializes degrees of freedom on the current mesh.

Sets the

Definition at line 196 of file system.C.

References libMesh::System::is_initialized(), libMesh::libmesh_assert(), and libMesh::System::reinit_mesh().

Referenced by libMesh::StaticCondensationDofMap::reinit().

{
  // Calling init() twice on the same system currently works evil
  // magic, whether done directly or via EquationSystems::read()
  libmesh_assert(!this->is_initialized());

  this->reinit_mesh();
}

init_context()

virtual void libMesh::RBConstruction::init_context ( FEMContext &  )

inlineprotectedvirtual

Initialize the FEMContext prior to performing an element loop.

Reimplement this in derived classes in order to call FE::get_*() as the particular physics requires.

Reimplemented in SimpleRBConstruction, SimpleRBConstruction, SimpleRBConstruction, SimpleRBConstruction, SimpleRBConstruction, and ElasticityRBConstruction.

Definition at line 825 of file rb_construction.h.

Referenced by add_scaled_matrix_and_vector().

  {
    // Failing to rederive init_context() means your FE objects don't
    // know what to compute.
    libmesh_deprecated();
  }

init_data()

void libMesh::RBConstructionBase< LinearImplicitSystem >::init_data ( )

protectedvirtualinherited

Initializes the member data fields associated with the system, so that, e.g., assemble() may be used.

Reimplemented from libMesh::LinearImplicitSystem.

Reimplemented in SimpleRBConstruction, SimpleRBConstruction, SimpleRBConstruction, SimpleRBConstruction, SimpleRBConstruction, SimpleRBConstruction, and ElasticityRBConstruction.

Definition at line 123 of file rb_construction_base.C.

Referenced by SimpleRBConstruction::init_data().

{
  Base::init_data();

  // Initialize the inner product storage vector, which is useful for
  // storing intermediate results when evaluating inner products
  inner_product_storage_vector = NumericVector<Number>::build(this->comm());
  inner_product_storage_vector->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
}

init_matrices()

void libMesh::System::init_matrices ( )

protectedvirtualinherited

Initializes the matrices associated with this system.

Reimplemented in libMesh::EigenSystem.

Definition at line 311 of file system.C.

References libMesh::System::_matrices, libMesh::System::_matrices_initialized, libMesh::System::_matrix_types, libMesh::System::_prefer_hash_table_matrix_assembly, libMesh::System::_require_sparsity_pattern, libMesh::DofMap::attach_matrix(), libMesh::DofMap::compute_sparsity(), libMesh::System::get_dof_map(), libMesh::System::get_mesh(), libMesh::SparseMatrix< T >::initialized(), libMesh::DofMap::is_attached(), libMesh::libmesh_assert(), and libMesh::System::name().

Referenced by libMesh::System::init_data(), and libMesh::EigenSystem::init_matrices().

{
  parallel_object_only();

  // No matrices to init
  if (_matrices.empty())
    {
      // any future matrices to be added will need their own
      // initialization
      _matrices_initialized = true;

      return;
    }

  // Check for quick return in case the first matrix
  // (and by extension all the matrices) has already
  // been initialized
  if (_matrices.begin()->second->initialized())
    {
      libmesh_assert(_matrices_initialized);
      return;
    }

  _matrices_initialized = true;

  // Tell the matrices about the dof map, and vice versa
  for (auto & pr : _matrices)
    {
      SparseMatrix<Number> & m = *(pr.second);
      libmesh_assert (!m.initialized());

      // We want to allow repeated init() on systems, but we don't
      // want to attach the same matrix to the DofMap twice
      if (!this->get_dof_map().is_attached(m))
        this->get_dof_map().attach_matrix(m);

      // If the user has already explicitly requested that this matrix use a hash table, then we
      // always honor that
      const bool use_hash =
          pr.second->use_hash_table() ||
          (this->_prefer_hash_table_matrix_assembly && pr.second->supports_hash_table());
      pr.second->use_hash_table(use_hash);
      // Make this call after we've determined whether the matrix is using a hash table
      if (pr.second->require_sparsity_pattern())
        this->_require_sparsity_pattern = true;
    }

  // Compute the sparsity pattern for the current
  // mesh and DOF distribution.  This also updates
  // additional matrices, \p DofMap now knows them
  if (this->_require_sparsity_pattern)
    this->get_dof_map().compute_sparsity(this->get_mesh());

  // Initialize matrices and set to zero
  for (auto & [name, mat] : _matrices)
    {
      mat->init(_matrix_types[name]);
      mat->zero();
    }
}

init_qois()

void libMesh::System::init_qois ( unsigned int  n_qois )

inherited

Accessors for qoi and qoi_error_estimates vectors.

Definition at line 2164 of file system.C.

References libMesh::System::_qoi, libMesh::System::_qoi_error_estimates, and libMesh::System::n_qois().

Referenced by CoupledSystemQoI::init_qoi_count(), LaplaceQoI::init_qoi_count(), and main().

{
  _qoi.resize(n_qois);
  _qoi_error_estimates.resize(n_qois);
}

initialize_parameters() [1/2]

void libMesh::RBParametrized::initialize_parameters ( const RBParameters &  mu_min_in, const RBParameters &  mu_max_in, const std::map< std::string, std::vector< Real >> &  discrete_parameter_values  )

inherited

Initialize the parameter ranges and set current_parameters.

Parameter ranges are inclusive. The input min/max RBParameters should have exactly 1 sample each. Vector-valued samples are not currently supported for the min/max parameters or for discrete parameters.

Definition at line 53 of file rb_parametrized.C.

References libMesh::RBParametrized::_discrete_parameter_values, libMesh::RBParameters::begin_serialized(), libMesh::RBParameters::end_serialized(), libMesh::RBParameters::n_parameters(), libMesh::RBParameters::n_samples(), libMesh::Quality::name(), libMesh::RBParametrized::parameters_initialized, libMesh::RBParametrized::parameters_max, libMesh::RBParametrized::parameters_min, libMesh::Real, libMesh::RBParametrized::set_parameters(), and libMesh::RBParameters::set_value().

Referenced by enrich_basis_from_rhs_terms(), libMesh::RBParametrized::initialize_parameters(), libMesh::RBDataDeserialization::load_parameter_ranges(), libMesh::RBSCMConstruction::perform_SCM_greedy(), libMesh::RBSCMConstruction::process_parameters_file(), libMesh::RBParametrized::read_parameter_data_from_files(), libMesh::RBEIMConstruction::set_rb_construction_parameters(), set_rb_construction_parameters(), RBParametersTest::testRBParametrized(), libMesh::RBEIMConstruction::train_eim_approximation_with_greedy(), libMesh::RBEIMConstruction::train_eim_approximation_with_POD(), train_reduced_basis_with_greedy(), and train_reduced_basis_with_POD().

{
  // Check that the min/max vectors have the same size.
  libmesh_error_msg_if(mu_min_in.n_parameters() != mu_max_in.n_parameters(),
                       "Error: Invalid mu_min/mu_max in initialize_parameters(), different number of parameters.");
  libmesh_error_msg_if(mu_min_in.n_samples() != 1 ||
                       mu_max_in.n_samples() != 1,
                       "Error: Invalid mu_min/mu_max in initialize_parameters(), only 1 sample supported.");

  // Ensure all the values are valid for min and max.
  auto pr_min = mu_min_in.begin_serialized();
  auto pr_max = mu_max_in.begin_serialized();
  for (; pr_min != mu_min_in.end_serialized(); ++pr_min, ++pr_max)
    libmesh_error_msg_if((*pr_min).second > (*pr_max).second,
                         "Error: Invalid mu_min/mu_max in RBParameters constructor.");

  parameters_min = mu_min_in;
  parameters_max = mu_max_in;

  // Add in min/max values due to the discrete parameters
  for (const auto & [name, vals] : discrete_parameter_values)
    {
      libmesh_error_msg_if(vals.empty(), "Error: List of discrete parameters for " << name << " is empty.");

      Real min_val = *std::min_element(vals.begin(), vals.end());
      Real max_val = *std::max_element(vals.begin(), vals.end());

      libmesh_assert_less_equal(min_val, max_val);

      parameters_min.set_value(name, min_val);
      parameters_max.set_value(name, max_val);
    }

    _discrete_parameter_values = discrete_parameter_values;

  parameters_initialized = true;

  // Initialize the current parameters to parameters_min
  set_parameters(parameters_min);
}

initialize_parameters() [2/2]

void libMesh::RBParametrized::initialize_parameters ( const RBParametrized &  rb_parametrized )

inherited

Initialize the parameter ranges and set current_parameters.

Definition at line 96 of file rb_parametrized.C.

References libMesh::RBParametrized::get_discrete_parameter_values(), libMesh::RBParametrized::get_parameters_max(), libMesh::RBParametrized::get_parameters_min(), and libMesh::RBParametrized::initialize_parameters().

{
  initialize_parameters(rb_parametrized.get_parameters_min(),
                        rb_parametrized.get_parameters_max(),
                        rb_parametrized.get_discrete_parameter_values());
}

initialize_rb_construction()

void libMesh::RBConstruction::initialize_rb_construction ( bool  skip_matrix_assembly = false, bool  skip_vector_assembly = false  )

virtual

Allocate all the data structures necessary for the construction stage of the RB method.

This function also performs matrix and vector assembly of the "truth" affine expansion.

We can optionally skip the matrix or vector assembly steps by setting skip_matrix_assembly = true, or skip_vector_assembly = true, respectively.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 466 of file rb_construction.C.

References allocate_data_structures(), assemble_affine_expansion(), libMesh::LinearImplicitSystem::get_linear_solver(), libMesh::RBThetaExpansion::get_n_outputs(), get_rb_assembly_expansion(), get_rb_theta_expansion(), inner_product_solver, and libMesh::LinearSolver< T >::reuse_preconditioner().

Referenced by libMesh::TransientRBConstruction::initialize_rb_construction(), and main().

{
  if (!skip_matrix_assembly && !skip_vector_assembly)
    {
      // Check that the theta and assembly objects are consistently sized
      libmesh_assert_equal_to (get_rb_theta_expansion().get_n_A_terms(), get_rb_assembly_expansion().get_n_A_terms());
      libmesh_assert_equal_to (get_rb_theta_expansion().get_n_F_terms(), get_rb_assembly_expansion().get_n_F_terms());
      libmesh_assert_equal_to (get_rb_theta_expansion().get_n_outputs(), get_rb_assembly_expansion().get_n_outputs());
      for (unsigned int i=0; i<get_rb_theta_expansion().get_n_outputs(); i++)
        {
          libmesh_assert_equal_to (get_rb_theta_expansion().get_n_output_terms(i),
                                   get_rb_assembly_expansion().get_n_output_terms(i));
        }
    }

  // Perform the initialization
  allocate_data_structures();
  assemble_affine_expansion(skip_matrix_assembly, skip_vector_assembly);

  // inner_product_solver performs solves with the same matrix every time
  // hence we can set reuse_preconditioner(true).
  inner_product_solver->reuse_preconditioner(true);

  // The primary solver is used for truth solves and other solves that
  // require different matrices, so set reuse_preconditioner(false).
  get_linear_solver()->reuse_preconditioner(false);

}

initialize_training_parameters()

void libMesh::RBConstructionBase< LinearImplicitSystem >::initialize_training_parameters ( const RBParameters &  mu_min, const RBParameters &  mu_max, const unsigned int  n_global_training_samples, const std::map< std::string, bool > &  log_param_scale, const bool  deterministic = true  )

virtualinherited

Initialize the parameter ranges and indicate whether deterministic or random training parameters should be used and whether or not we want the parameters to be scaled logarithmically.

n_global_training_samples is the total number of samples to generate, which will be distributed across all the processors.

Definition at line 292 of file rb_construction_base.C.

Referenced by set_rb_construction_parameters().

{
  if (!is_quiet())
    {
      // Print out some info about the training set initialization
      libMesh::out << "Initializing training parameters with "
                  << (deterministic ? "deterministic " : "random " )
                  << "training set..." << std::endl;

      for (const auto & pr : log_param_scale)
        libMesh::out << "Parameter "
                     << pr.first
                     << ": log scaling = "
                     << pr.second
                     << std::endl;

      libMesh::out << std::endl;
    }

  if (deterministic)
    {
      const auto [first_local_index, last_local_index] =
        generate_training_parameters_deterministic(this->comm(),
                                                   log_param_scale,
                                                   _training_parameters,
                                                   n_global_training_samples,
                                                   mu_min,
                                                   mu_max,
                                                   serial_training_set);
      _first_local_index = first_local_index;
      _n_local_training_samples = last_local_index-first_local_index;
    }
  else
    {
      // Generate random training samples for all parameters
      const auto [first_local_index, last_local_index] =
        generate_training_parameters_random(this->comm(),
                                            log_param_scale,
                                            _training_parameters,
                                            n_global_training_samples,
                                            mu_min,
                                            mu_max,
                                            this->_training_parameters_random_seed,
                                            serial_training_set);
      _first_local_index = first_local_index;
      _n_local_training_samples = last_local_index-first_local_index;
    }
  _n_global_training_samples = _n_local_training_samples;

  if (!serial_training_set)
    this->comm().sum(_n_global_training_samples);

  // For each parameter that only allows discrete values, we "snap" to the nearest
  // allowable discrete value
  if (get_n_discrete_params() > 0)
    {
      for (auto & [param_name, sample_vector] : _training_parameters)
        {
          if (is_discrete_parameter(param_name))
            {
              const std::vector<Real> & discrete_values =
                libmesh_map_find(get_discrete_parameter_values(), param_name);

              for (const auto sample_idx : index_range(sample_vector))
                {
                  // Round all values to the closest discrete value.
                  std::vector<Real> discretized_vector(sample_vector[sample_idx].size());
                  std::transform(sample_vector[sample_idx].cbegin(),
                                 sample_vector[sample_idx].cend(),
                                 discretized_vector.begin(),
                                 [&discrete_values](const Real & val) {
                                   return get_closest_value(val, discrete_values);
                                 });
                  sample_vector[sample_idx] = discretized_vector;
                }
            }
        }
    }

  _training_parameters_initialized = true;
}

is_adjoint_already_solved()

bool libMesh::System::is_adjoint_already_solved ( ) const

inlineinherited

Accessor for the adjoint_already_solved boolean.

Definition at line 411 of file system.h.

References libMesh::System::adjoint_already_solved.

Referenced by libMesh::ImplicitSystem::adjoint_qoi_parameter_sensitivity(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::AdjointResidualErrorEstimator::estimate_error(), libMesh::ImplicitSystem::qoi_parameter_hessian(), and libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product().

  { return adjoint_already_solved;}

is_discrete_parameter()

bool libMesh::RBParametrized::is_discrete_parameter ( const std::string &  mu_name ) const

inherited

Is parameter mu_name discrete?

Definition at line 351 of file rb_parametrized.C.

References libMesh::RBParametrized::_discrete_parameter_values, and libMesh::RBParametrized::parameters_initialized.

Referenced by libMesh::RBDataSerialization::add_parameter_ranges_to_builder(), libMesh::RBEIMConstruction::print_info(), print_info(), and libMesh::RBParametrized::write_parameter_ranges_to_file().

{
  libmesh_error_msg_if(!parameters_initialized,
                       "Error: parameters not initialized in RBParametrized::is_discrete_parameter");

  return _discrete_parameter_values.count(mu_name);
}

is_initialized()

bool libMesh::System::is_initialized ( ) const

inlineinherited

Returns

true iff this system has been initialized.

Definition at line 2457 of file system.h.

References libMesh::System::_is_initialized.

Referenced by libMesh::DofMap::add_variable(), libMesh::DofMap::add_variables(), libMesh::System::get_info(), libMesh::System::init(), and libMesh::StaticCondensationDofMap::reinit().

{
  return _is_initialized;
}

is_quiet()

bool libMesh::RBConstructionBase< LinearImplicitSystem >::is_quiet ( ) const

inlineinherited

Is the system in quiet mode?

Definition at line 106 of file rb_construction_base.h.

Referenced by compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), print_info(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

  { return this->quiet_mode; }

is_rb_eval_initialized()

bool libMesh::RBConstruction::is_rb_eval_initialized

(

)

const

Returns

true if rb_eval is initialized. False, otherwise.

Definition at line 193 of file rb_construction.C.

References rb_eval.

Referenced by libMesh::TransientRBConstruction::print_info(), and print_info().

{
  return (rb_eval != nullptr);
}

is_serial_training_type()

bool libMesh::RBConstruction::is_serial_training_type ( const std::string &  RB_training_type_in )

virtual

Returns

true if RB_training_type_in is a type of training that requires a serial training set. For example, POD training generally does require a serial training set.

Definition at line 1681 of file rb_construction.C.

Referenced by set_RB_training_type().

{
  return (RB_training_type_in == "POD");
}

load_basis_function()

void libMesh::RBConstruction::load_basis_function ( unsigned int  i )

virtual

Load the i^th RB function into the RBConstruction solution vector.

Definition at line 1708 of file rb_construction.C.

References libMesh::RBEvaluation::get_basis_function(), get_rb_evaluation(), libMesh::System::solution, and libMesh::System::update().

Referenced by main().

{
  LOG_SCOPE("load_basis_function()", "RBConstruction");

  libmesh_assert_less (i, get_rb_evaluation().get_n_basis_functions());

  *solution = get_rb_evaluation().get_basis_function(i);

  this->update();
}

load_rb_solution()

void libMesh::RBConstruction::load_rb_solution ( )

virtual

Load the RB solution from the most recent solve with rb_eval into this system's solution vector.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 2256 of file rb_construction.C.

References libMesh::RBEvaluation::get_n_basis_functions(), get_rb_evaluation(), libMesh::make_range(), libMesh::RBEvaluation::RB_solution, libMesh::DenseVector< T >::size(), libMesh::System::solution, and libMesh::System::update().

Referenced by main().

{
  LOG_SCOPE("load_rb_solution()", "RBConstruction");

  solution->zero();

  libmesh_error_msg_if(get_rb_evaluation().RB_solution.size() > get_rb_evaluation().get_n_basis_functions(),
                       "ERROR: System contains " << get_rb_evaluation().get_n_basis_functions() << " basis functions."
                       << " RB_solution vector contains " << get_rb_evaluation().RB_solution.size() << " entries."
                       << " RB_solution in RBConstruction::load_rb_solution is too long!");

  for (auto i : make_range(get_rb_evaluation().RB_solution.size()))
    solution->add(get_rb_evaluation().RB_solution(i), get_rb_evaluation().get_basis_function(i));

  update();
}

load_training_set()

void libMesh::RBConstructionBase< LinearImplicitSystem >::load_training_set ( const std::map< std::string, std::vector< RBParameter >> &  new_training_set )

virtualinherited

Overwrite the training parameters with new_training_set.

This training set is assumed to contain only the samples local to this processor.

Definition at line 379 of file rb_construction_base.C.

Referenced by set_rb_construction_parameters().

{
  // Make sure we're running this on all processors at the same time
  libmesh_parallel_only(this->comm());

  // First, make sure that an initial training set has already been generated
  libmesh_error_msg_if(!_training_parameters_initialized,
                       "Error: load_training_set cannot be used to initialize parameters");

  // Make sure that the training set has the correct number of parameters
  const unsigned int n_params = get_n_params();
  libmesh_error_msg_if(new_training_set.size() > n_params,
                       "Error: new_training_set should not have more than get_n_params() parameters.");

  // Check that (new_training_set.size() == get_n_params()) is the same on all processes so that
  // we go into the same branch of the "if" statement below on all processes.
  const bool size_matches = (new_training_set.size() == n_params);
  libmesh_assert(this->comm().verify(size_matches));

  if (size_matches)
    {
      // If new_training_set stores values for all parameters, then we overwrite
      // _training_parameters with new_training_set.

      // Get the number of local and global training parameters
      _first_local_index = 0;
      _n_local_training_samples =
        cast_int<numeric_index_type>(new_training_set.begin()->second.size());
      _n_global_training_samples = _n_local_training_samples;

      if (!serial_training_set)
        {
          this->comm().sum(_n_global_training_samples);

          // Set the first/last indices.
          std::vector<numeric_index_type> local_sizes (this->n_processors(), 0);
          local_sizes[this->processor_id()] = _n_local_training_samples;
          this->comm().sum(local_sizes);

          // first_local_index is the sum of local_sizes
          // for all processor ids less than ours
          for (auto p : make_range(this->processor_id()))
            _first_local_index += local_sizes[p];
        }

      // Ensure that the parameters are the same.
      for (const auto & pr : _training_parameters)
        libmesh_error_msg_if(!new_training_set.count(pr.first),
                             "Parameters must be identical in order to overwrite dataset.");

      // Copy the values from the new_training_set to the internal training_parameters.
      _training_parameters = new_training_set;
    }
  else
    {
      // If new_training_set stores values for a subset of the parameters, then we keep the
      // length of training_parameters unchanged and overwrite the entries of the specified
      // parameters from new_training_set. Note that we repeatedly loop over new_training_set
      // to fill up the entire length of the sample_vector.
      for (auto & [param_name, sample_vector]: _training_parameters)
        {
          if (new_training_set.count(param_name))
            {
              for (const auto i : make_range(get_local_n_training_samples()))
                {
                  const unsigned int num_new_samples = libmesh_map_find(new_training_set,param_name).size();
                  libmesh_error_msg_if (num_new_samples==0, "new_training_set set should not be empty");

                  const unsigned int new_training_set_index = i % num_new_samples;
                  sample_vector[i] = libmesh_map_find(new_training_set,param_name)[new_training_set_index];
                }
            }
        }
    }
}

local_dof_indices()

void libMesh::System::local_dof_indices ( const unsigned int  var, std::set< dof_id_type > &  var_indices  ) const

inherited

Fills the std::set with the degrees of freedom on the local processor corresponding the the variable number passed in.

Definition at line 1409 of file system.C.

References libMesh::DofMap::dof_indices(), libMesh::DofMapBase::end_dof(), libMesh::DofMapBase::first_dof(), libMesh::System::get_dof_map(), libMesh::System::get_mesh(), and libMesh::libmesh_assert().

Referenced by libMesh::System::discrete_var_norm(), SystemsTest::testBlockRestrictedVarNDofs(), and libMesh::DirectSolutionTransfer::transfer().

{
  // Make sure the set is clear
  var_indices.clear();

  std::vector<dof_id_type> dof_indices;

  const dof_id_type
    first_local = this->get_dof_map().first_dof(),
    end_local   = this->get_dof_map().end_dof();

  // Begin the loop over the elements
  for (const auto & elem : this->get_mesh().active_local_element_ptr_range())
    {
      this->get_dof_map().dof_indices (elem, dof_indices, var);

      for (dof_id_type dof : dof_indices)
        //If the dof is owned by the local processor
        if (first_local <= dof && dof < end_local)
          var_indices.insert(dof);
    }

  // we may have missed assigning DOFs to nodes that we own
  // but to which we have no connected elements matching our
  // variable restriction criterion.  this will happen, for example,
  // if variable V is restricted to subdomain S.  We may not own
  // any elements which live in S, but we may own nodes which are
  // *connected* to elements which do.
  for (const auto & node : this->get_mesh().local_node_ptr_range())
    {
      libmesh_assert(node);
      this->get_dof_map().dof_indices (node, dof_indices, var);
      for (auto dof : dof_indices)
        if (first_local <= dof && dof < end_local)
          var_indices.insert(dof);
    }
}

matrices_begin() [1/2]

System::matrices_iterator libMesh::System::matrices_begin ( )

inlineinherited

Beginning of matrices container.

Definition at line 2529 of file system.h.

References libMesh::System::_matrices.

{
  return _matrices.begin();
}

matrices_begin() [2/2]

System::const_matrices_iterator libMesh::System::matrices_begin ( ) const

inlineinherited

Beginning of matrices container.

Definition at line 2535 of file system.h.

References libMesh::System::_matrices.

{
  return _matrices.begin();
}

matrices_end() [1/2]

System::matrices_iterator libMesh::System::matrices_end ( )

inlineinherited

End of matrices container.

Definition at line 2541 of file system.h.

References libMesh::System::_matrices.

{
  return _matrices.end();
}

matrices_end() [2/2]

System::const_matrices_iterator libMesh::System::matrices_end ( ) const

inlineinherited

End of matrices container.

Definition at line 2547 of file system.h.

References libMesh::System::_matrices.

{
  return _matrices.end();
}

n_active_dofs()

dof_id_type libMesh::System::n_active_dofs ( ) const

inlineinherited

Returns

The number of active degrees of freedom for this System.

Definition at line 2483 of file system.h.

References libMesh::System::n_constrained_dofs(), and libMesh::System::n_dofs().

{
  return this->n_dofs() - this->n_constrained_dofs();
}

n_components()

unsigned int libMesh::System::n_components ( ) const

inherited

Returns

The total number of scalar components in the system's variables. This will equal n_vars() in the case of all scalar-valued variables.

Definition at line 2689 of file system.C.

References libMesh::System::get_dof_map(), libMesh::System::get_mesh(), and libMesh::DofMap::n_components().

{
  return this->get_dof_map().n_components(this->get_mesh());
}

n_constrained_dofs()

dof_id_type libMesh::System::n_constrained_dofs ( ) const

inherited

Returns

The total number of constrained degrees of freedom in the system.

Definition at line 125 of file system.C.

References libMesh::System::_dof_map.

Referenced by form_functionA(), form_functionB(), form_matrixA(), libMesh::System::get_info(), libMesh::System::n_active_dofs(), libMesh::EigenSystem::solve(), and BoundaryInfoTest::testShellFaceConstraints().

{
#ifdef LIBMESH_ENABLE_CONSTRAINTS

  return _dof_map->n_constrained_dofs();

#else

  return 0;

#endif
}

n_dofs()

dof_id_type libMesh::System::n_dofs ( ) const

inherited

Returns

The number of degrees of freedom in the system

Definition at line 118 of file system.C.

References libMesh::System::_dof_map.

Referenced by libMesh::TransientRBConstruction::add_IC_to_RB_space(), libMesh::System::add_vector(), libMesh::TransientRBConstruction::allocate_data_structures(), allocate_data_structures(), libMesh::TransientRBConstruction::assemble_affine_expansion(), libMesh::ClawSystem::assemble_avg_coupling_matrices(), libMesh::ClawSystem::assemble_boundary_condition_matrices(), libMesh::AdvectionSystem::assemble_claw_rhs(), libMesh::ClawSystem::assemble_jump_coupling_matrix(), compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), compute_residual_dual_norm_slow(), libMesh::TransientRBConstruction::enrich_RB_space(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::System::get_info(), libMesh::SecondOrderUnsteadySolver::init_data(), libMesh::AdvectionSystem::init_data(), libMesh::UnsteadySolver::init_data(), libMesh::System::init_data(), libMesh::OptimizationSystem::initialize_equality_constraints_storage(), libMesh::OptimizationSystem::initialize_inequality_constraints_storage(), main(), libMesh::TransientRBConstruction::mass_matrix_scaled_matvec(), libMesh::System::n_active_dofs(), libMesh::CondensedEigenSystem::n_global_non_condensed_dofs(), libMesh::FEMSystem::numerical_jacobian(), libMesh::RBSCMConstruction::perform_SCM_greedy(), libMesh::RBEvaluation::read_in_vectors_from_multiple_files(), libMesh::TransientRBConstruction::read_riesz_representors_from_files(), read_riesz_representors_from_files(), MeshFunctionTest::read_variable_info_from_output_data(), libMesh::SecondOrderUnsteadySolver::reinit(), libMesh::UnsteadySolver::reinit(), libMesh::System::restrict_vectors(), OverlappingAlgebraicGhostingTest::run_ghosting_test(), OverlappingCouplingGhostingTest::run_sparsity_pattern_test(), libMesh::TransientRBConstruction::set_error_temporal_data(), libMesh::PetscPreconditioner< T >::set_hypre_ads_data(), libMesh::PetscPreconditioner< T >::set_hypre_ams_data(), libMesh::ClawSystem::solve_conservation_law(), SystemsTest::test100KVariables(), MeshfunctionDFEM::test_mesh_function_dfem(), MeshfunctionDFEM::test_mesh_function_dfem_grad(), MeshFunctionTest::test_p_level(), SystemsTest::testPostInitAddVector(), SystemsTest::testPostInitAddVectorTypeChange(), SystemsTest::testProjectCubeWithMeshFunction(), SystemsTest::testProjectMatrix1D(), SystemsTest::testProjectMatrix2D(), SystemsTest::testProjectMatrix3D(), train_reduced_basis_with_POD(), libMesh::MeshFunctionSolutionTransfer::transfer(), libMesh::TransientRBConstruction::truth_assembly(), truth_assembly(), libMesh::TransientRBConstruction::update_RB_initial_condition_all_N(), libMesh::TransientRBConstruction::update_RB_system_matrices(), update_RB_system_matrices(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

{
  return _dof_map->n_dofs();
}

n_linear_iterations()

unsigned int libMesh::LinearImplicitSystem::n_linear_iterations ( ) const

inlineinherited

Returns

The number of iterations taken for the most recent linear solve.

Definition at line 155 of file linear_implicit_system.h.

References libMesh::LinearImplicitSystem::_n_linear_iterations.

Referenced by compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), main(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

{ return _n_linear_iterations; }

n_local_constrained_dofs()

dof_id_type libMesh::System::n_local_constrained_dofs ( ) const

inherited

Returns

The number of constrained degrees of freedom on this processor.

Definition at line 140 of file system.C.

References libMesh::System::_dof_map.

Referenced by libMesh::System::get_info().

{
#ifdef LIBMESH_ENABLE_CONSTRAINTS

  return _dof_map->n_local_constrained_dofs();

#else

  return 0;

#endif
}

n_local_dofs()

dof_id_type libMesh::System::n_local_dofs ( ) const

inherited

Returns

The number of degrees of freedom local to this processor

Definition at line 155 of file system.C.

References libMesh::System::_dof_map.

Referenced by libMesh::TransientRBConstruction::add_IC_to_RB_space(), libMesh::System::add_vector(), libMesh::TransientRBConstruction::allocate_data_structures(), allocate_data_structures(), libMesh::TransientRBConstruction::assemble_affine_expansion(), libMesh::AdvectionSystem::assemble_claw_rhs(), libMesh::PetscDMWrapper::build_section(), compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), compute_residual_dual_norm_slow(), libMesh::TransientRBConstruction::enrich_RB_space(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::System::get_info(), libMesh::SecondOrderUnsteadySolver::init_data(), libMesh::AdvectionSystem::init_data(), libMesh::UnsteadySolver::init_data(), libMesh::System::init_data(), libMesh::OptimizationSystem::initialize_equality_constraints_storage(), libMesh::OptimizationSystem::initialize_inequality_constraints_storage(), main(), libMesh::TransientRBConstruction::mass_matrix_scaled_matvec(), libMesh::RBEvaluation::read_in_vectors_from_multiple_files(), libMesh::TransientRBConstruction::read_riesz_representors_from_files(), read_riesz_representors_from_files(), libMesh::SecondOrderUnsteadySolver::reinit(), libMesh::UnsteadySolver::reinit(), libMesh::System::restrict_vectors(), OverlappingAlgebraicGhostingTest::run_ghosting_test(), OverlappingCouplingGhostingTest::run_sparsity_pattern_test(), libMesh::TransientRBConstruction::set_error_temporal_data(), libMesh::PetscPreconditioner< T >::set_hypre_ads_data(), libMesh::PetscPreconditioner< T >::set_hypre_ams_data(), libMesh::ClawSystem::solve_conservation_law(), MeshFunctionTest::test_p_level(), train_reduced_basis_with_POD(), libMesh::TransientRBConstruction::truth_assembly(), truth_assembly(), libMesh::TransientRBConstruction::update_RB_initial_condition_all_N(), libMesh::TransientRBConstruction::update_RB_system_matrices(), update_RB_system_matrices(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

{
  return _dof_map->n_local_dofs();
}

n_matrices()

unsigned int libMesh::System::n_matrices ( ) const

inlineinherited

Returns

The number of matrices handled by this system. This is the size of the _matrices map

Definition at line 2638 of file system.h.

References libMesh::System::_matrices.

Referenced by libMesh::ImplicitSystem::add_matrices(), and libMesh::System::get_info().

{
  return cast_int<unsigned int>(_matrices.size());
}

n_objects() [1/2]

static unsigned int libMesh::ReferenceCounter::n_objects ( )

inlinestaticinherited

Prints the number of outstanding (created, but not yet destroyed) objects.

Definition at line 85 of file reference_counter.h.

References libMesh::ReferenceCounter::_n_objects.

Referenced by libMesh::LibMeshInit::~LibMeshInit().

  { return _n_objects; }

n_objects() [2/2]

static unsigned int libMesh::ReferenceCounter::n_objects ( )

inlinestaticinherited

Prints the number of outstanding (created, but not yet destroyed) objects.

Definition at line 85 of file reference_counter.h.

References libMesh::ReferenceCounter::_n_objects.

Referenced by libMesh::LibMeshInit::~LibMeshInit().

  { return _n_objects; }

n_processors()

processor_id_type libMesh::ParallelObject::n_processors ( ) const

inlineinherited

Returns

The number of processors in the group.

Definition at line 103 of file parallel_object.h.

References libMesh::ParallelObject::_communicator, libMesh::libmesh_assert(), and TIMPI::Communicator::size().

Referenced by libMesh::Partitioner::_find_global_index_by_pid_map(), libMesh::BoundaryInfo::_find_id_maps(), libMesh::DofMap::add_constraints_to_send_list(), libMesh::PetscDMWrapper::add_dofs_to_section(), libMesh::DistributedMesh::add_elem(), libMesh::DofMap::add_neighbors_to_send_list(), libMesh::DistributedMesh::add_node(), libMesh::System::add_vector(), libMesh::LaplaceMeshSmoother::allgather_graph(), libMesh::DofMap::allgather_recursive_constraints(), libMesh::FEMSystem::assembly(), libMesh::Nemesis_IO::assert_symmetric_cmaps(), libMesh::Partitioner::assign_partitioning(), libMesh::AztecLinearSolver< T >::AztecLinearSolver(), libMesh::Partitioner::build_graph(), libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::DistributedMesh::clear(), libMesh::DistributedMesh::clear_elems(), libMesh::Nemesis_IO_Helper::compute_border_node_ids(), libMesh::Nemesis_IO_Helper::construct_nemesis_filename(), libMesh::UnstructuredMesh::copy_nodes_and_elements(), libMesh::Nemesis_IO::copy_scalar_solution(), libMesh::ExodusII_IO::copy_scalar_solution(), libMesh::UnstructuredMesh::create_pid_mesh(), libMesh::MeshTools::create_processor_bounding_box(), libMesh::DofMap::distribute_dofs(), libMesh::DofMap::distribute_scalar_dofs(), libMesh::DistributedMesh::DistributedMesh(), libMesh::EnsightIO::EnsightIO(), libMesh::RBEIMEvaluation::gather_bfs(), libMesh::MeshBase::get_info(), libMesh::StaticCondensation::init(), libMesh::SystemSubsetBySubdomain::init(), libMesh::PetscDMWrapper::init_petscdm(), libMesh::Nemesis_IO_Helper::initialize(), libMesh::ExodusII_IO_Helper::initialize(), libMesh::DistributedMesh::insert_elem(), libMesh::NumericVector< Number >::is_effectively_ghosted(), libMesh::NumericVector< Number >::is_effectively_serial(), libMesh::MeshTools::libmesh_assert_contiguous_dof_ids(), libMesh::MeshTools::libmesh_assert_parallel_consistent_new_node_procids(), libMesh::MeshTools::libmesh_assert_parallel_consistent_procids< Elem >(), libMesh::MeshTools::libmesh_assert_parallel_consistent_procids< Node >(), libMesh::MeshTools::libmesh_assert_topology_consistent_procids< Node >(), libMesh::MeshTools::libmesh_assert_valid_boundary_ids(), libMesh::MeshTools::libmesh_assert_valid_dof_ids(), libMesh::MeshTools::libmesh_assert_valid_neighbors(), libMesh::MeshTools::libmesh_assert_valid_refinement_flags(), libMesh::DofMap::local_variable_indices(), libMesh::MeshRefinement::make_coarsening_compatible(), libMesh::MeshBase::n_active_elem_on_proc(), libMesh::DofMap::n_dofs_per_processor(), libMesh::MeshBase::n_elem_on_proc(), libMesh::MeshBase::n_nodes_on_proc(), libMesh::RBEIMEvaluation::node_gather_bfs(), libMesh::Partitioner::partition(), libMesh::MeshBase::partition(), libMesh::Partitioner::partition_unpartitioned_elements(), libMesh::System::point_gradient(), libMesh::System::point_hessian(), libMesh::System::point_value(), libMesh::DofMap::prepare_send_list(), libMesh::MeshBase::print_constraint_rows(), libMesh::DofMap::print_dof_constraints(), libMesh::NameBasedIO::read(), libMesh::Nemesis_IO::read(), libMesh::CheckpointIO::read(), libMesh::CheckpointIO::read_connectivity(), libMesh::XdrIO::read_header(), libMesh::CheckpointIO::read_nodes(), libMesh::System::read_parallel_data(), libMesh::System::read_SCALAR_dofs(), libMesh::System::read_serialized_blocked_dof_objects(), libMesh::System::read_serialized_vector(), libMesh::DistributedMesh::renumber_dof_objects(), libMesh::Partitioner::repartition(), OverlappingFunctorTest::run_partitioner_test(), libMesh::DofMap::scatter_constraints(), libMesh::DistributedMesh::set_next_unique_id(), libMesh::DofMap::set_nonlocal_dof_objects(), libMesh::PetscDMWrapper::set_point_range_in_section(), WriteVecAndScalar::setupTests(), libMesh::RBEIMEvaluation::side_gather_bfs(), DistributedMeshTest::testRemoteElemError(), CheckpointIOTest::testSplitter(), libMesh::MeshRefinement::uniformly_coarsen(), libMesh::DistributedMesh::update_parallel_id_counts(), libMesh::GMVIO::write_binary(), libMesh::GMVIO::write_discontinuous_gmv(), libMesh::ExodusII_IO_Helper::write_nodal_coordinates(), libMesh::VTKIO::write_nodal_data(), libMesh::ExodusII_IO::write_nodal_data(), libMesh::System::write_parallel_data(), libMesh::System::write_SCALAR_dofs(), libMesh::XdrIO::write_serialized_bcs_helper(), libMesh::System::write_serialized_blocked_dof_objects(), libMesh::XdrIO::write_serialized_connectivity(), libMesh::XdrIO::write_serialized_nodes(), and libMesh::XdrIO::write_serialized_nodesets().

  {
    processor_id_type returnval =
      cast_int<processor_id_type>(_communicator.size());
    libmesh_assert(returnval); // We never have an empty comm
    return returnval;
  }

n_qois()

unsigned int libMesh::System::n_qois ( ) const

inlineinherited

Number of currently active quantities of interest.

Definition at line 2562 of file system.h.

References libMesh::System::_qoi, and libMesh::System::_qoi_error_estimates.

Referenced by libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::AdaptiveTimeSolver::adjoint_advance_timestep(), libMesh::UnsteadySolver::adjoint_advance_timestep(), libMesh::ImplicitSystem::adjoint_qoi_parameter_sensitivity(), libMesh::TwostepTimeSolver::adjoint_solve(), libMesh::ImplicitSystem::adjoint_solve(), libMesh::SensitivityData::allocate_data(), libMesh::SensitivityData::allocate_hessian_data(), libMesh::ExplicitSystem::assemble_qoi(), libMesh::FEMSystem::assemble_qoi(), libMesh::ExplicitSystem::assemble_qoi_derivative(), libMesh::FEMSystem::assemble_qoi_derivative(), libMesh::DiffContext::DiffContext(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::AdjointResidualErrorEstimator::estimate_error(), libMesh::FileSolutionHistory::FileSolutionHistory(), libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity(), libMesh::UnsteadySolver::init_adjoints(), libMesh::TimeSolver::init_adjoints(), libMesh::System::init_qois(), libMesh::Euler2Solver::integrate_adjoint_refinement_error_estimate(), libMesh::TwostepTimeSolver::integrate_adjoint_refinement_error_estimate(), libMesh::EulerSolver::integrate_adjoint_refinement_error_estimate(), libMesh::Euler2Solver::integrate_qoi_timestep(), libMesh::TwostepTimeSolver::integrate_qoi_timestep(), libMesh::EulerSolver::integrate_qoi_timestep(), main(), libMesh::FEMContext::pre_fe_reinit(), libMesh::ImplicitSystem::qoi_parameter_hessian(), libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product(), libMesh::FileSolutionHistory::retrieve(), libMesh::QoISet::size(), libMesh::UnsteadySolver::UnsteadySolver(), and libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve().

{
  libmesh_assert_equal_to(this->_qoi.size(), this->_qoi_error_estimates.size());

  return cast_int<unsigned int>(this->_qoi.size());
}

n_variable_groups()

unsigned int libMesh::System::n_variable_groups ( ) const

inherited

Returns

The number of VariableGroup variable groups in the system

Definition at line 2694 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::n_variable_groups().

Referenced by libMesh::FEMSystem::assembly(), and libMesh::System::get_info().

{
  return this->get_dof_map().n_variable_groups();
}

n_vars()

unsigned int libMesh::System::n_vars ( ) const

inherited

Returns

The number of variables in the system

Definition at line 2669 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::n_vars().

Referenced by libMesh::ExactSolution::_compute_error(), libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::PetscDMWrapper::add_dofs_helper(), libMesh::DiffContext::add_localized_vector(), add_scaled_matrix_and_vector(), libMesh::TwostepTimeSolver::adjoint_solve(), libMesh::FEMContext::attach_quadrature_rules(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::PetscDMWrapper::build_section(), libMesh::System::calculate_norm(), compute_stresses(), LinearElasticityWithContact::compute_stresses(), LinearElasticity::compute_stresses(), LargeDeformationElasticity::compute_stresses(), libMesh::DGFEMContext::DGFEMContext(), libMesh::DiffContext::DiffContext(), libMesh::JumpErrorEstimator::estimate_error(), libMesh::AdjointResidualErrorEstimator::estimate_error(), libMesh::ExactErrorEstimator::estimate_error(), libMesh::ErrorEstimator::estimate_errors(), libMesh::ExactSolution::ExactSolution(), libMesh::FEMContext::find_hardest_fe_type(), libMesh::EquationSystems::find_variable_numbers(), libMesh::FEMSystem::init_context(), libMesh::RBEIMConstruction::init_context(), libMesh::FEMContext::init_internal_data(), libMesh::PetscDMWrapper::init_petscdm(), libMesh::DifferentiablePhysics::init_physics(), AssemblyA0::interior_assembly(), AssemblyA1::interior_assembly(), AssemblyA2::interior_assembly(), InnerProductAssembly::interior_assembly(), main(), libMesh::WeightedPatchRecoveryErrorEstimator::EstimateError::operator()(), libMesh::SmoothnessEstimator::EstimateSmoothness::operator()(), libMesh::PatchRecoveryErrorEstimator::EstimateError::operator()(), output_norms(), libMesh::petsc_auto_fieldsplit(), libMesh::FEMContext::pre_fe_reinit(), libMesh::InterMeshProjection::project_system_vectors(), libMesh::System::re_update(), libMesh::System::read_parallel_data(), libMesh::System::read_serialized_blocked_dof_objects(), libMesh::System::read_serialized_vector(), libMesh::System::read_serialized_vectors(), libMesh::System::reinit_mesh(), libMesh::HPCoarsenTest::select_refinement(), libMesh::PetscPreconditioner< T >::set_petsc_aux_data(), libMesh::PetscDMWrapper::set_point_range_in_section(), libMesh::SystemSubsetBySubdomain::set_var_nums(), OverlappingTestBase::setup_coupling_matrix(), SystemsTest::testDofCouplingWithVarGroups(), SlitMeshRefinedSystemTest::testRestart(), SlitMeshRefinedSystemTest::testSystem(), libMesh::System::write_header(), libMesh::System::write_parallel_data(), libMesh::System::write_serialized_blocked_dof_objects(), libMesh::System::write_serialized_vector(), libMesh::System::write_serialized_vectors(), and libMesh::System::zero_variable().

{
  return this->get_dof_map().n_vars();
}

n_vectors()

unsigned int libMesh::System::n_vectors ( ) const

inlineinherited

Returns

The number of vectors (in addition to the solution) handled by this system This is the size of the _vectors map

Definition at line 2499 of file system.h.

References libMesh::System::_vectors.

Referenced by libMesh::ExplicitSystem::add_system_rhs(), libMesh::System::compare(), libMesh::System::get_info(), main(), libMesh::InterMeshProjection::project_system_vectors(), and libMesh::System::write_header().

{
  return cast_int<unsigned int>(_vectors.size());
}

name()

const std::string & libMesh::System::name ( ) const

inlineinherited

Returns

The system name.

Definition at line 2385 of file system.h.

References libMesh::System::_sys_name.

Referenced by libMesh::System::compare(), compute_jacobian(), compute_residual(), DMlibMeshSetUpName_Private(), enrich_basis_from_rhs_terms(), libMesh::ExactErrorEstimator::estimate_error(), libMesh::ExactSolution::ExactSolution(), libMesh::ExactErrorEstimator::find_squared_element_error(), libMesh::System::get_info(), libMesh::NewtonSolver::init(), libMesh::StaticCondensation::init(), libMesh::TimeSolver::init_data(), libMesh::System::init_matrices(), main(), output_norms(), libMesh::petsc_auto_fieldsplit(), libMesh::System::prefix(), libMesh::RBSCMConstruction::print_info(), libMesh::ClawSystem::print_info(), libMesh::RBEIMConstruction::print_info(), print_info(), libMesh::TimeSolver::reinit(), libMesh::PetscDiffSolver::setup_petsc_data(), libMesh::FrequencySystem::solve(), truth_solve(), libMesh::System::user_assembly(), libMesh::System::user_constrain(), libMesh::System::user_initialization(), libMesh::System::user_QOI(), libMesh::System::user_QOI_derivative(), libMesh::System::write_header(), libMesh::System::write_parallel_data(), and libMesh::System::write_serialized_data().

{
  return _sys_name;
}

number()

unsigned int libMesh::System::number ( ) const

inlineinherited

Returns

The system number.

Definition at line 2393 of file system.h.

References libMesh::System::_sys_number.

Referenced by libMesh::ExactSolution::_compute_error(), libMesh::PetscDMWrapper::add_dofs_helper(), assemble_matrix_and_rhs(), assemble_shell(), libMesh::VariationalSmootherSystem::assembly(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::VariationalSmootherConstraint::constrain_node_to_line(), libMesh::VariationalSmootherConstraint::constrain_node_to_plane(), libMesh::Nemesis_IO::copy_elemental_solution(), libMesh::ExodusII_IO::copy_elemental_solution(), libMesh::Nemesis_IO::copy_nodal_solution(), libMesh::ExodusII_IO::copy_nodal_solution(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::ExactErrorEstimator::find_squared_element_error(), libMesh::EquationSystems::find_variable_numbers(), libMesh::VariationalSmootherConstraint::fix_node(), libMesh::System::get_info(), libMesh::VariationalSmootherSystem::init_data(), main(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::SortAndCopy::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectVertices::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectEdges::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectSides::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectInteriors::operator()(), libMesh::System::read_parallel_data(), libMesh::System::read_serialized_blocked_dof_objects(), libMesh::StaticCondensationDofMap::reinit(), LinearElasticityWithContact::residual_and_jacobian(), SolidSystem::save_initial_mesh(), libMesh::HPCoarsenTest::select_refinement(), libMesh::PetscPreconditioner< T >::set_hypre_ads_data(), libMesh::PetscPreconditioner< T >::set_hypre_ams_data(), libMesh::PetscDMWrapper::set_point_range_in_section(), MeshInputTest::testCopyElementVectorImpl(), libMesh::MeshFunctionSolutionTransfer::transfer(), libMesh::MeshfreeSolutionTransfer::transfer(), libMesh::BoundaryVolumeSolutionTransfer::transfer_boundary_volume(), libMesh::BoundaryVolumeSolutionTransfer::transfer_volume_boundary(), libMesh::DTKAdapter::update_variable_values(), libMesh::System::write_parallel_data(), libMesh::System::write_serialized_blocked_dof_objects(), and libMesh::System::zero_variable().

{
  return _sys_number;
}

operator=() [1/2]

RBConstruction& libMesh::RBConstruction::operator= ( const RBConstruction &  )

delete

operator=() [2/2]

RBConstruction& libMesh::RBConstruction::operator= ( RBConstruction &&  )

delete

point_gradient() [1/4]

Gradient libMesh::System::point_gradient ( unsigned int  var, const Point &  p, const bool  insist_on_success = true, const NumericVector< Number > *  sol = nullptr  ) const

inherited

Returns

The gradient of the solution variable var at the physical point p in the mesh, similarly to point_value.

Definition at line 2343 of file system.C.

References libMesh::Variable::active_subdomains(), TIMPI::Communicator::broadcast(), libMesh::ParallelObject::comm(), libMesh::PointLocatorBase::enable_out_of_mesh_mode(), libMesh::System::get_dof_map(), libMesh::System::get_mesh(), libMesh::libmesh_assert(), mesh, TIMPI::Communicator::min(), libMesh::ParallelObject::n_processors(), libMesh::ParallelObject::processor_id(), libMesh::DofObject::processor_id(), and libMesh::System::variable().

Referenced by line_print(), and libMesh::System::point_gradient().

{
  // This function must be called on every processor; there's no
  // telling where in the partition p falls.
  parallel_object_only();

  // And every processor had better agree about which point we're
  // looking for
#ifndef NDEBUG
  libmesh_assert(this->comm().verify(p(0)));
#if LIBMESH_DIM > 1
  libmesh_assert(this->comm().verify(p(1)));
#endif
#if LIBMESH_DIM > 2
  libmesh_assert(this->comm().verify(p(2)));
#endif
#endif // NDEBUG

  // Get a reference to the mesh object associated with the system object that calls this function
  const MeshBase & mesh = this->get_mesh();

  // Use an existing PointLocator or create a new one
  std::unique_ptr<PointLocatorBase> locator_ptr = mesh.sub_point_locator();
  PointLocatorBase & locator = *locator_ptr;

  if (!insist_on_success || !mesh.is_serial())
    locator.enable_out_of_mesh_mode();

  // Get a pointer to an element that contains p and allows us to
  // evaluate var
  const std::set<subdomain_id_type> & raw_subdomains =
    this->variable(var).active_subdomains();
  const std::set<subdomain_id_type> * implicit_subdomains =
    raw_subdomains.empty() ? nullptr : &raw_subdomains;
  const Elem * e = locator(p, implicit_subdomains);

  Gradient grad_u;

  if (e && this->get_dof_map().is_evaluable(*e, var))
    grad_u = point_gradient(var, p, *e, sol);

  // If I have an element containing p, then let's let everyone know
  processor_id_type lowest_owner =
    (e && (e->processor_id() == this->processor_id())) ?
    this->processor_id() : this->n_processors();
  this->comm().min(lowest_owner);

  // Everybody should get their value from a processor that was able
  // to compute it.
  // If nobody admits owning the point, we may have a problem.
  if (lowest_owner != this->n_processors())
    this->comm().broadcast(grad_u, lowest_owner);
  else
    libmesh_assert(!insist_on_success);

  return grad_u;
}

point_gradient() [2/4]

Gradient libMesh::System::point_gradient ( unsigned int  var, const Point &  p, const Elem &  e, const NumericVector< Number > *  sol = nullptr  ) const

inherited

Returns

The gradient of the solution variable var at the physical point p in local Elem e in the mesh, similarly to point_value.

Definition at line 2405 of file system.C.

References libMesh::FEInterface::compute_data(), libMesh::Elem::contains_point(), libMesh::System::current_local_solution, dim, libMesh::Elem::dim(), libMesh::DofMap::dof_indices(), libMesh::FEComputeData::dshape, libMesh::FEComputeData::enable_derivative(), libMesh::System::get_dof_map(), libMesh::System::get_equation_systems(), libMesh::FEMap::inverse_map(), libMesh::DofMap::is_evaluable(), libMesh::libmesh_assert(), libMesh::FEComputeData::local_transform, and libMesh::DofMap::variable_type().

{
  // Ensuring that the given point is really in the element is an
  // expensive assert, but as long as debugging is turned on we might
  // as well try to catch a particularly nasty potential error
  libmesh_assert (e.contains_point(p));

  if (!sol)
    sol = this->current_local_solution.get();

  // Get the dof map to get the proper indices for our computation
  const DofMap & dof_map = this->get_dof_map();

  // write the element dimension into a separate variable.
  const unsigned int dim = e.dim();

  // Make sure we can evaluate on this element.
  libmesh_assert (dof_map.is_evaluable(e, var));

  // Need dof_indices for phi[i][j]
  std::vector<dof_id_type> dof_indices;

  // Fill in the dof_indices for our element
  dof_map.dof_indices (&e, dof_indices, var);

  // Get the no of dofs associated with this point
  const unsigned int num_dofs = cast_int<unsigned int>
    (dof_indices.size());

  FEType fe_type = dof_map.variable_type(var);

  // Map the physical co-ordinates to the master co-ordinates
  Point coor = FEMap::inverse_map(dim, &e, p);

  // get the shape function value via the FEInterface to also handle the case
  // of infinite elements correctly, the shape function is not fe->phi().
  FEComputeData fe_data(this->get_equation_systems(), coor);
  fe_data.enable_derivative();
  FEInterface::compute_data(dim, fe_type, &e, fe_data);

  // Get ready to accumulate a gradient
  Gradient grad_u;

  for (unsigned int l=0; l<num_dofs; l++)
    {
      // Chartesian coordinates have always LIBMESH_DIM entries,
      // local coordinates have as many coordinates as the element has.
      for (std::size_t v=0; v<dim; v++)
        for (std::size_t xyz=0; xyz<LIBMESH_DIM; xyz++)
          {
            // FIXME: this needs better syntax: It is matrix-vector multiplication.
            grad_u(xyz) += fe_data.local_transform[v][xyz]
              * fe_data.dshape[l](v)
              * (*sol)(dof_indices[l]);
          }
    }

  return grad_u;
}

point_gradient() [3/4]

Gradient libMesh::System::point_gradient ( unsigned int  var, const Point &  p, const Elem *  e  ) const

inherited

Calls the version of point_gradient() which takes a reference.

This function exists only to prevent people from calling the version of point_gradient() that has a boolean third argument, which would result in unnecessary PointLocator calls.

Definition at line 2470 of file system.C.

References libMesh::libmesh_assert(), and libMesh::System::point_gradient().

{
  libmesh_assert(e);
  return this->point_gradient(var, p, *e);
}

point_gradient() [4/4]

Gradient libMesh::System::point_gradient ( unsigned int  var, const Point &  p, const NumericVector< Number > *  sol  ) const

inherited

Calls the parallel version of point_gradient().

This function exists only to prevent people from accidentally calling the version of point_gradient() that has a boolean third argument, which would result in incorrect output.

Definition at line 2478 of file system.C.

References libMesh::System::point_gradient().

{
  return this->point_gradient(var, p, true, sol);
}

point_hessian() [1/4]

Tensor libMesh::System::point_hessian ( unsigned int  var, const Point &  p, const bool  insist_on_success = true, const NumericVector< Number > *  sol = nullptr  ) const

inherited

Returns

The second derivative tensor of the solution variable var at the physical point p in the mesh, similarly to point_value.

Definition at line 2487 of file system.C.

References libMesh::Variable::active_subdomains(), TIMPI::Communicator::broadcast(), libMesh::ParallelObject::comm(), libMesh::PointLocatorBase::enable_out_of_mesh_mode(), libMesh::System::get_dof_map(), libMesh::System::get_mesh(), libMesh::libmesh_assert(), mesh, TIMPI::Communicator::min(), libMesh::ParallelObject::n_processors(), libMesh::ParallelObject::processor_id(), libMesh::DofObject::processor_id(), and libMesh::System::variable().

Referenced by libMesh::System::point_hessian().

{
  // This function must be called on every processor; there's no
  // telling where in the partition p falls.
  parallel_object_only();

  // And every processor had better agree about which point we're
  // looking for
#ifndef NDEBUG
  libmesh_assert(this->comm().verify(p(0)));
#if LIBMESH_DIM > 1
  libmesh_assert(this->comm().verify(p(1)));
#endif
#if LIBMESH_DIM > 2
  libmesh_assert(this->comm().verify(p(2)));
#endif
#endif // NDEBUG

  // Get a reference to the mesh object associated with the system object that calls this function
  const MeshBase & mesh = this->get_mesh();

  // Use an existing PointLocator or create a new one
  std::unique_ptr<PointLocatorBase> locator_ptr = mesh.sub_point_locator();
  PointLocatorBase & locator = *locator_ptr;

  if (!insist_on_success || !mesh.is_serial())
    locator.enable_out_of_mesh_mode();

  // Get a pointer to an element that contains p and allows us to
  // evaluate var
  const std::set<subdomain_id_type> & raw_subdomains =
    this->variable(var).active_subdomains();
  const std::set<subdomain_id_type> * implicit_subdomains =
    raw_subdomains.empty() ? nullptr : &raw_subdomains;
  const Elem * e = locator(p, implicit_subdomains);

  Tensor hess_u;

  if (e && this->get_dof_map().is_evaluable(*e, var))
    hess_u = point_hessian(var, p, *e, sol);

  // If I have an element containing p, then let's let everyone know
  processor_id_type lowest_owner =
    (e && (e->processor_id() == this->processor_id())) ?
    this->processor_id() : this->n_processors();
  this->comm().min(lowest_owner);

  // Everybody should get their value from a processor that was able
  // to compute it.
  // If nobody admits owning the point, we may have a problem.
  if (lowest_owner != this->n_processors())
    this->comm().broadcast(hess_u, lowest_owner);
  else
    libmesh_assert(!insist_on_success);

  return hess_u;
}

point_hessian() [2/4]

Tensor libMesh::System::point_hessian ( unsigned int  var, const Point &  p, const Elem &  e, const NumericVector< Number > *  sol = nullptr  ) const

inherited

Returns

The second derivative tensor of the solution variable var at the physical point p in local Elem e in the mesh, similarly to point_value.

Definition at line 2548 of file system.C.

References libMesh::TypeTensor< T >::add_scaled(), libMesh::FEGenericBase< OutputType >::build(), libMesh::Elem::contains_point(), libMesh::System::current_local_solution, libMesh::Elem::dim(), libMesh::DofMap::dof_indices(), libMesh::System::get_dof_map(), libMesh::Elem::infinite(), libMesh::FEMap::inverse_map(), libMesh::DofMap::is_evaluable(), libMesh::libmesh_assert(), and libMesh::DofMap::variable_type().

{
  // Ensuring that the given point is really in the element is an
  // expensive assert, but as long as debugging is turned on we might
  // as well try to catch a particularly nasty potential error
  libmesh_assert (e.contains_point(p));

  if (!sol)
    sol = this->current_local_solution.get();

  if (e.infinite())
    libmesh_not_implemented();

  // Get the dof map to get the proper indices for our computation
  const DofMap & dof_map = this->get_dof_map();

  // Make sure we can evaluate on this element.
  libmesh_assert (dof_map.is_evaluable(e, var));

  // Need dof_indices for phi[i][j]
  std::vector<dof_id_type> dof_indices;

  // Fill in the dof_indices for our element
  dof_map.dof_indices (&e, dof_indices, var);

  // Get the no of dofs associated with this point
  const unsigned int num_dofs = cast_int<unsigned int>
    (dof_indices.size());

  FEType fe_type = dof_map.variable_type(var);

  // Build a FE again so we can calculate u(p)
  std::unique_ptr<FEBase> fe (FEBase::build(e.dim(), fe_type));

  // Map the physical co-ordinates to the master co-ordinates
  // Build a vector of point co-ordinates to send to reinit
  std::vector<Point> coor(1, FEMap::inverse_map(e.dim(), &e, p));

  // Get the values of the shape function derivatives
  const std::vector<std::vector<RealTensor>> &  d2phi = fe->get_d2phi();

  // Reinitialize the element and compute the shape function values at coor
  fe->reinit (&e, &coor);

  // Get ready to accumulate a hessian
  Tensor hess_u;

  for (unsigned int l=0; l<num_dofs; l++)
    {
      hess_u.add_scaled (d2phi[l][0], (*sol)(dof_indices[l]));
    }

  return hess_u;
}

point_hessian() [3/4]

Tensor libMesh::System::point_hessian ( unsigned int  var, const Point &  p, const Elem *  e  ) const

inherited

Calls the version of point_hessian() which takes a reference.

This function exists only to prevent people from calling the version of point_hessian() that has a boolean third argument, which would result in unnecessary PointLocator calls.

Definition at line 2608 of file system.C.

References libMesh::libmesh_assert(), and libMesh::System::point_hessian().

{
  libmesh_assert(e);
  return this->point_hessian(var, p, *e);
}

point_hessian() [4/4]

Tensor libMesh::System::point_hessian ( unsigned int  var, const Point &  p, const NumericVector< Number > *  sol  ) const

inherited

Calls the parallel version of point_hessian().

This function exists only to prevent people from accidentally calling the version of point_hessian() that has a boolean third argument, which would result in incorrect output.

Definition at line 2616 of file system.C.

References libMesh::System::point_hessian().

{
  return this->point_hessian(var, p, true, sol);
}

point_value() [1/4]

Number libMesh::System::point_value ( unsigned int  var, const Point &  p, const bool  insist_on_success = true, const NumericVector< Number > *  sol = nullptr  ) const

inherited

Returns

The value of the solution variable var at the physical point p in the mesh, without knowing a priori which element contains p, using the degree of freedom coefficients in sol (or in current_local_solution if sol is left null).

Note

This function uses MeshBase::sub_point_locator(); users may or may not want to call MeshBase::clear_point_locator() afterward. Also, point_locator() is expensive (N log N for initial construction, log N for evaluations). Avoid using this function in any context where you are already looping over elements.

Because the element containing p may lie on any processor, this function is parallel-only.

By default this method expects the point to reside inside the domain and will abort if no element can be found which contains p. The optional parameter insist_on_success can be set to false to allow the method to return 0 when the point is not located.

Definition at line 2214 of file system.C.

References libMesh::Variable::active_subdomains(), TIMPI::Communicator::broadcast(), libMesh::ParallelObject::comm(), libMesh::PointLocatorBase::enable_out_of_mesh_mode(), libMesh::System::get_dof_map(), libMesh::System::get_mesh(), libMesh::libmesh_assert(), mesh, TIMPI::Communicator::min(), libMesh::ParallelObject::n_processors(), libMesh::ParallelObject::processor_id(), libMesh::DofObject::processor_id(), and libMesh::System::variable().

Referenced by line_print(), main(), libMesh::System::point_value(), MeshInputTest::testCopyElementSolutionImpl(), MeshInputTest::testCopyElementVectorImpl(), MeshInputTest::testCopyNodalSolutionImpl(), DefaultCouplingTest::testCoupling(), PointNeighborCouplingTest::testCoupling(), MeshInputTest::testExodusWriteElementDataFromDiscontinuousNodalData(), PeriodicBCTest::testPeriodicBC(), SystemsTest::testProjectCubeWithMeshFunction(), EquationSystemsTest::testRepartitionThenReinit(), DisjointNeighborTest::testTempJump(), and DisjointNeighborTest::testTempJumpRefine().

{
  // This function must be called on every processor; there's no
  // telling where in the partition p falls.
  parallel_object_only();

  // And every processor had better agree about which point we're
  // looking for
#ifndef NDEBUG
  libmesh_assert(this->comm().verify(p(0)));
#if LIBMESH_DIM > 1
  libmesh_assert(this->comm().verify(p(1)));
#endif
#if LIBMESH_DIM > 2
  libmesh_assert(this->comm().verify(p(2)));
#endif
#endif // NDEBUG

  // Get a reference to the mesh object associated with the system object that calls this function
  const MeshBase & mesh = this->get_mesh();

  // Use an existing PointLocator or create a new one
  std::unique_ptr<PointLocatorBase> locator_ptr = mesh.sub_point_locator();
  PointLocatorBase & locator = *locator_ptr;

  if (!insist_on_success || !mesh.is_serial())
    locator.enable_out_of_mesh_mode();

  // Get a pointer to an element that contains p and allows us to
  // evaluate var
  const std::set<subdomain_id_type> & raw_subdomains =
    this->variable(var).active_subdomains();
  const std::set<subdomain_id_type> * implicit_subdomains =
    raw_subdomains.empty() ? nullptr : &raw_subdomains;
  const Elem * e = locator(p, implicit_subdomains);

  Number u = 0;

  if (e && this->get_dof_map().is_evaluable(*e, var))
    u = point_value(var, p, *e, sol);

  // If I have an element containing p, then let's let everyone know
  processor_id_type lowest_owner =
    (e && (e->processor_id() == this->processor_id())) ?
    this->processor_id() : this->n_processors();
  this->comm().min(lowest_owner);

  // Everybody should get their value from a processor that was able
  // to compute it.
  // If nobody admits owning the point, we have a problem.
  if (lowest_owner != this->n_processors())
    this->comm().broadcast(u, lowest_owner);
  else
    libmesh_assert(!insist_on_success);

  return u;
}

point_value() [2/4]

Number libMesh::System::point_value ( unsigned int  var, const Point &  p, const Elem &  e, const NumericVector< Number > *  sol = nullptr  ) const

inherited

Returns

The value of the solution variable var at the physical point p contained in local Elem e, using the degree of freedom coefficients in sol (or in current_local_solution if sol is left null).

This version of point_value can be run in serial, but assumes e is in the local mesh partition or is algebraically ghosted.

Definition at line 2275 of file system.C.

References libMesh::FEInterface::compute_data(), libMesh::Elem::contains_point(), libMesh::System::current_local_solution, libMesh::Elem::dim(), libMesh::DofMap::dof_indices(), libMesh::System::get_dof_map(), libMesh::System::get_equation_systems(), libMesh::FEMap::inverse_map(), libMesh::DofMap::is_evaluable(), libMesh::libmesh_assert(), and libMesh::DofMap::variable_type().

{
  // Ensuring that the given point is really in the element is an
  // expensive assert, but as long as debugging is turned on we might
  // as well try to catch a particularly nasty potential error
  libmesh_assert (e.contains_point(p));

  if (!sol)
    sol = this->current_local_solution.get();

  // Get the dof map to get the proper indices for our computation
  const DofMap & dof_map = this->get_dof_map();

  // Make sure we can evaluate on this element.
  libmesh_assert (dof_map.is_evaluable(e, var));

  // Need dof_indices for phi[i][j]
  std::vector<dof_id_type> dof_indices;

  // Fill in the dof_indices for our element
  dof_map.dof_indices (&e, dof_indices, var);

  // Get the no of dofs associated with this point
  const unsigned int num_dofs = cast_int<unsigned int>
    (dof_indices.size());

  FEType fe_type = dof_map.variable_type(var);

  // Map the physical co-ordinates to the master co-ordinates
  Point coor = FEMap::inverse_map(e.dim(), &e, p);

  // get the shape function value via the FEInterface to also handle the case
  // of infinite elements correctly, the shape function is not fe->phi().
  FEComputeData fe_data(this->get_equation_systems(), coor);
  FEInterface::compute_data(e.dim(), fe_type, &e, fe_data);

  // Get ready to accumulate a value
  Number u = 0;

  for (unsigned int l=0; l<num_dofs; l++)
    {
      u += fe_data.shape[l] * (*sol)(dof_indices[l]);
    }

  return u;
}

point_value() [3/4]

Number libMesh::System::point_value ( unsigned int  var, const Point &  p, const Elem *  e  ) const

inherited

Calls the version of point_value() which takes a reference.

This function exists only to prevent people from calling the version of point_value() that has a boolean third argument, which would result in unnecessary PointLocator calls.

Definition at line 2327 of file system.C.

References libMesh::libmesh_assert(), and libMesh::System::point_value().

{
  libmesh_assert(e);
  return this->point_value(var, p, *e);
}

point_value() [4/4]

Number libMesh::System::point_value ( unsigned int  var, const Point &  p, const NumericVector< Number > *  sol  ) const

inherited

Calls the parallel version of point_value().

This function exists only to prevent people from accidentally calling the version of point_value() that has a boolean third argument, which would result in incorrect output.

Definition at line 2335 of file system.C.

References libMesh::System::point_value().

{
  return this->point_value(var, p, true, sol);
}

post_process_elem_matrix_and_vector()

virtual void libMesh::RBConstruction::post_process_elem_matrix_and_vector ( DGFEMContext &  )

inlineprotectedvirtual

This function is called from add_scaled_matrix_and_vector() before each element matrix and vector are assembled into their global counterparts.

By default it is a no-op, but it could be used to apply any user-defined transformations immediately prior to assembly. We use DGFEMContext since it allows for both DG and continuous Galerkin formulations.

Definition at line 724 of file rb_construction.h.

Referenced by add_scaled_matrix_and_vector().

{}

post_process_truth_solution()

virtual void libMesh::RBConstruction::post_process_truth_solution ( )

inlineprotectedvirtual

Similarly, provide an opportunity to post-process the truth solution after the solve is complete.

By default this is a no-op, but it could be used to apply any required user-defined post processing to the solution vector. Note: the truth solution is stored in the "solution" member of this class, which is inherited from the parent System class several levels up.

Definition at line 734 of file rb_construction.h.

Referenced by enrich_basis_from_rhs_terms(), and truth_solve().

{}

preevaluate_thetas()

void libMesh::RBConstruction::preevaluate_thetas ( )

protectedvirtual

Definition at line 2787 of file rb_construction.C.

References _evaluated_thetas, _preevaluate_thetas_completed, libMesh::RBConstructionBase< LinearImplicitSystem >::get_first_local_training_index(), libMesh::RBConstructionBase< LinearImplicitSystem >::get_local_n_training_samples(), libMesh::RBThetaExpansion::get_n_A_terms(), libMesh::RBParametrized::get_parameters(), get_rb_evaluation(), libMesh::RBEvaluation::get_rb_theta_expansion(), libMesh::index_range(), libMesh::make_range(), and libMesh::RBConstructionBase< LinearImplicitSystem >::set_params_from_training_set().

Referenced by train_reduced_basis_with_greedy().

{
  LOG_SCOPE("preevaluate_thetas()", "RBConstruction");

  _evaluated_thetas.resize(get_local_n_training_samples());

  // Early return if we've already preevaluated thetas.
  if (_preevaluate_thetas_completed)
    return;

  if ( get_local_n_training_samples() == 0 )
    return;

  auto & rb_theta_expansion = get_rb_evaluation().get_rb_theta_expansion();
  const unsigned int n_A_terms = rb_theta_expansion.get_n_A_terms();
  const unsigned int n_F_terms = rb_theta_expansion.get_n_F_terms();
  const unsigned int n_outputs = rb_theta_expansion.get_total_n_output_terms();

  // Collect all training parameters
  // TODO: Here instead of using a vector of RBParameters objects,
  // we could use a single RBParameters object with multiple samples.
  // This would save memory over the current approach, but that may
  // not be a big deal in practice unless the number of training samples
  // is very large for some reason.
  std::vector<RBParameters> mus(get_local_n_training_samples());
  const numeric_index_type first_index = get_first_local_training_index();
  for (unsigned int i=0; i<get_local_n_training_samples(); i++)
    {
      // Load training parameter i, this is only loaded
      // locally since the RB solves are local.
      set_params_from_training_set( first_index+i );
      mus[i] = get_parameters();
      _evaluated_thetas[i].resize(n_A_terms + n_F_terms + n_outputs);
    }

  // Evaluate thetas for all training parameters simultaneously
  for (unsigned int q_a=0; q_a<n_A_terms; q_a++)
    {
      const auto A_vals = rb_theta_expansion.eval_A_theta(q_a, mus);
      for (auto i : make_range(get_local_n_training_samples()))
        _evaluated_thetas[i][q_a] = A_vals[i];
    }

  for (unsigned int q_f=0; q_f<n_F_terms; q_f++)
    {
      const auto F_vals = rb_theta_expansion.eval_F_theta(q_f, mus);
      for (auto i : make_range(get_local_n_training_samples()))
        _evaluated_thetas[i][n_A_terms + q_f] = F_vals[i];
    }

  {
    unsigned int output_counter = 0;
    for (unsigned int n=0; n<rb_theta_expansion.get_n_outputs(); n++)
      for (unsigned int q_l=0; q_l<rb_theta_expansion.get_n_output_terms(n); q_l++)
        {
          // Evaluate the current output functional term for all
          // training parameters simultaneously.
          const auto output_vals = rb_theta_expansion.eval_output_theta(n, q_l, mus);

          // TODO: the size of _evaluated_thetas is currently assumed to be
          // the same as get_local_n_training_samples(), but this won't be
          // the case if we use RBParameters objects that have multiple samples.
          // So just make sure that's the case for now.
          libmesh_error_msg_if(output_vals.size() != get_local_n_training_samples(),
                               "We currently only support single-sample RBParameters "
                               "objects during the training stage.");

          for (auto i : index_range(output_vals))
            _evaluated_thetas[i][n_A_terms + n_F_terms + output_counter] = output_vals[i];

          // Go to next output term
          output_counter++;
        }
  }

  _preevaluate_thetas_completed = true;
}

prefer_hash_table_matrix_assembly()

void libMesh::System::prefer_hash_table_matrix_assembly ( bool  preference )

inlineinherited

Sets whether to use hash table matrix assembly if the matrix sub-classes support it.

Definition at line 2667 of file system.h.

References libMesh::System::_matrices_initialized, and libMesh::System::_prefer_hash_table_matrix_assembly.

Referenced by main().

{
  libmesh_error_msg_if(
      _matrices_initialized,
      "System::prefer_hash_table_matrix_assembly() should be called before matrices are initialized");
  _prefer_hash_table_matrix_assembly = preference;
}

prefix()

std::string libMesh::System::prefix ( ) const

inlineinherited

Returns

A prefix that may be applied to solver options. Note that this prefix is only used if prefix_with_name()

Definition at line 1980 of file system.h.

References libMesh::System::name().

Referenced by libMesh::ImplicitSystem::get_linear_solver(), libMesh::LinearImplicitSystem::solve(), and libMesh::NonlinearImplicitSystem::solve().

{ return this->name() + "_"; }

prefix_with_name() [1/2]

void libMesh::System::prefix_with_name ( bool  value )

inlineinherited

Instructs this system to prefix solve options with its name for solvers that leverage prefixes.

Definition at line 1969 of file system.h.

References libMesh::System::_prefix_with_name, and value.

{ _prefix_with_name = value; }

prefix_with_name() [2/2]

bool libMesh::System::prefix_with_name ( ) const

inlineinherited

Returns

Whether we are name prefixing

Definition at line 1974 of file system.h.

References libMesh::System::_prefix_with_name.

Referenced by libMesh::ContinuationSystem::ContinuationSystem(), libMesh::ImplicitSystem::get_linear_solver(), libMesh::LinearImplicitSystem::solve(), libMesh::NonlinearImplicitSystem::solve(), and libMesh::System::System().

{ return _prefix_with_name; }

print_basis_function_orthogonality()

void libMesh::RBConstruction::print_basis_function_orthogonality

(

)

const

Print out a matrix that shows the orthogonality of the RB basis functions.

This is a helpful debugging tool, e.g. orthogonality can be degraded due to finite precision arithmetic.

Definition at line 391 of file rb_construction.C.

References libMesh::RBEvaluation::get_n_basis_functions(), get_non_dirichlet_inner_product_matrix_if_avail(), get_rb_evaluation(), libMesh::out, libMesh::System::solution, value, and libMesh::SparseMatrix< T >::vector_mult().

Referenced by main().

{
  std::unique_ptr<NumericVector<Number>> temp = solution->clone();

  for (unsigned int i=0; i<get_rb_evaluation().get_n_basis_functions(); i++)
    {
      for (unsigned int j=0; j<get_rb_evaluation().get_n_basis_functions(); j++)
        {
          get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*temp, get_rb_evaluation().get_basis_function(j));
          Number value = temp->dot( get_rb_evaluation().get_basis_function(i) );

          libMesh::out << value << " ";
        }
      libMesh::out << std::endl;
    }
  libMesh::out << std::endl;
}

print_discrete_parameter_values()

void libMesh::RBParametrized::print_discrete_parameter_values ( ) const

inherited

Print out all the discrete parameter values.

Definition at line 366 of file rb_parametrized.C.

References libMesh::RBParametrized::get_discrete_parameter_values(), libMesh::Quality::name(), libMesh::out, and value.

Referenced by libMesh::RBSCMConstruction::print_info(), libMesh::RBEIMConstruction::print_info(), and print_info().

{
  for (const auto & [name, values] : get_discrete_parameter_values())
    {
      libMesh::out << "Discrete parameter " << name << ", values: ";

      for (const auto & value : values)
        libMesh::out << value << " ";
      libMesh::out << std::endl;
    }
}

print_info() [1/3]

void libMesh::ReferenceCounter::print_info ( std::ostream &  out_stream = libMesh::out )

staticinherited

Prints the reference information, by default to libMesh::out.

Definition at line 81 of file reference_counter.C.

References libMesh::ReferenceCounter::_enable_print_counter, and libMesh::ReferenceCounter::get_info().

Referenced by libMesh::LibMeshInit::~LibMeshInit().

{
  if (_enable_print_counter)
    out_stream << ReferenceCounter::get_info();
}

print_info() [2/3]

void libMesh::ReferenceCounter::print_info ( std::ostream &  out_stream = libMesh::out )

staticinherited

Prints the reference information, by default to libMesh::out.

Definition at line 81 of file reference_counter.C.

References libMesh::ReferenceCounter::_enable_print_counter, and libMesh::ReferenceCounter::get_info().

Referenced by libMesh::LibMeshInit::~LibMeshInit().

{
  if (_enable_print_counter)
    out_stream << ReferenceCounter::get_info();
}

print_info() [3/3]

void libMesh::RBConstruction::print_info ( ) const

virtual

Print out info that describes the current setup of this RBConstruction.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 354 of file rb_construction.C.

References get_abs_training_tolerance(), libMesh::RBThetaExpansion::get_n_A_terms(), libMesh::RBThetaExpansion::get_n_F_terms(), libMesh::RBThetaExpansion::get_n_output_terms(), libMesh::RBThetaExpansion::get_n_outputs(), libMesh::RBParametrized::get_n_params(), libMesh::RBConstructionBase< LinearImplicitSystem >::get_n_training_samples(), get_normalize_rb_bound_in_greedy(), libMesh::RBParametrized::get_parameter_max(), libMesh::RBParametrized::get_parameter_min(), libMesh::RBParametrized::get_parameters(), get_rb_theta_expansion(), get_RB_training_type(), get_rel_training_tolerance(), libMesh::RBParametrized::is_discrete_parameter(), libMesh::RBConstructionBase< LinearImplicitSystem >::is_quiet(), is_rb_eval_initialized(), libMesh::System::name(), Nmax, libMesh::out, and libMesh::RBParametrized::print_discrete_parameter_values().

Referenced by main(), and libMesh::TransientRBConstruction::print_info().

{
  // Print out info that describes the current setup
  libMesh::out << std::endl << "RBConstruction parameters:" << std::endl;
  libMesh::out << "system name: " << this->name() << std::endl;
  libMesh::out << "Nmax: " << Nmax << std::endl;
  libMesh::out << "Greedy relative error tolerance: " << get_rel_training_tolerance() << std::endl;
  libMesh::out << "Greedy absolute error tolerance: " << get_abs_training_tolerance() << std::endl;
  libMesh::out << "Do we normalize RB error bound in greedy? " << get_normalize_rb_bound_in_greedy() << std::endl;
  libMesh::out << "RB training type: " << get_RB_training_type() << std::endl;
  if (is_rb_eval_initialized())
    {
      libMesh::out << "Aq operators attached: " << get_rb_theta_expansion().get_n_A_terms() << std::endl;
      libMesh::out << "Fq functions attached: " << get_rb_theta_expansion().get_n_F_terms() << std::endl;
      libMesh::out << "n_outputs: " << get_rb_theta_expansion().get_n_outputs() << std::endl;
      for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
        libMesh::out << "output " << n << ", Q_l = " << get_rb_theta_expansion().get_n_output_terms(n) << std::endl;
    }
  else
    {
      libMesh::out << "RBThetaExpansion member is not set yet" << std::endl;
    }
  libMesh::out << "Number of parameters: " << get_n_params() << std::endl;
  for (const auto & pr : get_parameters())
    if (!is_discrete_parameter(pr.first))
      {
        libMesh::out <<   "Parameter " << pr.first
                     << ": Min = " << get_parameter_min(pr.first)
                     << ", Max = " << get_parameter_max(pr.first) << std::endl;
      }

  print_discrete_parameter_values();
  libMesh::out << "n_training_samples: " << get_n_training_samples() << std::endl;
  libMesh::out << "quiet mode? " << is_quiet() << std::endl;
  libMesh::out << std::endl;
}

print_parameters()

void libMesh::RBParametrized::print_parameters ( ) const

inherited

Print the current parameters.

Definition at line 178 of file rb_parametrized.C.

References libMesh::RBParametrized::get_parameters(), libMesh::RBParametrized::parameters_initialized, and libMesh::RBParameters::print().

Referenced by libMesh::RBEIMConstruction::train_eim_approximation_with_greedy(), and train_reduced_basis_with_greedy().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::print_current_parameters");

  get_parameters().print();
}

process_parameters_file()

void libMesh::RBConstruction::process_parameters_file ( const std::string &  parameters_filename )

virtual

Read in from the file specified by parameters_filename and set the this system's member variables accordingly.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 208 of file rb_construction.C.

References abs_training_tolerance, libMesh::make_range(), Nmax, libMesh::RBConstructionBase< LinearImplicitSystem >::quiet_mode, libMesh::Real, rel_training_tolerance, set_rb_construction_parameters(), and libMesh::RBParameters::set_value().

Referenced by main(), and libMesh::TransientRBConstruction::process_parameters_file().

{
  // First read in data from input_filename
  GetPot infile(parameters_filename);

  const unsigned int n_training_samples = infile("n_training_samples",0);
  const bool deterministic_training = infile("deterministic_training",false);
  unsigned int training_parameters_random_seed_in =
    static_cast<unsigned int>(-1);
  training_parameters_random_seed_in = infile("training_parameters_random_seed",
                                              training_parameters_random_seed_in);
  const bool quiet_mode_in = infile("quiet_mode", quiet_mode);
  const unsigned int Nmax_in = infile("Nmax", Nmax);
  const Real rel_training_tolerance_in = infile("rel_training_tolerance",
                                                rel_training_tolerance);
  const Real abs_training_tolerance_in = infile("abs_training_tolerance",
                                                abs_training_tolerance);

  // Initialize value to false, let the input file value override.
  const bool normalize_rb_bound_in_greedy_in = infile("normalize_rb_bound_in_greedy",
                                                      false);

  const std::string RB_training_type_in = infile("RB_training_type", "Greedy");

  // Read in the parameters from the input file too
  unsigned int n_continuous_parameters = infile.vector_variable_size("parameter_names");
  RBParameters mu_min_in;
  RBParameters mu_max_in;
  for (unsigned int i=0; i<n_continuous_parameters; i++)
    {
      // Read in the parameter names
      std::string param_name = infile("parameter_names", "NONE", i);

      {
        Real min_val = infile(param_name, 0., 0);
        mu_min_in.set_value(param_name, min_val);
      }

      {
        Real max_val = infile(param_name, 0., 1);
        mu_max_in.set_value(param_name, max_val);
      }
    }

  std::map<std::string, std::vector<Real>> discrete_parameter_values_in;

  unsigned int n_discrete_parameters = infile.vector_variable_size("discrete_parameter_names");
  for (unsigned int i=0; i<n_discrete_parameters; i++)
    {
      std::string param_name = infile("discrete_parameter_names", "NONE", i);

      unsigned int n_vals_for_param = infile.vector_variable_size(param_name);
      std::vector<Real> vals_for_param(n_vals_for_param);
      for (auto j : make_range(vals_for_param.size()))
        vals_for_param[j] = infile(param_name, 0., j);

      discrete_parameter_values_in[param_name] = vals_for_param;
    }

  std::map<std::string,bool> log_scaling_in;
  // For now, just set all entries to false.
  // TODO: Implement a decent way to specify log-scaling true/false
  // in the input text file
  for (const auto & pr : mu_min_in)
    log_scaling_in[pr.first] = false;

  // Set the parameters that have been read in
  set_rb_construction_parameters(n_training_samples,
                                 deterministic_training,
                                 static_cast<int>(training_parameters_random_seed_in),
                                 quiet_mode_in,
                                 Nmax_in,
                                 rel_training_tolerance_in,
                                 abs_training_tolerance_in,
                                 normalize_rb_bound_in_greedy_in,
                                 RB_training_type_in,
                                 mu_min_in,
                                 mu_max_in,
                                 discrete_parameter_values_in,
                                 log_scaling_in);
}

processor_id()

processor_id_type libMesh::ParallelObject::processor_id ( ) const

inlineinherited

Returns

The rank of this processor in the group.

Definition at line 114 of file parallel_object.h.

References libMesh::ParallelObject::_communicator, and TIMPI::Communicator::rank().

Referenced by libMesh::BoundaryInfo::_find_id_maps(), libMesh::PetscDMWrapper::add_dofs_to_section(), libMesh::DistributedMesh::add_elem(), libMesh::BoundaryInfo::add_elements(), libMesh::DofMap::add_neighbors_to_send_list(), libMesh::DistributedMesh::add_node(), libMesh::MeshTools::Modification::all_tri(), libMesh::DofMap::allgather_recursive_constraints(), libMesh::FEMSystem::assembly(), libMesh::Nemesis_IO::assert_symmetric_cmaps(), libMesh::Partitioner::assign_partitioning(), libMesh::Nemesis_IO_Helper::build_element_and_node_maps(), libMesh::Partitioner::build_graph(), libMesh::InfElemBuilder::build_inf_elem(), libMesh::BoundaryInfo::build_node_list_from_side_list(), libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::MeshFunction::check_found_elem(), libMesh::DistributedMesh::clear(), libMesh::DistributedMesh::clear_elems(), libMesh::ExodusII_IO_Helper::close(), libMesh::Nemesis_IO_Helper::compute_border_node_ids(), libMesh::Nemesis_IO_Helper::compute_communication_map_parameters(), libMesh::Nemesis_IO_Helper::compute_internal_and_border_elems_and_internal_nodes(), compute_max_error_bound(), libMesh::Nemesis_IO_Helper::compute_node_communication_maps(), libMesh::Nemesis_IO_Helper::compute_num_global_elem_blocks(), libMesh::Nemesis_IO_Helper::compute_num_global_nodesets(), libMesh::Nemesis_IO_Helper::compute_num_global_sidesets(), libMesh::Nemesis_IO_Helper::construct_nemesis_filename(), libMesh::ExodusII_IO::copy_elemental_solution(), libMesh::ExodusII_IO::copy_nodal_solution(), libMesh::Nemesis_IO::copy_scalar_solution(), libMesh::ExodusII_IO::copy_scalar_solution(), libMesh::MeshTools::correct_node_proc_ids(), libMesh::ExodusII_IO_Helper::create(), libMesh::DistributedMesh::delete_elem(), libMesh::MeshCommunication::delete_remote_elements(), libMesh::DofMap::distribute_dofs(), libMesh::DofMap::distribute_scalar_dofs(), libMesh::DistributedMesh::DistributedMesh(), libMesh::DofMapBase::end_dof(), libMesh::DofMapBase::end_old_dof(), libMesh::EnsightIO::EnsightIO(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::SubFunctor::find_dofs_to_send(), libMesh::UnstructuredMesh::find_neighbors(), libMesh::DofMapBase::first_dof(), libMesh::DofMapBase::first_old_dof(), libMesh::RBEIMEvaluation::gather_bfs(), libMesh::Nemesis_IO_Helper::get_cmap_params(), libMesh::Nemesis_IO_Helper::get_eb_info_global(), libMesh::Nemesis_IO_Helper::get_elem_cmap(), libMesh::Nemesis_IO_Helper::get_elem_map(), libMesh::MeshBase::get_info(), libMesh::DofMap::get_info(), libMesh::Nemesis_IO_Helper::get_init_global(), libMesh::Nemesis_IO_Helper::get_init_info(), libMesh::RBEIMEvaluation::get_interior_basis_functions_as_vecs(), libMesh::Nemesis_IO_Helper::get_loadbal_param(), libMesh::DofMap::get_local_constraints(), libMesh::MeshBase::get_local_constraints(), libMesh::Nemesis_IO_Helper::get_node_cmap(), libMesh::Nemesis_IO_Helper::get_node_map(), libMesh::Nemesis_IO_Helper::get_ns_param_global(), libMesh::Nemesis_IO_Helper::get_ss_param_global(), libMesh::SparsityPattern::Build::handle_vi_vj(), libMesh::LaplaceMeshSmoother::init(), libMesh::SystemSubsetBySubdomain::init(), HeatSystem::init_data(), libMesh::ExodusII_IO_Helper::initialize(), libMesh::ExodusII_IO_Helper::initialize_element_variables(), libMesh::ExodusII_IO_Helper::initialize_global_variables(), libMesh::ExodusII_IO_Helper::initialize_nodal_variables(), libMesh::DistributedMesh::insert_elem(), libMesh::DofMap::is_evaluable(), libMesh::SparsityPattern::Build::join(), libMesh::TransientRBEvaluation::legacy_write_offline_data_to_files(), libMesh::RBSCMEvaluation::legacy_write_offline_data_to_files(), libMesh::RBEvaluation::legacy_write_offline_data_to_files(), libMesh::MeshTools::libmesh_assert_consistent_distributed(), libMesh::MeshTools::libmesh_assert_consistent_distributed_nodes(), libMesh::MeshTools::libmesh_assert_contiguous_dof_ids(), libMesh::MeshTools::libmesh_assert_parallel_consistent_procids< Elem >(), libMesh::MeshTools::libmesh_assert_valid_neighbors(), libMesh::DistributedMesh::libmesh_assert_valid_parallel_object_ids(), libMesh::DofMap::local_variable_indices(), main(), libMesh::MeshRefinement::make_coarsening_compatible(), AugmentSparsityOnInterface::mesh_reinit(), libMesh::TriangulatorInterface::MeshedHole::MeshedHole(), libMesh::MeshBase::n_active_local_elem(), libMesh::BoundaryInfo::n_boundary_conds(), libMesh::MeshTools::n_connected_components(), libMesh::MeshBase::n_constraint_rows(), libMesh::BoundaryInfo::n_edge_conds(), libMesh::DofMapBase::n_local_dofs(), libMesh::MeshBase::n_local_elem(), libMesh::MeshBase::n_local_nodes(), libMesh::BoundaryInfo::n_nodeset_conds(), libMesh::BoundaryInfo::n_shellface_conds(), libMesh::RBEIMEvaluation::node_gather_bfs(), libMesh::DistributedMesh::own_node(), libMesh::BoundaryInfo::parallel_sync_node_ids(), libMesh::BoundaryInfo::parallel_sync_side_ids(), libMesh::System::point_gradient(), libMesh::System::point_hessian(), libMesh::System::point_value(), libMesh::MeshBase::print_constraint_rows(), libMesh::DofMap::print_dof_constraints(), libMesh::DofMap::process_mesh_constraint_rows(), libMesh::Nemesis_IO_Helper::put_cmap_params(), libMesh::Nemesis_IO_Helper::put_elem_cmap(), libMesh::Nemesis_IO_Helper::put_elem_map(), libMesh::Nemesis_IO_Helper::put_loadbal_param(), libMesh::Nemesis_IO_Helper::put_node_cmap(), libMesh::Nemesis_IO_Helper::put_node_map(), libMesh::NameBasedIO::read(), libMesh::Nemesis_IO::read(), libMesh::XdrIO::read(), libMesh::CheckpointIO::read(), libMesh::EquationSystems::read(), libMesh::ExodusII_IO_Helper::read_elem_num_map(), libMesh::ExodusII_IO_Helper::read_global_values(), libMesh::ExodusII_IO::read_header(), libMesh::CheckpointIO::read_header(), libMesh::XdrIO::read_header(), libMesh::System::read_header(), libMesh::DynaIO::read_mesh(), libMesh::ExodusII_IO_Helper::read_node_num_map(), libMesh::System::read_parallel_data(), libMesh::TransientRBConstruction::read_riesz_representors_from_files(), read_riesz_representors_from_files(), libMesh::System::read_SCALAR_dofs(), libMesh::XdrIO::read_serialized_bc_names(), libMesh::XdrIO::read_serialized_bcs_helper(), libMesh::System::read_serialized_blocked_dof_objects(), libMesh::XdrIO::read_serialized_connectivity(), libMesh::System::read_serialized_data(), libMesh::XdrIO::read_serialized_nodes(), libMesh::XdrIO::read_serialized_nodesets(), libMesh::XdrIO::read_serialized_subdomain_names(), libMesh::System::read_serialized_vector(), libMesh::System::read_serialized_vectors(), libMesh::Nemesis_IO_Helper::read_var_names_impl(), libMesh::SimplexRefiner::refine_via_edges(), libMesh::StaticCondensationDofMap::reinit(), libMesh::DistributedMesh::renumber_dof_objects(), libMesh::DistributedMesh::renumber_nodes_and_elements(), libMesh::DofMap::scatter_constraints(), libMesh::CheckpointIO::select_split_config(), libMesh::DistributedMesh::set_next_unique_id(), libMesh::DofMap::set_nonlocal_dof_objects(), libMesh::PetscDMWrapper::set_point_range_in_section(), libMesh::RBEIMEvaluation::side_gather_bfs(), MeshFunctionTest::test_bad_gradient_var_with_out_of_mesh_value(), MeshFunctionTest::test_bad_hessian_var_with_out_of_mesh_value(), ExodusTest< elem_type >::test_read_gold(), ExodusTest< elem_type >::test_write(), MeshInputTest::testAbaqusRead(), MeshInputTest::testBadGmsh(), BoundaryInfoTest::testBoundaryIDs(), MeshInputTest::testCopyElementSolutionImpl(), MeshInputTest::testCopyElementVectorImpl(), MeshInputTest::testCopyNodalSolutionImpl(), DefaultCouplingTest::testCoupling(), PointNeighborCouplingTest::testCoupling(), MeshInputTest::testDynaFileMappings(), MeshInputTest::testDynaNoSplines(), MeshInputTest::testDynaReadElem(), MeshInputTest::testDynaReadPatch(), MeshInputTest::testExodusFileMappings(), MeshInputTest::testExodusIGASidesets(), MeshInputTest::testExodusWriteElementDataFromDiscontinuousNodalData(), MeshInputTest::testGmshBCIDOverlap(), MeshInputTest::testGoodGmsh(), MeshInputTest::testGoodSTL(), MeshInputTest::testGoodSTLBinary(), BoundaryInfoTest::testInternalBoundary(), MeshInputTest::testLowOrderEdgeBlocks(), SystemsTest::testProjectMatrix3D(), BoundaryInfoTest::testShellFaceConstraints(), MeshInputTest::testSingleElementImpl(), WriteVecAndScalar::testSolution(), CheckpointIOTest::testSplitter(), MeshInputTest::testTetgenIO(), MeshSmootherTest::testVariationalSmoother(), libMesh::MeshTools::total_weight(), libMesh::NetGenMeshInterface::triangulate(), libMesh::MeshRefinement::uniformly_coarsen(), libMesh::DistributedMesh::update_parallel_id_counts(), libMesh::DTKAdapter::update_variable_values(), libMesh::MeshTools::volume(), libMesh::STLIO::write(), libMesh::NameBasedIO::write(), libMesh::XdrIO::write(), libMesh::CheckpointIO::write(), libMesh::EquationSystems::write(), libMesh::GMVIO::write_discontinuous_gmv(), libMesh::ExodusII_IO::write_element_data(), libMesh::ExodusII_IO_Helper::write_element_values(), libMesh::ExodusII_IO_Helper::write_element_values_element_major(), libMesh::ExodusII_IO_Helper::write_elements(), libMesh::ExodusII_IO_Helper::write_elemset_data(), libMesh::ExodusII_IO_Helper::write_elemsets(), libMesh::ExodusII_IO::write_global_data(), libMesh::ExodusII_IO_Helper::write_global_values(), libMesh::System::write_header(), libMesh::ExodusII_IO::write_information_records(), libMesh::ExodusII_IO_Helper::write_information_records(), libMesh::ExodusII_IO_Helper::write_nodal_coordinates(), libMesh::UCDIO::write_nodal_data(), libMesh::VTKIO::write_nodal_data(), libMesh::ExodusII_IO::write_nodal_data(), libMesh::ExodusII_IO::write_nodal_data_common(), libMesh::ExodusII_IO::write_nodal_data_discontinuous(), libMesh::ExodusII_IO_Helper::write_nodal_values(), libMesh::ExodusII_IO_Helper::write_nodeset_data(), libMesh::Nemesis_IO_Helper::write_nodesets(), libMesh::ExodusII_IO_Helper::write_nodesets(), libMesh::RBEIMEvaluation::write_out_interior_basis_functions(), libMesh::RBEIMEvaluation::write_out_node_basis_functions(), libMesh::RBEIMEvaluation::write_out_side_basis_functions(), write_output_solvedata(), libMesh::System::write_parallel_data(), write_riesz_representors_to_files(), libMesh::System::write_SCALAR_dofs(), libMesh::XdrIO::write_serialized_bc_names(), libMesh::XdrIO::write_serialized_bcs_helper(), libMesh::System::write_serialized_blocked_dof_objects(), libMesh::XdrIO::write_serialized_connectivity(), libMesh::System::write_serialized_data(), libMesh::XdrIO::write_serialized_nodes(), libMesh::XdrIO::write_serialized_nodesets(), libMesh::XdrIO::write_serialized_subdomain_names(), libMesh::System::write_serialized_vector(), libMesh::System::write_serialized_vectors(), libMesh::ExodusII_IO_Helper::write_sideset_data(), libMesh::Nemesis_IO_Helper::write_sidesets(), libMesh::ExodusII_IO_Helper::write_sidesets(), libMesh::ExodusII_IO::write_timestep(), libMesh::ExodusII_IO_Helper::write_timestep(), and libMesh::ExodusII_IO::write_timestep_discontinuous().

  { return cast_int<processor_id_type>(_communicator.rank()); }

project_solution() [1/3]

void libMesh::System::project_solution ( FunctionBase< Number > *  f, FunctionBase< Gradient > *  g = nullptr, std::optional< ConstElemRange >  active_local_range = std::nullopt, std::optional< std::vector< unsigned int >>  variable_numbers = std::nullopt  ) const

inherited

Projects arbitrary functions onto the current solution.

This method projects an arbitrary function onto the solution via L2 projections and nodal interpolations on each element.

The function value f and its gradient g are user-provided cloneable functors. A gradient g is only required/used for projecting onto finite element spaces with continuous derivatives. elem_range active_local_range, if provided, indicates the range of elements over which to perform the projection. variable_numbers variable_numbers, if provided, indicates the variable numbers onto which to project.

Definition at line 1064 of file system_projection.C.

Referenced by init_sys(), initialize(), main(), set_initial_condition(), SlitMeshRefinedSystemTest::setUp(), FETestBase< order, family, elem_type, N_x >::setUp(), MeshFunctionTest::test_bad_gradient_var_with_out_of_mesh_value(), MeshFunctionTest::test_bad_hessian_var_with_out_of_mesh_value(), MeshfunctionDFEM::test_mesh_function_dfem(), MeshfunctionDFEM::test_mesh_function_dfem_grad(), MeshFunctionTest::test_p_level(), ProjectSolutionTest::test_partial_project_solution(), MeshFunctionTest::test_subdomain_id_sets(), MeshInputTest::testCopyElementSolutionImpl(), MeshInputTest::testCopyNodalSolutionImpl(), DefaultCouplingTest::testCoupling(), PointNeighborCouplingTest::testCoupling(), SystemsTest::testProjectCubeWithMeshFunction(), MeshInputTest::testProjectionRegression(), EquationSystemsTest::testRepartitionThenReinit(), and libMesh::MeshfreeSolutionTransfer::transfer().

{
  this->project_vector(*solution, f, g, /*is_adjoint=*/-1, active_local_range, variable_numbers);

  solution->localize(*current_local_solution, _dof_map->get_send_list());
}

project_solution() [2/3]

void libMesh::System::project_solution ( FEMFunctionBase< Number > *  f, FEMFunctionBase< Gradient > *  g = nullptr, std::optional< ConstElemRange >  active_local_range = std::nullopt, std::optional< std::vector< unsigned int >>  variable_numbers = std::nullopt  ) const

inherited

Projects arbitrary functions onto the current solution.

This method projects an arbitrary function onto the solution via L2 projections and nodal interpolations on each element.

The function value f and its gradient g are user-provided cloneable functors. A gradient g is only required/used for projecting onto finite element spaces with continuous derivatives. elem_range active_local_range, if provided, indicates the range of elements over which to perform the projection. variable_numbers variable_numbers, if provided, indicates the variable numbers onto which to project.

Definition at line 1079 of file system_projection.C.

{
  this->project_vector(*solution, f, g, /*is_adjoint=*/-1, active_local_range, variable_numbers);

  solution->localize(*current_local_solution, _dof_map->get_send_list());
}

project_solution() [3/3]

void libMesh::System::project_solution ( ValueFunctionPointer  fptr, GradientFunctionPointer  gptr, const Parameters &  parameters, std::optional< ConstElemRange >  active_local_range = std::nullopt, std::optional< std::vector< unsigned int >>  variable_numbers = std::nullopt  ) const

inherited

This method projects an arbitrary function onto the solution via L2 projections and nodal interpolations on each element.

Definition at line 1048 of file system_projection.C.

References fptr(), and gptr().

{
  WrappedFunction<Number> f(*this, fptr, &function_parameters);
  WrappedFunction<Gradient> g(*this, gptr, &function_parameters);
  this->project_solution(&f, &g, active_local_range, variable_numbers);
}

project_solution_on_reinit()

bool& libMesh::System::project_solution_on_reinit ( void  )

inlineinherited

Tells the System whether or not to project the solution vector onto new grids when the system is reinitialized.

The solution will be projected unless project_solution_on_reinit() = false is called.

Definition at line 884 of file system.h.

References libMesh::System::_solution_projection.

Referenced by libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::AdjointRefinementEstimator::estimate_error(), and libMesh::MemoryHistoryData::store_vectors().

  { return _solution_projection; }

project_vector() [1/5]

void libMesh::System::project_vector ( NumericVector< Number > &  new_vector, FunctionBase< Number > *  f, FunctionBase< Gradient > *  g = nullptr, int  is_adjoint = -1, std::optional< ConstElemRange >  active_local_range = std::nullopt, std::optional< std::vector< unsigned int >>  variable_numbers = std::nullopt  ) const

inherited

Projects arbitrary functions onto a vector of degree of freedom values for the current system.

This method projects an arbitrary function via L2 projections and nodal interpolations on each element.

The function value f and its gradient g are user-provided cloneable functors. A gradient g is only required/used for projecting onto finite element spaces with continuous derivatives. elem_range active_local_range, if provided, indicates the range of elements over which to perform the projection. variable_numbers variable_numbers, if provided, indicates the variable numbers onto which to project.

Constrain the new vector using the requested adjoint rather than primal constraints if is_adjoint is non-negative.

Definition at line 1111 of file system_projection.C.

References libMesh::libmesh_assert().

Referenced by main(), libMesh::NewmarkSolver::project_initial_accel(), libMesh::SecondOrderUnsteadySolver::project_initial_rate(), libMesh::InterMeshProjection::project_system_vectors(), and libMesh::System::restrict_vectors().

{
  LOG_SCOPE ("project_vector(FunctionBase)", "System");

  libmesh_assert(f);

  WrappedFunctor<Number> f_fem(*f);

  if (g)
    {
      WrappedFunctor<Gradient> g_fem(*g);

      this->project_vector(new_vector, &f_fem, &g_fem, is_adjoint, active_local_range, variable_numbers);
    }
  else
    this->project_vector(new_vector, &f_fem, nullptr, is_adjoint, active_local_range, variable_numbers);
}

project_vector() [2/5]

void libMesh::System::project_vector ( NumericVector< Number > &  new_vector, FEMFunctionBase< Number > *  f, FEMFunctionBase< Gradient > *  g = nullptr, int  is_adjoint = -1, std::optional< ConstElemRange >  active_local_range = std::nullopt, std::optional< std::vector< unsigned int >>  variable_numbers = std::nullopt  ) const

inherited

Projects arbitrary functions onto a vector of degree of freedom values for the current system.

This method projects an arbitrary function via L2 projections and nodal interpolations on each element.

The function value f and its gradient g are user-provided cloneable functors. A gradient g is only required/used for projecting onto finite element spaces with continuous derivatives. elem_range active_local_range, if provided, indicates the range of elements over which to perform the projection. variable_numbers variable_numbers, if provided, indicates the variable numbers onto which to project.

Constrain the new vector using the requested adjoint rather than primal constraints if is_adjoint is non-negative.

Definition at line 1139 of file system_projection.C.

References libMesh::NumericVector< T >::close(), libMesh::FEMFunctionBase< Output >::component(), libMesh::FEType::family, libMesh::libmesh_assert(), libMesh::libmesh_ignore(), n_vars, libMesh::NODEELEM, libMesh::FEMContext::pre_fe_reinit(), libMesh::RATIONAL_BERNSTEIN, libMesh::SCALAR, libMesh::DofMap::SCALAR_dof_indices(), libMesh::NumericVector< T >::set(), and libMesh::Variable::type().

{
  LOG_SCOPE ("project_fem_vector()", "System");

  libmesh_assert (f);

  if (!active_local_range)
    {
      active_local_range.emplace
        (this->get_mesh().active_local_elements_begin(),
         this->get_mesh().active_local_elements_end());
    }

  VectorSetAction<Number> setter(new_vector);

  const unsigned int n_variables = this->n_vars();

  std::vector<unsigned int> vars;
  if (variable_numbers)
    {
      vars = *variable_numbers;
      for (auto v : vars)
        if (v >= n_variables)
          libmesh_error_msg("ERROR: variable number " << v <<
                           " out of range for system with " <<
                           n_variables << " variables.");
    }
  else
    {
      vars.resize(n_variables);
      std::iota(vars.begin(), vars.end(), 0);
    }


  // Use a typedef to make the calling sequence for parallel_for() a bit more readable
  typedef
    GenericProjector<FEMFunctionWrapper<Number>, FEMFunctionWrapper<Gradient>,
                     Number, VectorSetAction<Number>> FEMProjector;

  FEMFunctionWrapper<Number> fw(*f);

  if (g)
    {
      FEMFunctionWrapper<Gradient> gw(*g);

      FEMProjector projector(*this, fw, &gw, setter, vars);
      projector.project(active_local_range.value());
    }
  else
    {
      FEMProjector projector(*this, fw, nullptr, setter, vars);
      projector.project(active_local_range.value());
    }

  // Also, load values into the SCALAR dofs
  // Note: We assume that all SCALAR dofs are on the
  // processor with highest ID
  if (this->processor_id() == (this->n_processors()-1))
    {
      // FIXME: Do we want to first check for SCALAR vars before building this? [PB]
      FEMContext context( *this );

      const DofMap & dof_map = this->get_dof_map();
      for (auto var : vars)
        if (this->variable(var).type().family == SCALAR)
          {
            // FIXME: We reinit with an arbitrary element in case the user
            //        doesn't override FEMFunctionBase::component. Is there
            //        any use case we're missing? [PB]
            context.pre_fe_reinit(*this, *(this->get_mesh().active_local_elements_begin()));

            std::vector<dof_id_type> SCALAR_indices;
            dof_map.SCALAR_dof_indices (SCALAR_indices, var);
            const unsigned int n_SCALAR_dofs =
              cast_int<unsigned int>(SCALAR_indices.size());

            for (unsigned int i=0; i<n_SCALAR_dofs; i++)
              {
                const dof_id_type global_index = SCALAR_indices[i];
                const unsigned int component_index =
                  this->variable_scalar_number(var,i);

                new_vector.set(global_index, f->component(context, component_index, Point(), this->time));
              }
          }
    }

  new_vector.close();

  // Look for spline bases, in which case we need to backtrack
  // to calculate the spline DoF values.
  std::vector<const Variable *> rational_vars;
  for (auto varnum : vars)
    {
      const Variable & var = this->get_dof_map().variable(varnum);
      if (var.type().family == RATIONAL_BERNSTEIN)
        rational_vars.push_back(&var);
    }

  // Okay, but are we really using any *spline* bases, or just
  // unconstrained rational bases?
  bool using_spline_bases = false;
  if (!rational_vars.empty())
    {
      // Look for a spline node: a NodeElem with a rational variable
      // on it.
      for (auto & elem : active_local_range.value())
        if (elem->type() == NODEELEM)
          for (auto rational_var : rational_vars)
            if (rational_var->active_on_subdomain(elem->subdomain_id()))
              {
                using_spline_bases = true;
                goto checked_on_splines;
              }
    }

checked_on_splines:

  // Not every processor may have a NodeElem, especially while
  // we're not partitioning them efficiently yet.
  this->comm().max(using_spline_bases);

  if (using_spline_bases)
    this->solve_for_unconstrained_dofs(new_vector, is_adjoint);

#ifdef LIBMESH_ENABLE_CONSTRAINTS
  if (is_adjoint == -1)
    this->get_dof_map().enforce_constraints_exactly(*this, &new_vector);
  else if (is_adjoint >= 0)
    this->get_dof_map().enforce_adjoint_constraints_exactly(new_vector,
                                                            is_adjoint);
#else
  libmesh_ignore(is_adjoint);
#endif
}

project_vector() [3/5]

void libMesh::System::project_vector ( ValueFunctionPointer  fptr, GradientFunctionPointer  gptr, const Parameters &  parameters, NumericVector< Number > &  new_vector, int  is_adjoint = -1, std::optional< ConstElemRange >  active_local_range = std::nullopt, std::optional< std::vector< unsigned int >>  variable_numbers = std::nullopt  ) const

inherited

Projects arbitrary functions onto a vector of degree of freedom values for the current system.

This method projects an arbitrary function via L2 projections and nodal interpolations on each element.

The function value fptr and its gradient gptr are represented by function pointers. A gradient gptr is only required/used for projecting onto finite element spaces with continuous derivatives. elem_range active_local_range, if provided, indicates the range of elements over which to perform the projection. variable_numbers variable_numbers, if provided, indicates the variable numbers onto which to project.

Constrain the new vector using the requested adjoint rather than primal constraints if is_adjoint is non-negative.

Definition at line 1094 of file system_projection.C.

References fptr(), and gptr().

{
  WrappedFunction<Number> f(*this, fptr, &function_parameters);
  WrappedFunction<Gradient> g(*this, gptr, &function_parameters);
  this->project_vector(new_vector, &f, &g, is_adjoint, active_local_range, variable_numbers);
}

project_vector() [4/5]

void libMesh::System::project_vector ( NumericVector< Number > &  vector, int  is_adjoint = -1, std::optional< ConstElemRange >  active_local_range = std::nullopt, std::optional< std::vector< unsigned int >>  variable_numbers = std::nullopt  ) const

protectedinherited

Projects the vector defined on the old mesh onto the new mesh.

Constrain the new vector using the requested adjoint rather than primal constraints if is_adjoint is non-negative.

Definition at line 247 of file system_projection.C.

References libMesh::NumericVector< T >::clone().

{
  // Create a copy of the vector, which currently
  // contains the old data.
  std::unique_ptr<NumericVector<Number>>
    old_vector (vector.clone());

  // Project the old vector to the new vector
  this->project_vector (*old_vector, vector, is_adjoint, active_local_range, variable_numbers);
}

project_vector() [5/5]

void libMesh::System::project_vector ( const NumericVector< Number > &  old_v, NumericVector< Number > &  new_v, int  is_adjoint = -1, std::optional< ConstElemRange >  active_local_range = std::nullopt, std::optional< std::vector< unsigned int >>  variable_numbers = std::nullopt  ) const

protectedinherited

Projects the vector defined on the old mesh onto the new mesh.

This method projects the vector via L2 projections or nodal interpolations on each element.

The original vector is unchanged and the new vector is passed through the second argument.

Constrain the new vector using the requested adjoint rather than primal constraints if is_adjoint is non-negative.

This method projects a solution from an old mesh to a current, refined mesh. The input vector old_v gives the solution on the old mesh, while the new_v gives the solution (to be computed) on the new mesh.

Definition at line 267 of file system_projection.C.

References libMesh::NumericVector< T >::clear(), libMesh::NumericVector< T >::close(), libMesh::NumericVector< T >::get(), libMesh::GHOSTED, libMesh::index_range(), libMesh::NumericVector< T >::init(), libMesh::libmesh_assert(), libMesh::libmesh_ignore(), libMesh::NumericVector< T >::local_size(), libMesh::NumericVector< T >::localize(), libMesh::make_range(), n_vars, libMesh::PARALLEL, libMesh::Threads::parallel_reduce(), libMesh::SCALAR, libMesh::DofMap::SCALAR_dof_indices(), libMesh::BuildProjectionList::send_list, libMesh::SERIAL, libMesh::NumericVector< T >::set(), libMesh::NumericVector< T >::size(), libMesh::NumericVector< T >::type(), libMesh::TYPE_SCALAR, and libMesh::BuildProjectionList::unique().

{
  LOG_SCOPE ("project_vector(old,new)", "System");

  new_v.clear();

#ifdef LIBMESH_ENABLE_AMR

  // Resize the new vector and get a serial version.
  NumericVector<Number> * new_vector_ptr = nullptr;
  std::unique_ptr<NumericVector<Number>> new_vector_built;
  NumericVector<Number> * local_old_vector;
  std::unique_ptr<NumericVector<Number>> local_old_vector_built;
  const NumericVector<Number> * old_vector_ptr = nullptr;

  if (!active_local_range)
    {
      active_local_range.emplace
        (this->get_mesh().active_local_elements_begin(),
         this->get_mesh().active_local_elements_end());
    }

  // If the old vector was uniprocessor, make the new
  // vector uniprocessor
  if (old_v.type() == SERIAL)
    {
      new_v.init (this->n_dofs(), false, SERIAL);
      new_vector_ptr = &new_v;
      old_vector_ptr = &old_v;
    }

  // Otherwise it is a parallel, distributed vector, which
  // we need to localize.
  else if (old_v.type() == PARALLEL)
    {
      // Build a send list for efficient localization
      BuildProjectionList projection_list(*this);
      Threads::parallel_reduce (active_local_range.value(),
                                projection_list);

      // Create a sorted, unique send_list
      projection_list.unique();

      new_v.init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
      new_vector_built = NumericVector<Number>::build(this->comm());
      local_old_vector_built = NumericVector<Number>::build(this->comm());
      new_vector_ptr = new_vector_built.get();
      local_old_vector = local_old_vector_built.get();
      new_vector_ptr->init(this->n_dofs(), this->n_local_dofs(),
                           this->get_dof_map().get_send_list(), false,
                           GHOSTED);
      local_old_vector->init(old_v.size(), old_v.local_size(),
                             projection_list.send_list, false, GHOSTED);
      old_v.localize(*local_old_vector, projection_list.send_list);
      local_old_vector->close();
      old_vector_ptr = local_old_vector;
    }
  else if (old_v.type() == GHOSTED)
    {
      // Build a send list for efficient localization
      BuildProjectionList projection_list(*this);
      Threads::parallel_reduce (active_local_range.value(),
                                projection_list);

      // Create a sorted, unique send_list
      projection_list.unique();

      new_v.init (this->n_dofs(), this->n_local_dofs(),
                  this->get_dof_map().get_send_list(), false, GHOSTED);

      local_old_vector_built = NumericVector<Number>::build(this->comm());
      new_vector_ptr = &new_v;
      local_old_vector = local_old_vector_built.get();
      local_old_vector->init(old_v.size(), old_v.local_size(),
                             projection_list.send_list, false, GHOSTED);
      old_v.localize(*local_old_vector, projection_list.send_list);
      local_old_vector->close();
      old_vector_ptr = local_old_vector;
    }
  else // unknown old_v.type()
    libmesh_error_msg("ERROR: Unknown old_v.type() == " << old_v.type());

  // Note that the above will have zeroed the new_vector.
  // Just to be sure, assert that new_vector_ptr and old_vector_ptr
  // were successfully set before trying to deref them.
  libmesh_assert(new_vector_ptr);
  libmesh_assert(old_vector_ptr);

  NumericVector<Number> & new_vector = *new_vector_ptr;
  const NumericVector<Number> & old_vector = *old_vector_ptr;

  const unsigned int n_variables = this->n_vars();

  if (n_variables)
    {
      std::vector<unsigned int> vars;
      if (variable_numbers)
        {
          vars = *variable_numbers;
          for (auto v : vars)
            if (v >= n_variables)
              libmesh_error_msg("ERROR: variable number " << v <<
                               " out of range for system with " <<
                               n_variables << " variables.");
        }
      else
        {
          vars.resize(n_variables);
          std::iota(vars.begin(), vars.end(), 0);
        }

      std::vector<unsigned int> regular_vars, vector_vars;
      for (auto var : vars)
      {
        if (FEInterface::field_type(this->variable_type(var)) == TYPE_SCALAR)
          regular_vars.push_back(var);
        else
          vector_vars.push_back(var);
      }

      VectorSetAction<Number> setter(new_vector);

      if (!regular_vars.empty())
        {
          // Use a typedef to make the calling sequence for parallel_for() a bit more readable
          typedef
            GenericProjector<OldSolutionValue<Number,   &FEMContext::point_value>,
                             OldSolutionValue<Gradient, &FEMContext::point_gradient>,
                             Number, VectorSetAction<Number>> FEMProjector;

          OldSolutionValue<Number,   &FEMContext::point_value>
            f(*this, old_vector, &regular_vars);
          OldSolutionValue<Gradient, &FEMContext::point_gradient>
            g(*this, old_vector, &regular_vars);

          FEMProjector projector(*this, f, &g, setter, regular_vars);
          projector.project(active_local_range.value());
        }

      if (!vector_vars.empty())
        {
          typedef
            GenericProjector<OldSolutionValue<Gradient,   &FEMContext::point_value>,
                             OldSolutionValue<Tensor, &FEMContext::point_gradient>,
                             Gradient, VectorSetAction<Number>> FEMVectorProjector;

          OldSolutionValue<Gradient, &FEMContext::point_value> f_vector(*this, old_vector, &vector_vars);
          OldSolutionValue<Tensor, &FEMContext::point_gradient> g_vector(*this, old_vector, &vector_vars);

          FEMVectorProjector vector_projector(*this, f_vector, &g_vector, setter, vector_vars);
          vector_projector.project(active_local_range.value());
        }

      // Copy the SCALAR dofs from old_vector to new_vector
      // Note: We assume that all SCALAR dofs are on the
      // processor with highest ID
      if (this->processor_id() == (this->n_processors()-1))
        {
          const DofMap & dof_map = this->get_dof_map();
          for (auto var : vars)
            if (this->variable(var).type().family == SCALAR)
              {
                // We can just map SCALAR dofs directly across
                std::vector<dof_id_type> new_SCALAR_indices, old_SCALAR_indices;
                dof_map.SCALAR_dof_indices (new_SCALAR_indices, var, false);
                dof_map.SCALAR_dof_indices (old_SCALAR_indices, var, true);
                for (auto i : index_range(new_SCALAR_indices))
                  new_vector.set(new_SCALAR_indices[i], old_vector(old_SCALAR_indices[i]));
              }
        }
    }

  new_vector.close();

  // If the old vector was serial, we probably need to send our values
  // to other processors
  //
  // FIXME: I'm not sure how to make a NumericVector do that without
  // creating a temporary parallel vector to use localize! - RHS
  if (old_v.type() == SERIAL)
    {
      std::unique_ptr<NumericVector<Number>> dist_v = NumericVector<Number>::build(this->comm());
      dist_v->init(this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
      dist_v->close();

      for (auto i : make_range(dist_v->size()))
        if (new_vector(i) != 0.0)
          dist_v->set(i, new_vector(i));

      dist_v->close();

      dist_v->localize (new_v, this->get_dof_map().get_send_list());
      new_v.close();
    }
  // If the old vector was parallel, we need to update it
  // and free the localized copies
  else if (old_v.type() == PARALLEL)
    {
      // We may have to set dof values that this processor doesn't
      // own in certain special cases, like LAGRANGE FIRST or
      // HERMITE THIRD elements on second-order meshes?
      new_v = new_vector;
      new_v.close();
    }


  // Apply constraints only if we we are asked to
  if(this->project_with_constraints)
  {
    if (is_adjoint == -1)
    {
      this->get_dof_map().enforce_constraints_exactly(*this, &new_v);
    }
    else if (is_adjoint >= 0)
    {
      this->get_dof_map().enforce_adjoint_constraints_exactly(new_v,
                                                            is_adjoint);
    }
  }
#else

  // AMR is disabled: simply copy the vector
  new_v = old_v;

  libmesh_ignore(is_adjoint, active_local_range, variable_numbers);

#endif // #ifdef LIBMESH_ENABLE_AMR
}

projection_matrix()

void libMesh::System::projection_matrix ( SparseMatrix< Number > &  proj_mat ) const

inherited

This method creates a projection matrix which corresponds to the operation of project_vector between old and new solution spaces.

Heterogeneous Dirichlet boundary conditions are not taken into account here; if this matrix is used for prolongation (mesh refinement) on a side with a heterogeneous BC, the newly created degrees of freedom on that side will still match the coarse grid approximation of the BC, not the fine grid approximation.

Definition at line 982 of file system_projection.C.

References libMesh::make_range(), n_vars, libMesh::SCALAR, libMesh::DofMap::SCALAR_dof_indices(), and libMesh::SparseMatrix< T >::set().

Referenced by libMesh::PetscDMWrapper::init_petscdm(), SystemsTest::testProjectMatrix1D(), SystemsTest::testProjectMatrix2D(), and SystemsTest::testProjectMatrix3D().

{
  LOG_SCOPE ("projection_matrix()", "System");

  const unsigned int n_variables = this->n_vars();

  if (n_variables)
    {
      ConstElemRange active_local_elem_range
        (this->get_mesh().active_local_elements_begin(),
         this->get_mesh().active_local_elements_end());

      std::vector<unsigned int> vars(n_variables);
      std::iota(vars.begin(), vars.end(), 0);

      // Use a typedef to make the calling sequence for parallel_for() a bit more readable
      typedef OldSolutionCoefs<Real, &FEMContext::point_value> OldSolutionValueCoefs;
      typedef OldSolutionCoefs<RealGradient, &FEMContext::point_gradient> OldSolutionGradientCoefs;

      typedef
        GenericProjector<OldSolutionValueCoefs,
                         OldSolutionGradientCoefs,
                         DynamicSparseNumberArray<Real,dof_id_type>,
                         MatrixFillAction<Real, Number> > ProjMatFiller;

      OldSolutionValueCoefs    f(*this, &vars);
      OldSolutionGradientCoefs g(*this, &vars);
      MatrixFillAction<Real, Number> setter(proj_mat);

      ProjMatFiller mat_filler(*this, f, &g, setter, vars);
      mat_filler.project(active_local_elem_range);

      // Set the SCALAR dof transfer entries too.
      // Note: We assume that all SCALAR dofs are on the
      // processor with highest ID
      if (this->processor_id() == (this->n_processors()-1))
        {
          const DofMap & dof_map = this->get_dof_map();
          for (auto var : make_range(this->n_vars()))
            if (this->variable(var).type().family == SCALAR)
              {
                // We can just map SCALAR dofs directly across
                std::vector<dof_id_type> new_SCALAR_indices, old_SCALAR_indices;
                dof_map.SCALAR_dof_indices (new_SCALAR_indices, var, false);
                dof_map.SCALAR_dof_indices (old_SCALAR_indices, var, true);
                const unsigned int new_n_dofs =
                  cast_int<unsigned int>(new_SCALAR_indices.size());

                for (unsigned int i=0; i<new_n_dofs; i++)
                  {
                    proj_mat.set( new_SCALAR_indices[i],
                                  old_SCALAR_indices[i], 1);
                  }
              }
        }
    }
}

prolong_vectors()

void libMesh::System::prolong_vectors ( )

virtualinherited

Prolong vectors after the mesh has refined.

Definition at line 432 of file system.C.

References libMesh::System::restrict_vectors().

Referenced by libMesh::EquationSystems::reinit_solutions().

{
#ifdef LIBMESH_ENABLE_AMR
  // Currently project_vector handles both restriction and prolongation
  this->restrict_vectors();
#endif
}

qoi_parameter_hessian()

void libMesh::ImplicitSystem::qoi_parameter_hessian ( const QoISet &  qoi_indices, const ParameterVector &  parameters, SensitivityData &  hessian  )

overridevirtualinherited

For each of the system's quantities of interest q in qoi[qoi_indices], and for a vector of parameters p, the parameter sensitivity Hessian H_ij is defined as H_ij = (d^2 q)/(d p_i d p_j) This Hessian is the output of this method, where for each q_i, H_jk is stored in hessian.second_derivative(i,j,k).

Note that in some cases only current_local_solution is used during assembly, and, therefore, if solution has been altered without update() being called, then the user must call update() before calling this function.

Reimplemented from libMesh::System.

Definition at line 922 of file implicit_system.C.

References libMesh::ImplicitSystem::adjoint_solve(), libMesh::SensitivityData::allocate_hessian_data(), libMesh::ExplicitSystem::assemble_qoi(), libMesh::ExplicitSystem::assemble_qoi_derivative(), libMesh::ImplicitSystem::assembly(), libMesh::NumericVector< T >::close(), libMesh::SparseMatrix< T >::close(), libMesh::NumericVector< T >::dot(), libMesh::System::get_adjoint_rhs(), libMesh::System::get_adjoint_solution(), libMesh::System::get_qoi_value(), libMesh::System::get_qoi_values(), libMesh::System::get_sensitivity_solution(), libMesh::QoISet::has_index(), libMesh::System::is_adjoint_already_solved(), libMesh::ImplicitSystem::matrix, libMesh::System::n_qois(), libMesh::Real, libMesh::ExplicitSystem::rhs, libMesh::SensitivityData::second_derivative(), libMesh::ImplicitSystem::sensitivity_solve(), libMesh::ParameterVector::size(), libMesh::System::solution, libMesh::TOLERANCE, libMesh::System::update(), and libMesh::SparseMatrix< T >::vector_mult().

{
  // We currently get partial derivatives via finite differencing
  const Real delta_p = TOLERANCE;

  ParameterVector & parameters_vec =
    const_cast<ParameterVector &>(parameters_in);

  // We'll use one temporary vector for matrix-vector-vector products
  std::unique_ptr<NumericVector<Number>> tempvec = this->solution->zero_clone();

  // And another temporary vector to hold a copy of the true solution
  // so we can safely perturb this->solution.
  std::unique_ptr<NumericVector<Number>> oldsolution = this->solution->clone();

  const unsigned int Np = cast_int<unsigned int>
    (parameters_vec.size());
  const unsigned int Nq = this->n_qois();

  // For each quantity of interest q, the parameter sensitivity
  // Hessian is defined as q''_{kl} = {d^2 q}/{d p_k d p_l}.
  //
  // We calculate it from values and partial derivatives of the
  // quantity of interest function Q, solution u, adjoint solution z,
  // and residual R, as:
  //
  // q''_{kl} =
  // Q''_{kl} + Q''_{uk}(u)*u'_l + Q''_{ul}(u) * u'_k +
  // Q''_{uu}(u)*u'_k*u'_l -
  // R''_{kl}(u,z) -
  // R''_{uk}(u,z)*u'_l - R''_{ul}(u,z)*u'_k -
  // R''_{uu}(u,z)*u'_k*u'_l
  //
  // See the adjoints model document for more details.

  // We first do an adjoint solve to get z for each quantity of
  // interest
  // if we haven't already or dont have an initial condition for the adjoint
  if (!this->is_adjoint_already_solved())
    {
      this->adjoint_solve(qoi_indices);
    }

  // And a sensitivity solve to get u_k for each parameter
  this->sensitivity_solve(parameters_vec);

  // Get ready to fill in second derivatives:
  sensitivities.allocate_hessian_data(qoi_indices, *this, parameters_vec);

  for (unsigned int k=0; k != Np; ++k)
    {
      Number old_parameterk = *parameters_vec[k];

      // The Hessian is symmetric, so we just calculate the lower
      // triangle and the diagonal, and we get the upper triangle from
      // the transpose of the lower

      for (unsigned int l=0; l != k+1; ++l)
        {
          // The second partial derivatives with respect to parameters_vec
          // are all calculated via a central finite difference
          // stencil:
          // F''_{kl} ~= (F(p+dp*e_k+dp*e_l) - F(p+dp*e_k-dp*e_l) -
          //              F(p-dp*e_k+dp*e_l) + F(p-dp*e_k-dp*e_l))/(4*dp^2)
          // We will add Q''_{kl}(u) and subtract R''_{kl}(u,z) at the
          // same time.
          //
          // We have to be careful with the perturbations to handle
          // the k=l case

          Number old_parameterl = *parameters_vec[l];

          *parameters_vec[k] += delta_p;
          *parameters_vec[l] += delta_p;
          this->assemble_qoi(qoi_indices);
          this->assembly(true, false, true);
          this->rhs->close();
          std::vector<Number> partial2q_term = this->get_qoi_values();
          std::vector<Number> partial2R_term(this->n_qois());
          for (unsigned int i=0; i != Nq; ++i)
            if (qoi_indices.has_index(i))
              partial2R_term[i] = this->rhs->dot(this->get_adjoint_solution(i));

          *parameters_vec[l] -= 2.*delta_p;
          this->assemble_qoi(qoi_indices);
          this->assembly(true, false, true);
          this->rhs->close();
          for (unsigned int i=0; i != Nq; ++i)
            if (qoi_indices.has_index(i))
              {
                partial2q_term[i] -= this->get_qoi_value(i);
                partial2R_term[i] -= this->rhs->dot(this->get_adjoint_solution(i));
              }

          *parameters_vec[k] -= 2.*delta_p;
          this->assemble_qoi(qoi_indices);
          this->assembly(true, false, true);
          this->rhs->close();
          for (unsigned int i=0; i != Nq; ++i)
            if (qoi_indices.has_index(i))
              {
                partial2q_term[i] += this->get_qoi_value(i);
                partial2R_term[i] += this->rhs->dot(this->get_adjoint_solution(i));
              }

          *parameters_vec[l] += 2.*delta_p;
          this->assemble_qoi(qoi_indices);
          this->assembly(true, false, true);
          this->rhs->close();
          for (unsigned int i=0; i != Nq; ++i)
            if (qoi_indices.has_index(i))
              {
                partial2q_term[i] -= this->get_qoi_value(i);
                partial2R_term[i] -= this->rhs->dot(this->get_adjoint_solution(i));
                partial2q_term[i] /= (4. * delta_p * delta_p);
                partial2R_term[i] /= (4. * delta_p * delta_p);
              }

          for (unsigned int i=0; i != Nq; ++i)
            if (qoi_indices.has_index(i))
              {
                Number current_terms = partial2q_term[i] - partial2R_term[i];
                sensitivities.second_derivative(i,k,l) += current_terms;
                if (k != l)
                  sensitivities.second_derivative(i,l,k) += current_terms;
              }

          // Don't leave the parameters_vec perturbed
          *parameters_vec[l] = old_parameterl;
          *parameters_vec[k] = old_parameterk;
        }

      // We get (partial q / partial u) and
      // (partial R / partial u) from the user, but centrally
      // difference to get q_uk and R_uk terms:
      // (partial^2 q / partial u partial k)
      // q_uk*u'_l = (q_u(p+dp*e_k)*u'_l - q_u(p-dp*e_k)*u'_l)/(2*dp)
      // R_uk*z*u'_l = (R_u(p+dp*e_k)*z*u'_l - R_u(p-dp*e_k)*z*u'_l)/(2*dp)
      //
      // To avoid creating Nq temporary vectors, we add these
      // subterms to the sensitivities output one by one.
      //
      // FIXME: this is probably a bad order of operations for
      // controlling floating point error.

      *parameters_vec[k] = old_parameterk + delta_p;
      this->assembly(false, true);
      this->matrix->close();
      this->assemble_qoi_derivative(qoi_indices,
                                    /* include_liftfunc = */ true,
                                    /* apply_constraints = */ false);

      for (unsigned int l=0; l != Np; ++l)
        {
          this->matrix->vector_mult(*tempvec, this->get_sensitivity_solution(l));
          for (unsigned int i=0; i != Nq; ++i)
            if (qoi_indices.has_index(i))
              {
                this->get_adjoint_rhs(i).close();
                Number current_terms =
                  (this->get_adjoint_rhs(i).dot(this->get_sensitivity_solution(l)) -
                   tempvec->dot(this->get_adjoint_solution(i))) / (2.*delta_p);
                sensitivities.second_derivative(i,k,l) += current_terms;

                // We use the _uk terms twice; symmetry lets us reuse
                // these calculations for the _ul terms.

                sensitivities.second_derivative(i,l,k) += current_terms;
              }
        }

      *parameters_vec[k] = old_parameterk - delta_p;
      this->assembly(false, true);
      this->matrix->close();
      this->assemble_qoi_derivative(qoi_indices,
                                    /* include_liftfunc = */ true,
                                    /* apply_constraints = */ false);

      for (unsigned int l=0; l != Np; ++l)
        {
          this->matrix->vector_mult(*tempvec, this->get_sensitivity_solution(l));
          for (unsigned int i=0; i != Nq; ++i)
            if (qoi_indices.has_index(i))
              {
                this->get_adjoint_rhs(i).close();
                Number current_terms =
                  (-this->get_adjoint_rhs(i).dot(this->get_sensitivity_solution(l)) +
                   tempvec->dot(this->get_adjoint_solution(i))) / (2.*delta_p);
                sensitivities.second_derivative(i,k,l) += current_terms;

                // We use the _uk terms twice; symmetry lets us reuse
                // these calculations for the _ul terms.

                sensitivities.second_derivative(i,l,k) += current_terms;
              }
        }

      // Don't leave the parameter perturbed
      *parameters_vec[k] = old_parameterk;

      // Our last remaining terms are -R_uu(u,z)*u_k*u_l and
      // Q_uu(u)*u_k*u_l
      //
      // We take directional central finite differences of R_u and Q_u
      // to approximate these terms, e.g.:
      //
      // Q_uu(u)*u_k ~= (Q_u(u+dp*u_k) - Q_u(u-dp*u_k))/(2*dp)

      *this->solution = this->get_sensitivity_solution(k);
      *this->solution *= delta_p;
      *this->solution += *oldsolution;

      // We've modified solution, so we need to update before calling
      // assembly since assembly may only use current_local_solution
      this->update();
      this->assembly(false, true);
      this->matrix->close();
      this->assemble_qoi_derivative(qoi_indices,
                                    /* include_liftfunc = */ true,
                                    /* apply_constraints = */ false);

      // The Hessian is symmetric, so we just calculate the lower
      // triangle and the diagonal, and we get the upper triangle from
      // the transpose of the lower
      //
      // Note that, because we took the directional finite difference
      // with respect to k and not l, we've added an O(delta_p^2)
      // error to any permutational symmetry in the Hessian...
      for (unsigned int l=0; l != k+1; ++l)
        {
          this->matrix->vector_mult(*tempvec, this->get_sensitivity_solution(l));
          for (unsigned int i=0; i != Nq; ++i)
            if (qoi_indices.has_index(i))
              {
                this->get_adjoint_rhs(i).close();
                Number current_terms =
                  (this->get_adjoint_rhs(i).dot(this->get_sensitivity_solution(l)) -
                   tempvec->dot(this->get_adjoint_solution(i))) / (2.*delta_p);
                sensitivities.second_derivative(i,k,l) += current_terms;
                if (k != l)
                  sensitivities.second_derivative(i,l,k) += current_terms;
              }
        }

      *this->solution = this->get_sensitivity_solution(k);
      *this->solution *= -delta_p;
      *this->solution += *oldsolution;

      // We've modified solution, so we need to update before calling
      // assembly since assembly may only use current_local_solution
      this->update();
      this->assembly(false, true);
      this->matrix->close();
      this->assemble_qoi_derivative(qoi_indices,
                                    /* include_liftfunc = */ true,
                                    /* apply_constraints = */ false);

      for (unsigned int l=0; l != k+1; ++l)
        {
          this->matrix->vector_mult(*tempvec, this->get_sensitivity_solution(l));
          for (unsigned int i=0; i != Nq; ++i)
            if (qoi_indices.has_index(i))
              {
                this->get_adjoint_rhs(i).close();
                Number current_terms =
                  (-this->get_adjoint_rhs(i).dot(this->get_sensitivity_solution(l)) +
                   tempvec->dot(this->get_adjoint_solution(i))) / (2.*delta_p);
                sensitivities.second_derivative(i,k,l) += current_terms;
                if (k != l)
                  sensitivities.second_derivative(i,l,k) += current_terms;
              }
        }

      // Don't leave the solution perturbed
      *this->solution = *oldsolution;
    }

  // All parameters_vec have been reset.
  // Don't leave the qoi or system changed - principle of least
  // surprise.
  // We've modified solution, so we need to update before calling
  // assembly since assembly may only use current_local_solution
  this->update();
  this->assembly(true, true);
  this->rhs->close();
  this->matrix->close();
  this->assemble_qoi(qoi_indices);
}

qoi_parameter_hessian_vector_product()

void libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product ( const QoISet &  qoi_indices, const ParameterVector &  parameters, const ParameterVector &  vector, SensitivityData &  product  )

overridevirtualinherited

For each of the system's quantities of interest q in qoi[qoi_indices], and for a vector of parameters p, the parameter sensitivity Hessian H_ij is defined as H_ij = (d^2 q)/(d p_i d p_j) The Hessian-vector product, for a vector v_k in parameter space, is S_j = H_jk v_k This product is the output of this method, where for each q_i, S_j is stored in sensitivities[i][j].

Reimplemented from libMesh::System.

Definition at line 717 of file implicit_system.C.

References libMesh::ImplicitSystem::adjoint_solve(), libMesh::SensitivityData::allocate_data(), libMesh::ExplicitSystem::assemble_qoi(), libMesh::ExplicitSystem::assemble_qoi_derivative(), libMesh::ImplicitSystem::assembly(), libMesh::NumericVector< T >::close(), libMesh::SparseMatrix< T >::close(), libMesh::ParameterVector::deep_copy(), libMesh::NumericVector< T >::dot(), libMesh::System::get_adjoint_rhs(), libMesh::System::get_adjoint_solution(), libMesh::System::get_qoi_value(), libMesh::System::get_qoi_values(), libMesh::System::get_weighted_sensitivity_adjoint_solution(), libMesh::System::get_weighted_sensitivity_solution(), libMesh::QoISet::has_index(), libMesh::System::is_adjoint_already_solved(), libMesh::ImplicitSystem::matrix, libMesh::System::n_qois(), libMesh::Real, libMesh::ExplicitSystem::rhs, libMesh::ParameterVector::size(), libMesh::System::solution, libMesh::TOLERANCE, libMesh::ParameterVector::value_copy(), libMesh::SparseMatrix< T >::vector_mult(), libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve(), and libMesh::ImplicitSystem::weighted_sensitivity_solve().

{
  // We currently get partial derivatives via finite differencing
  const Real delta_p = TOLERANCE;

  ParameterVector & parameters_vec =
    const_cast<ParameterVector &>(parameters_in);

  // We'll use a single temporary vector for matrix-vector-vector products
  std::unique_ptr<NumericVector<Number>> tempvec = this->solution->zero_clone();

  const unsigned int Np = cast_int<unsigned int>
    (parameters_vec.size());
  const unsigned int Nq = this->n_qois();

  // For each quantity of interest q, the parameter sensitivity
  // Hessian is defined as q''_{kl} = {d^2 q}/{d p_k d p_l}.
  // Given a vector of parameter perturbation weights w_l, this
  // function evaluates the hessian-vector product sum_l(q''_{kl}*w_l)
  //
  // We calculate it from values and partial derivatives of the
  // quantity of interest function Q, solution u, adjoint solution z,
  // parameter sensitivity adjoint solutions z^l, and residual R, as:
  //
  // sum_l(q''_{kl}*w_l) =
  // sum_l(w_l * Q''_{kl}) + Q''_{uk}(u)*(sum_l(w_l u'_l)) -
  // R'_k(u, sum_l(w_l*z^l)) - R'_{uk}(u,z)*(sum_l(w_l u'_l) -
  // sum_l(w_l*R''_{kl}(u,z))
  //
  // See the adjoints model document for more details.

  // We first do an adjoint solve to get z for each quantity of
  // interest
  // if we haven't already or dont have an initial condition for the adjoint
  if (!this->is_adjoint_already_solved())
    {
      this->adjoint_solve(qoi_indices);
    }

  // Get ready to fill in sensitivities:
  sensitivities.allocate_data(qoi_indices, *this, parameters_vec);

  // We can't solve for all the solution sensitivities u'_l or for all
  // of the parameter sensitivity adjoint solutions z^l without
  // requiring O(Nq*Np) linear solves.  So we'll solve directly for their
  // weighted sum - this is just O(Nq) solves.

  // First solve for sum_l(w_l u'_l).
  this->weighted_sensitivity_solve(parameters_vec, vector);

  // Then solve for sum_l(w_l z^l).
  this->weighted_sensitivity_adjoint_solve(parameters_vec, vector, qoi_indices);

  for (unsigned int k=0; k != Np; ++k)
    {
      // We approximate sum_l(w_l * Q''_{kl}) with a central
      // differencing perturbation:
      // sum_l(w_l * Q''_{kl}) ~=
      // (Q(p + dp*w_l*e_l + dp*e_k) - Q(p - dp*w_l*e_l + dp*e_k) -
      // Q(p + dp*w_l*e_l - dp*e_k) + Q(p - dp*w_l*e_l - dp*e_k))/(4*dp^2)

      // The sum(w_l*R''_kl) term requires the same sort of perturbation,
      // and so we subtract it in at the same time:
      // sum_l(w_l * R''_{kl}) ~=
      // (R(p + dp*w_l*e_l + dp*e_k) - R(p - dp*w_l*e_l + dp*e_k) -
      // R(p + dp*w_l*e_l - dp*e_k) + R(p - dp*w_l*e_l - dp*e_k))/(4*dp^2)

      ParameterVector oldparameters, parameterperturbation;
      parameters_vec.deep_copy(oldparameters);
      vector.deep_copy(parameterperturbation);
      parameterperturbation *= delta_p;
      parameters_vec += parameterperturbation;

      Number old_parameter = *parameters_vec[k];

      *parameters_vec[k] = old_parameter + delta_p;
      this->assemble_qoi(qoi_indices);
      this->assembly(true, false, true);
      this->rhs->close();
      std::vector<Number> partial2q_term = this->get_qoi_values();
      std::vector<Number> partial2R_term(this->n_qois());
      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          partial2R_term[i] = this->rhs->dot(this->get_adjoint_solution(i));

      *parameters_vec[k] = old_parameter - delta_p;
      this->assemble_qoi(qoi_indices);
      this->assembly(true, false, true);
      this->rhs->close();
      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          {
            partial2q_term[i] -= this->get_qoi_value(i);
            partial2R_term[i] -= this->rhs->dot(this->get_adjoint_solution(i));
          }

      oldparameters.value_copy(parameters_vec);
      parameterperturbation *= -1.0;
      parameters_vec += parameterperturbation;

      // Re-center old_parameter, which may be affected by vector
      old_parameter = *parameters_vec[k];

      *parameters_vec[k] = old_parameter + delta_p;
      this->assemble_qoi(qoi_indices);
      this->assembly(true, false, true);
      this->rhs->close();
      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          {
            partial2q_term[i] -= this->get_qoi_value(i);
            partial2R_term[i] -= this->rhs->dot(this->get_adjoint_solution(i));
          }

      *parameters_vec[k] = old_parameter - delta_p;
      this->assemble_qoi(qoi_indices);
      this->assembly(true, false, true);
      this->rhs->close();
      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          {
            partial2q_term[i] += this->get_qoi_value(i);
            partial2R_term[i] += this->rhs->dot(this->get_adjoint_solution(i));
          }

      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          {
            partial2q_term[i] /= (4. * delta_p * delta_p);
            partial2R_term[i] /= (4. * delta_p * delta_p);
          }

      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          sensitivities[i][k] = partial2q_term[i] - partial2R_term[i];

      // We get (partial q / partial u), R, and
      // (partial R / partial u) from the user, but centrally
      // difference to get q_uk, R_k, and R_uk terms:
      // (partial R / partial k)
      // R_k*sum(w_l*z^l) = (R(p+dp*e_k)*sum(w_l*z^l) - R(p-dp*e_k)*sum(w_l*z^l))/(2*dp)
      // (partial^2 q / partial u partial k)
      // q_uk = (q_u(p+dp*e_k) - q_u(p-dp*e_k))/(2*dp)
      // (partial^2 R / partial u partial k)
      // R_uk*z*sum(w_l*u'_l) = (R_u(p+dp*e_k)*z*sum(w_l*u'_l) - R_u(p-dp*e_k)*z*sum(w_l*u'_l))/(2*dp)

      // To avoid creating Nq temporary vectors for q_uk or R_uk, we add
      // subterms to the sensitivities output one by one.
      //
      // FIXME: this is probably a bad order of operations for
      // controlling floating point error.

      *parameters_vec[k] = old_parameter + delta_p;
      this->assembly(true, true);
      this->rhs->close();
      this->matrix->close();
      this->assemble_qoi_derivative(qoi_indices,
                                    /* include_liftfunc = */ true,
                                    /* apply_constraints = */ false);

      this->matrix->vector_mult(*tempvec, this->get_weighted_sensitivity_solution());

      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          {
            this->get_adjoint_rhs(i).close();
            sensitivities[i][k] += (this->get_adjoint_rhs(i).dot(this->get_weighted_sensitivity_solution()) -
                                    this->rhs->dot(this->get_weighted_sensitivity_adjoint_solution(i)) -
                                    this->get_adjoint_solution(i).dot(*tempvec)) / (2.*delta_p);
          }

      *parameters_vec[k] = old_parameter - delta_p;
      this->assembly(true, true);
      this->rhs->close();
      this->matrix->close();
      this->assemble_qoi_derivative(qoi_indices,
                                    /* include_liftfunc = */ true,
                                    /* apply_constraints = */ false);

      this->matrix->vector_mult(*tempvec, this->get_weighted_sensitivity_solution());

      for (unsigned int i=0; i != Nq; ++i)
        if (qoi_indices.has_index(i))
          {
            this->get_adjoint_rhs(i).close();
            sensitivities[i][k] += (-this->get_adjoint_rhs(i).dot(this->get_weighted_sensitivity_solution()) +
                                    this->rhs->dot(this->get_weighted_sensitivity_adjoint_solution(i)) +
                                    this->get_adjoint_solution(i).dot(*tempvec)) / (2.*delta_p);
          }
    }

  // All parameters have been reset.
  // Don't leave the qoi or system changed - principle of least
  // surprise.
  this->assembly(true, true);
  this->rhs->close();
  this->matrix->close();
  this->assemble_qoi(qoi_indices);
}

qoi_parameter_sensitivity()

void libMesh::System::qoi_parameter_sensitivity ( const QoISet &  qoi_indices, const ParameterVector &  parameters, SensitivityData &  sensitivities  )

virtualinherited

Solves for the derivative of each of the system's quantities of interest q in qoi[qoi_indices] with respect to each parameter in parameters, placing the result for qoi i and parameter j into sensitivities[i][j].

Note

parameters is a const vector, not a vector-of-const; parameter values in this vector need to be mutable for finite differencing to work.

Automatically chooses the forward method for problems with more quantities of interest than parameters, or the adjoint method otherwise.

This method is only usable in derived classes which override an implementation.

Definition at line 590 of file system.C.

References libMesh::System::adjoint_qoi_parameter_sensitivity(), libMesh::System::forward_qoi_parameter_sensitivity(), libMesh::ParameterVector::size(), and libMesh::QoISet::size().

{
  // Forward sensitivities are more efficient for Nq > Np
  if (qoi_indices.size(*this) > parameters_vec.size())
    forward_qoi_parameter_sensitivity(qoi_indices, parameters_vec, sensitivities);
  // Adjoint sensitivities are more efficient for Np > Nq,
  // and an adjoint may be more reusable than a forward
  // solution sensitivity in the Np == Nq case.
  else
    adjoint_qoi_parameter_sensitivity(qoi_indices, parameters_vec, sensitivities);
}

re_update()

void libMesh::System::re_update ( )

virtualinherited

Re-update the local values when the mesh has changed.

This method takes the data updated by update() and makes it up-to-date on the current mesh.

Reimplemented in libMesh::TransientSystem< RBConstruction >.

Definition at line 521 of file system.C.

References libMesh::System::current_local_solution, libMesh::System::get_dof_map(), libMesh::DofMap::get_send_list(), libMesh::System::n_vars(), and libMesh::System::solution.

{
  parallel_object_only();

  // If this system is empty... don't do anything!
  if (!this->n_vars())
    return;

  const std::vector<dof_id_type> & send_list = this->get_dof_map().get_send_list ();

  // Check sizes
  libmesh_assert_equal_to (current_local_solution->size(), solution->size());
  // Not true with ghosted vectors
  // libmesh_assert_equal_to (current_local_solution->local_size(), solution->size());
  // libmesh_assert (!send_list.empty());
  libmesh_assert_less_equal (send_list.size(), solution->size());

  // Create current_local_solution from solution.  This will
  // put a local copy of solution into current_local_solution.
  solution->localize (*current_local_solution, send_list);
}

read_header()

void libMesh::System::read_header ( Xdr &  io, std::string_view  version, const bool  read_header = true, const bool  read_additional_data = true, const bool  read_legacy_format = false  )

inherited

Reads the basic data header for this System.

Definition at line 97 of file system_io.C.

References libMesh::System::_additional_data_written, libMesh::System::_written_var_indices, libMesh::System::add_variable(), libMesh::System::add_vector(), TIMPI::Communicator::broadcast(), libMesh::System::clear(), libMesh::ParallelObject::comm(), libMesh::Utility::contains(), libMesh::Xdr::data(), libMesh::FEType::family, libMesh::System::get_mesh(), libMesh::OrderWrapper::get_order(), libMesh::FEType::inf_map, libMesh::libmesh_assert(), libMesh::MeshBase::mesh_dimension(), libMesh::MONOMIAL, libMesh::on_command_line(), libMesh::FEType::order, libMesh::out, libMesh::ParallelObject::processor_id(), libMesh::FEType::radial_family, libMesh::FEType::radial_order, libMesh::Xdr::reading(), libMesh::System::variable_number(), libMesh::Xdr::version(), and libMesh::XYZ.

Referenced by libMesh::EquationSystems::read(), and libMesh::RBEvaluation::read_in_vectors_from_multiple_files().

{
  // This method implements the input of a
  // System object, embedded in the output of
  // an EquationSystems<T_sys>.  This warrants some
  // documentation.  The output file essentially
  // consists of 5 sections:
  //
  // for this system
  //
  //   5.) The number of variables in the system (unsigned int)
  //
  //   for each variable in the system
  //
  //     6.) The name of the variable (string)
  //
  //     6.1.) Variable subdomains
  //
  //     7.) Combined in an FEType:
  //         - The approximation order(s) of the variable
  //           (Order Enum, cast to int/s)
  //         - The finite element family/ies of the variable
  //           (FEFamily Enum, cast to int/s)
  //
  //   end variable loop
  //
  //   8.) The number of additional vectors (unsigned int),
  //
  //     for each additional vector in the system object
  //
  //     9.) the name of the additional vector  (string)
  //
  // end system
  libmesh_assert (io.reading());

  // Possibly clear data structures and start from scratch.
  if (read_header_in)
    this->clear ();

  // Figure out if we need to read infinite element information.
  // This will be true if the version string contains " with infinite elements"
  const bool read_ifem_info =
    Utility::contains(version, " with infinite elements") ||
    libMesh::on_command_line ("--read-ifem-systems");


  {
    // 5.)
    // Read the number of variables in the system
    unsigned int nv=0;
    if (this->processor_id() == 0)
      io.data (nv);
    this->comm().broadcast(nv);

    _written_var_indices.clear();
    _written_var_indices.resize(nv, 0);

    for (unsigned int var=0; var<nv; var++)
      {
        // 6.)
        // Read the name of the var-th variable
        std::string var_name;
        if (this->processor_id() == 0)
          io.data (var_name);
        this->comm().broadcast(var_name);

        // 6.1.)
        std::set<subdomain_id_type> domains;
        if (io.version() >= LIBMESH_VERSION_ID(0,7,2))
          {
            std::vector<subdomain_id_type> domain_array;
            if (this->processor_id() == 0)
              io.data (domain_array);
            for (const auto & id : domain_array)
              domains.insert(id);
          }
        this->comm().broadcast(domains);

        // 7.)
        // Read the approximation order(s) of the var-th variable
        int order=0;
        if (this->processor_id() == 0)
          io.data (order);
        this->comm().broadcast(order);


        // do the same for infinite element radial_order
        int rad_order=0;
        if (read_ifem_info)
          {
            if (this->processor_id() == 0)
              io.data(rad_order);
            this->comm().broadcast(rad_order);
          }

        // Read the finite element type of the var-th variable
        int fam=0;
        if (this->processor_id() == 0)
          io.data (fam);
        this->comm().broadcast(fam);
        FEType type;
        type.order  = static_cast<Order>(order);
        type.family = static_cast<FEFamily>(fam);

        // Check for incompatibilities.  The shape function indexing was
        // changed for the monomial and xyz finite element families to
        // simplify extension to arbitrary p.  The consequence is that
        // old restart files will not be read correctly.  This is expected
        // to be an unlikely occurrence, but catch it anyway.
        if (read_legacy_format)
          if ((type.family == MONOMIAL || type.family == XYZ) &&
              ((type.order.get_order() > 2 && this->get_mesh().mesh_dimension() == 2) ||
               (type.order.get_order() > 1 && this->get_mesh().mesh_dimension() == 3)))
            {
              libmesh_here();
              libMesh::out << "*****************************************************************\n"
                           << "* WARNING: reading a potentially incompatible restart file!!!   *\n"
                           << "*  contact [email protected] for more details *\n"
                           << "*****************************************************************"
                           << std::endl;
            }

        // Read additional information for infinite elements
        int radial_fam=0;
        int i_map=0;
        if (read_ifem_info)
          {
            if (this->processor_id() == 0)
              io.data (radial_fam);
            this->comm().broadcast(radial_fam);
            if (this->processor_id() == 0)
              io.data (i_map);
            this->comm().broadcast(i_map);
          }

#ifdef LIBMESH_ENABLE_INFINITE_ELEMENTS

        type.radial_order  = static_cast<Order>(rad_order);
        type.radial_family = static_cast<FEFamily>(radial_fam);
        type.inf_map       = static_cast<InfMapType>(i_map);

#endif

        if (read_header_in)
          {
            if (domains.empty())
              _written_var_indices[var] = this->add_variable (var_name, type);
            else
              _written_var_indices[var] = this->add_variable (var_name, type, &domains);
          }
        else
          _written_var_indices[var] = this->variable_number(var_name);
      }
  }

  // 8.)
  // Read the number of additional vectors.
  unsigned int nvecs=0;
  if (this->processor_id() == 0)
    io.data (nvecs);
  this->comm().broadcast(nvecs);

  // If nvecs > 0, this means that write_additional_data
  // was true when this file was written.  We will need to
  // make use of this fact later.
  this->_additional_data_written = nvecs;

  for (unsigned int vec=0; vec<nvecs; vec++)
    {
      // 9.)
      // Read the name of the vec-th additional vector
      std::string vec_name;
      if (this->processor_id() == 0)
        io.data (vec_name);
      this->comm().broadcast(vec_name);
      if (io.version() >= LIBMESH_VERSION_ID(1,7,0))
        {
          int vec_projection = 0;
          if (this->processor_id() == 0)
            io.data (vec_projection);
          this->comm().broadcast(vec_projection);
          int vec_type;
          if (this->processor_id() == 0)
            io.data (vec_type);
          this->comm().broadcast(vec_type);

          if (read_additional_data)
            this->add_vector(vec_name, bool(vec_projection), ParallelType(vec_type));
        }
      else if (read_additional_data)
          // Systems now can handle adding post-initialization vectors
          //  libmesh_assert(this->_can_add_vectors);
          // Some systems may have added their own vectors already
          //  libmesh_assert_equal_to (this->_vectors.count(vec_name), 0);
        this->add_vector(vec_name);
    }
}

read_parallel_data() [1/2]

template<typename InValType >

void libMesh::System::read_parallel_data ( Xdr &  io, const bool  read_additional_data  )

inherited

Reads additional data, namely vectors, for this System.

This method may safely be called on a distributed-memory mesh. This method will read an individual file for each processor in the simulation where the local solution components for that processor are stored.

This method implements the output of the vectors contained in this System object, embedded in the output of an EquationSystems<T_sys>.

9.) The global solution vector, re-ordered to be node-major (More on this later.)

for each additional vector in the object

10.) The global additional vector, re-ordered to be node-major (More on this later.)

Note that the actual IO is handled through the Xdr class (to be renamed later?) which provides a uniform interface to both the XDR (eXternal Data Representation) interface and standard ASCII output. Thus this one section of code will read XDR or ASCII files with no changes.

Definition at line 302 of file system_io.C.

References libMesh::System::_additional_data_written, libMesh::System::_vectors, libMesh::System::_written_var_indices, libMesh::Xdr::data(), libMesh::FEType::family, libMesh::System::get_dof_map(), libMesh::System::get_mesh(), libMesh::DofObject::invalid_id, libMesh::Xdr::is_open(), libMesh::libmesh_assert(), libMesh::make_range(), libMesh::ParallelObject::n_processors(), libMesh::System::n_vars(), libMesh::System::number(), libMesh::ParallelObject::processor_id(), libMesh::Xdr::reading(), libMesh::SCALAR, libMesh::DofMap::SCALAR_dof_indices(), libMesh::System::solution, libMesh::Variable::type(), and libMesh::System::variable().

{
  // PerfLog pl("IO Performance",false);
  // pl.push("read_parallel_data");
  [[maybe_unused]] dof_id_type total_read_size = 0;

  libmesh_assert (io.reading());
  libmesh_assert (io.is_open());

  // build the ordered nodes and element maps.
  // when writing/reading parallel files we need to iterate
  // over our nodes/elements in order of increasing global id().
  // however, this is not guaranteed to be ordering we obtain
  // by using the node_iterators/element_iterators directly.
  // so build a set, sorted by id(), that provides the ordering.
  // further, for memory economy build the set but then transfer
  // its contents to vectors, which will be sorted.
  std::vector<const DofObject *> ordered_nodes, ordered_elements;
  {
    std::set<const DofObject *, CompareDofObjectsByID>
      ordered_nodes_set (this->get_mesh().local_nodes_begin(),
                         this->get_mesh().local_nodes_end());

    ordered_nodes.insert(ordered_nodes.end(),
                         ordered_nodes_set.begin(),
                         ordered_nodes_set.end());
  }
  {
    std::set<const DofObject *, CompareDofObjectsByID>
      ordered_elements_set (this->get_mesh().local_elements_begin(),
                            this->get_mesh().local_elements_end());

    ordered_elements.insert(ordered_elements.end(),
                            ordered_elements_set.begin(),
                            ordered_elements_set.end());
  }

  //  std::vector<Number> io_buffer;
  std::vector<InValType> io_buffer;

  // 9.)
  //
  // Actually read the solution components
  // for the ith system to disk
  io.data(io_buffer);

  total_read_size += cast_int<dof_id_type>(io_buffer.size());

  const unsigned int sys_num = this->number();
  const unsigned int nv      = cast_int<unsigned int>
    (this->_written_var_indices.size());
  libmesh_assert_less_equal (nv, this->n_vars());

  dof_id_type cnt=0;

  // Loop over each non-SCALAR variable and each node, and read out the value.
  for (unsigned int data_var=0; data_var<nv; data_var++)
    {
      const unsigned int var = _written_var_indices[data_var];
      if (this->variable(var).type().family != SCALAR)
        {
          // First read the node DOF values
          for (const auto & node : ordered_nodes)
            for (auto comp : make_range(node->n_comp(sys_num,var)))
              {
                libmesh_assert_not_equal_to (node->dof_number(sys_num, var, comp),
                                             DofObject::invalid_id);
                libmesh_assert_less (cnt, io_buffer.size());
                this->solution->set(node->dof_number(sys_num, var, comp), io_buffer[cnt++]);
              }

          // Then read the element DOF values
          for (const auto & elem : ordered_elements)
            for (auto comp : make_range(elem->n_comp(sys_num,var)))
              {
                libmesh_assert_not_equal_to (elem->dof_number(sys_num, var, comp),
                                             DofObject::invalid_id);
                libmesh_assert_less (cnt, io_buffer.size());
                this->solution->set(elem->dof_number(sys_num, var, comp), io_buffer[cnt++]);
              }
        }
    }

  // Finally, read the SCALAR variables on the last processor
  for (unsigned int data_var=0; data_var<nv; data_var++)
    {
      const unsigned int var = _written_var_indices[data_var];
      if (this->variable(var).type().family == SCALAR)
        {
          if (this->processor_id() == (this->n_processors()-1))
            {
              const DofMap & dof_map = this->get_dof_map();
              std::vector<dof_id_type> SCALAR_dofs;
              dof_map.SCALAR_dof_indices(SCALAR_dofs, var);

              for (auto dof : SCALAR_dofs)
                this->solution->set(dof, io_buffer[cnt++]);
            }
        }
    }

  // And we're done setting solution entries
  this->solution->close();

  // For each additional vector, simply go through the list.
  // ONLY attempt to do this IF additional data was actually
  // written to the file for this system (controlled by the
  // _additional_data_written flag).
  if (this->_additional_data_written)
    {
      const std::size_t nvecs = this->_vectors.size();

      // If the number of additional vectors written is non-zero, and
      // the number of additional vectors we have is non-zero, and
      // they don't match, then something is wrong and we can't be
      // sure we're reading data into the correct places.
      if (read_additional_data && nvecs &&
          nvecs != this->_additional_data_written)
        libmesh_error_msg
          ("Additional vectors in file do not match system");

      auto pos = _vectors.begin();

      for (std::size_t i = 0; i != this->_additional_data_written; ++i)
        {
          cnt=0;
          io_buffer.clear();

          // 10.)
          //
          // Actually read the additional vector components
          // for the ith system from disk
          io.data(io_buffer);

          total_read_size += cast_int<dof_id_type>(io_buffer.size());

          // If read_additional_data==true and we have additional vectors,
          // then we will keep this vector data; otherwise we are going to
          // throw it away.
          if (read_additional_data && nvecs)
            {
              // Loop over each non-SCALAR variable and each node, and read out the value.
              for (unsigned int data_var=0; data_var<nv; data_var++)
                {
                  const unsigned int var = _written_var_indices[data_var];
                  if (this->variable(var).type().family != SCALAR)
                    {
                      // First read the node DOF values
                      for (const auto & node : ordered_nodes)
                        for (auto comp : make_range(node->n_comp(sys_num,var)))
                          {
                            libmesh_assert_not_equal_to (node->dof_number(sys_num, var, comp),
                                                         DofObject::invalid_id);
                            libmesh_assert_less (cnt, io_buffer.size());
                            pos->second->set(node->dof_number(sys_num, var, comp), io_buffer[cnt++]);
                          }

                      // Then read the element DOF values
                      for (const auto & elem : ordered_elements)
                        for (auto comp : make_range(elem->n_comp(sys_num,var)))
                          {
                            libmesh_assert_not_equal_to (elem->dof_number(sys_num, var, comp),
                                                         DofObject::invalid_id);
                            libmesh_assert_less (cnt, io_buffer.size());
                            pos->second->set(elem->dof_number(sys_num, var, comp), io_buffer[cnt++]);
                          }
                    }
                }

              // Finally, read the SCALAR variables on the last processor
              for (unsigned int data_var=0; data_var<nv; data_var++)
                {
                  const unsigned int var = _written_var_indices[data_var];
                  if (this->variable(var).type().family == SCALAR)
                    {
                      if (this->processor_id() == (this->n_processors()-1))
                        {
                          const DofMap & dof_map = this->get_dof_map();
                          std::vector<dof_id_type> SCALAR_dofs;
                          dof_map.SCALAR_dof_indices(SCALAR_dofs, var);

                          for (auto dof : SCALAR_dofs)
                            pos->second->set(dof, io_buffer[cnt++]);
                        }
                    }
                }

              // And we're done setting entries for this variable
              pos->second->close();
            }

          // If we've got vectors then we need to be iterating through
          // those too
          if (pos != this->_vectors.end())
            ++pos;
        }
    }

  // const Real
  //   dt   = pl.get_elapsed_time(),
  //   rate = total_read_size*sizeof(Number)/dt;

  // libMesh::err << "Read " << total_read_size << " \"Number\" values\n"
  //     << " Elapsed time = " << dt << '\n'
  //     << " Rate = " << rate/1.e6 << "(MB/sec)\n\n";

  // pl.pop("read_parallel_data");
}

read_parallel_data() [2/2]

template LIBMESH_EXPORT void libMesh::System::read_parallel_data< Real > ( Xdr &  io, const bool  read_additional_data  )

inlineinherited

Non-templated version for backward compatibility.

Reads additional data, namely vectors, for this System. This method may safely be called on a distributed-memory mesh. This method will read an individual file for each processor in the simulation where the local solution components for that processor are stored.

Definition at line 1412 of file system.h.

  { read_parallel_data<Number>(io, read_additional_data); }

read_parameter_data_from_files()

void libMesh::RBParametrized::read_parameter_data_from_files ( const std::string &  continuous_param_file_name, const std::string &  discrete_param_file_name, const bool  read_binary_data  )

inherited

Read in the parameter ranges from files.

Definition at line 262 of file rb_parametrized.C.

References libMesh::RBParametrized::initialize_parameters(), libMesh::RBParametrized::read_discrete_parameter_values_from_file(), and libMesh::RBParametrized::read_parameter_ranges_from_file().

Referenced by libMesh::RBSCMEvaluation::legacy_read_offline_data_from_files(), and libMesh::RBEvaluation::legacy_read_offline_data_from_files().

{
  RBParameters param_min;
  RBParameters param_max;
  read_parameter_ranges_from_file(continuous_param_file_name,
                                  read_binary_data,
                                  param_min,
                                  param_max);

  std::map<std::string, std::vector<Real>> discrete_parameter_values_in;
  read_discrete_parameter_values_from_file(discrete_param_file_name,
                                           read_binary_data,
                                           discrete_parameter_values_in);

  initialize_parameters(param_min, param_max, discrete_parameter_values_in);
}

read_riesz_representors_from_files()

void libMesh::RBConstruction::read_riesz_representors_from_files ( const std::string &  riesz_representors_dir, const bool  write_binary_residual_representors  )

virtual

Read in all the Riesz representor data from files.

directory_name specifies which directory to read from. io_flag specifies whether we read in all data, or only a basis (in)dependent subset.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 2649 of file rb_construction.C.

References libMesh::RBEvaluation::Aq_representor, libMesh::NumericVector< T >::build(), libMesh::ParallelObject::comm(), libMesh::DECODE, Fq_representor, Fq_representor_innerprods_computed, libMesh::RBEvaluation::get_n_basis_functions(), get_rb_evaluation(), libMesh::index_range(), libMesh::System::n_dofs(), libMesh::System::n_local_dofs(), libMesh::out, libMesh::PARALLEL, libMesh::ParallelObject::processor_id(), libMesh::READ, libMesh::System::read_serialized_data(), and libMesh::System::solution.

{
  LOG_SCOPE("read_riesz_representors_from_files()", "RBConstruction");

  libMesh::out << "Reading in the Fq_representors..." << std::endl;

  const std::string riesz_representor_suffix = (read_binary_residual_representors ? ".xdr" : ".dat");
  std::ostringstream file_name;
  struct stat stat_info;

  // Read in the Fq_representors.  There should be Q_f of these.  FIXME:
  // should we be worried about leaks here?
  for (const auto & rep : Fq_representor)
    libmesh_error_msg_if(rep, "Error, must delete existing Fq_representor before reading in from file.");

  for (auto i : index_range(Fq_representor))
    {
      file_name.str(""); // reset filename
      file_name << riesz_representors_dir
                << "/Fq_representor" << i << riesz_representor_suffix;

      // On processor zero check to be sure the file exists
      if (this->processor_id() == 0)
        {
          int stat_result = stat(file_name.str().c_str(), &stat_info);

          libmesh_error_msg_if(stat_result != 0, "File does not exist: " << file_name.str());
        }

      Xdr fqr_data(file_name.str(),
                   read_binary_residual_representors ? DECODE : READ);

      read_serialized_data(fqr_data, false);

      Fq_representor[i] = NumericVector<Number>::build(this->comm());
      Fq_representor[i]->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);

      // No need to copy, just swap
      // *Fq_representor[i] = *solution;
      Fq_representor[i]->swap(*solution);
    }

  // Alert the update_residual_terms() function that we don't need to recompute
  // the Fq_representors as we have already read them in from file!
  Fq_representor_innerprods_computed = true;


  libMesh::out << "Reading in the Aq_representors..." << std::endl;

  // Read in the Aq representors.  The class makes room for [Q_a][Nmax] of these.  We are going to
  // read in [Q_a][get_rb_evaluation().get_n_basis_functions()].  FIXME:
  // should we be worried about leaks in the locations where we're about to fill entries?
  RBEvaluation & rbe = get_rb_evaluation();
  for (const auto & row : rbe.Aq_representor)
    for (const auto & rep : row)
      libmesh_error_msg_if(rep, "Error, must delete existing Aq_representor before reading in from file.");

  // Now ready to read them in from file!
  for (auto i : index_range(rbe.Aq_representor))
    for (unsigned int j=0; j<rbe.get_n_basis_functions(); ++j)
      {
        file_name.str(""); // reset filename
        file_name << riesz_representors_dir
                  << "/Aq_representor" << i << "_" << j << riesz_representor_suffix;

        // On processor zero check to be sure the file exists
        if (this->processor_id() == 0)
          {
            int stat_result = stat(file_name.str().c_str(), &stat_info);

            libmesh_error_msg_if(stat_result != 0, "File does not exist: " << file_name.str());
          }

        Xdr aqr_data(file_name.str(), read_binary_residual_representors ? DECODE : READ);

        read_serialized_data(aqr_data, false);

        rbe.Aq_representor[i][j] = NumericVector<Number>::build(this->comm());
        rbe.Aq_representor[i][j]->init (n_dofs(), n_local_dofs(),
                                            false, PARALLEL);

        // No need to copy, just swap
        rbe.Aq_representor[i][j]->swap(*solution);
      }
}

read_serialized_data() [1/2]

template<typename InValType >

void libMesh::System::read_serialized_data ( Xdr &  io, const bool  read_additional_data = true  )

inherited

Reads additional data, namely vectors, for this System.

This method may safely be called on a distributed-memory mesh.

Definition at line 533 of file system_io.C.

References libMesh::System::_additional_data_written, libMesh::System::_vectors, libMesh::ParallelObject::processor_id(), and libMesh::System::solution.

Referenced by libMesh::TransientRBConstruction::initialize_truth(), libMesh::TransientRBConstruction::read_riesz_representors_from_files(), and read_riesz_representors_from_files().

{
  // This method implements the input of the vectors
  // contained in this System object, embedded in the
  // output of an EquationSystems<T_sys>.
  //
  //   10.) The global solution vector, re-ordered to be node-major
  //       (More on this later.)
  //
  //      for each additional vector in the object
  //
  //      11.) The global additional vector, re-ordered to be
  //          node-major (More on this later.)
  parallel_object_only();
  std::string comment;

  // PerfLog pl("IO Performance",false);
  // pl.push("read_serialized_data");
  // std::size_t total_read_size = 0;

  // 10.)
  // Read the global solution vector
  {
    // total_read_size +=
    this->read_serialized_vector<InValType>(io, this->solution.get());

    // get the comment
    if (this->processor_id() == 0)
      io.comment (comment);
  }

  // 11.)
  // Only read additional vectors if data is available, and only use
  // that data to fill our vectors if the user requested it.
  if (this->_additional_data_written)
    {
      const std::size_t nvecs = this->_vectors.size();

      // If the number of additional vectors written is non-zero, and
      // the number of additional vectors we have is non-zero, and
      // they don't match, then we can't read additional vectors
      // and be sure we're reading data into the correct places.
      if (read_additional_data && nvecs &&
          nvecs != this->_additional_data_written)
        libmesh_error_msg
          ("Additional vectors in file do not match system");

      auto pos = _vectors.begin();

      for (std::size_t i = 0; i != this->_additional_data_written; ++i)
        {
          // Read data, but only put it into a vector if we've been
          // asked to and if we have a corresponding vector to read.

          // total_read_size +=
          this->read_serialized_vector<InValType>
            (io, (read_additional_data && nvecs) ? pos->second.get() : nullptr);

          // get the comment
          if (this->processor_id() == 0)
            io.comment (comment);


          // If we've got vectors then we need to be iterating through
          // those too
          if (pos != this->_vectors.end())
            ++pos;
        }
    }

  // const Real
  //   dt   = pl.get_elapsed_time(),
  //   rate = total_read_size*sizeof(Number)/dt;

  // libMesh::out << "Read " << total_read_size << " \"Number\" values\n"
  //     << " Elapsed time = " << dt << '\n'
  //     << " Rate = " << rate/1.e6 << "(MB/sec)\n\n";

  // pl.pop("read_serialized_data");
}

read_serialized_data() [2/2]

template LIBMESH_EXPORT void libMesh::System::read_serialized_data< Real > ( Xdr &  io, const bool  read_additional_data = true  )

inlineinherited

Non-templated version for backward compatibility.

Reads additional data, namely vectors, for this System. This method may safely be called on a distributed-memory mesh.

Definition at line 1370 of file system.h.

  { read_serialized_data<Number>(io, read_additional_data); }

read_serialized_vectors() [1/2]

template<typename InValType >

std::size_t libMesh::System::read_serialized_vectors ( Xdr &  io, const std::vector< NumericVector< Number > *> &  vectors  ) const

inherited

Read a number of identically distributed vectors.

This method allows for optimization for the multiple vector case by only communicating the metadata once.

Definition at line 2018 of file system_io.C.

References libMesh::Xdr::data(), libMesh::System::get_mesh(), libMesh::libmesh_assert(), libMesh::make_range(), libMesh::MeshTools::n_elem(), libMesh::MeshBase::n_elem(), n_nodes, libMesh::MeshBase::n_nodes(), libMesh::System::n_vars(), libMesh::ParallelObject::processor_id(), libMesh::System::read_SCALAR_dofs(), libMesh::System::read_serialized_blocked_dof_objects(), libMesh::Xdr::reading(), libMesh::SCALAR, and libMesh::System::variable().

Referenced by libMesh::RBEvaluation::read_in_vectors_from_multiple_files().

{
  parallel_object_only();

  // Error checking
  // #ifndef NDEBUG
  //   // In parallel we better be reading a parallel vector -- if not
  //   // we will not set all of its components below!!
  //   if (this->n_processors() > 1)
  //     {
  //       libmesh_assert (vec.type() == PARALLEL ||
  //       vec.type() == GHOSTED);
  //     }
  // #endif

  libmesh_assert (io.reading());

  if (this->processor_id() == 0)
    {
      // sizes
      unsigned int num_vecs=0;
      dof_id_type vector_length=0;

      // Get the number of vectors
      io.data(num_vecs);
      // Get the buffer size
      io.data(vector_length);

      libmesh_error_msg_if
        (num_vecs != vectors.size(),
         "Xdr file header declares " << num_vecs << " vectors, but we were asked to read " << vectors.size());

      if (num_vecs != 0)
        {
          libmesh_error_msg_if (vectors[0] == nullptr, "vectors[0] should not be null");
          libmesh_error_msg_if (vectors[0]->size() != vector_length, "Inconsistent vector sizes");
        }
    }

  // no need to actually communicate these.
  // this->comm().broadcast(num_vecs);
  // this->comm().broadcast(vector_length);

  // Cache these - they are not free!
  const dof_id_type
    n_nodes = this->get_mesh().n_nodes(),
    n_elem  = this->get_mesh().n_elem();

  std::size_t read_length = 0;

  //---------------------------------
  // Collect the values for all nodes
  read_length +=
    this->read_serialized_blocked_dof_objects (n_nodes,
                                               this->get_mesh().local_nodes_begin(),
                                               this->get_mesh().local_nodes_end(),
                                               InValType(),
                                               io,
                                               vectors);

  //------------------------------------
  // Collect the values for all elements
  read_length +=
    this->read_serialized_blocked_dof_objects (n_elem,
                                               this->get_mesh().local_elements_begin(),
                                               this->get_mesh().local_elements_end(),
                                               InValType(),
                                               io,
                                               vectors);

  //-------------------------------------------
  // Finally loop over all the SCALAR variables
  for (NumericVector<Number> * vec : vectors)
    for (auto var : make_range(this->n_vars()))
      if (this->variable(var).type().family == SCALAR)
        {
          libmesh_assert_not_equal_to (vec, 0);

          read_length +=
            this->read_SCALAR_dofs (var, io, vec);
        }

  //---------------------------------------
  // last step - must close all the vectors
  for (NumericVector<Number> * vec : vectors)
    {
      libmesh_assert_not_equal_to (vec, 0);
      vec->close();
    }

  return read_length;
}

read_serialized_vectors() [2/2]

template LIBMESH_EXPORT std::size_t libMesh::System::read_serialized_vectors< Real > ( Xdr &  io, const std::vector< NumericVector< Number > *> &  vectors  ) const

inlineinherited

Non-templated version for backward compatibility.

Read a number of identically distributed vectors. This method allows for optimization for the multiple vector case by only communicating the metadata once.

Definition at line 1390 of file system.h.

  { return read_serialized_vectors<Number>(io, vectors); }

recompute_all_residual_terms()

void libMesh::RBConstruction::recompute_all_residual_terms ( const bool  compute_inner_products = true )

virtual

This function computes all of the residual representors, can be useful when restarting a basis training computation.

If compute_inner_products is false, we just compute the residual Riesz representors, whereas if true, we also compute all the corresponding inner product terms.

Definition at line 1818 of file rb_construction.C.

References compute_Fq_representor_innerprods(), delta_N, Fq_representor_innerprods_computed, libMesh::RBEvaluation::get_n_basis_functions(), get_rb_evaluation(), and update_residual_terms().

Referenced by train_reduced_basis_with_POD().

{
  // Compute the basis independent terms
  Fq_representor_innerprods_computed = false;
  compute_Fq_representor_innerprods(compute_inner_products);

  // and all the basis dependent terms
  unsigned int saved_delta_N = delta_N;
  delta_N = get_rb_evaluation().get_n_basis_functions();

  update_residual_terms(compute_inner_products);

  delta_N = saved_delta_N;
}

reinit()

void libMesh::LinearImplicitSystem::reinit ( )

overridevirtualinherited

Reinitializes the member data fields associated with the system, so that, e.g., assemble() may be used.

Reimplemented from libMesh::System.

Reimplemented in libMesh::NewmarkSystem.

Definition at line 99 of file linear_implicit_system.C.

References libMesh::ImplicitSystem::linear_solver, and libMesh::System::reinit().

Referenced by fe_assembly().

{
  // re-initialize the linear solver interface
  linear_solver->clear();

  // initialize parent data
  Parent::reinit();
}

reinit_constraints()

void libMesh::System::reinit_constraints ( )

virtualinherited

Reinitializes the constraints for this system.

Also prepares the send_list, whether or not constraints have changed.

Definition at line 483 of file system.C.

References libMesh::System::_mesh, libMesh::DofMap::create_dof_constraints(), libMesh::System::get_dof_map(), libMesh::on_command_line(), libMesh::out, libMesh::DofMap::prepare_send_list(), libMesh::DofMap::print_dof_constraints(), libMesh::DofMap::process_constraints(), libMesh::System::time, and libMesh::System::user_constrain().

Referenced by libMesh::EquationSystems::allgather(), libMesh::System::init_data(), libMesh::PetscDMWrapper::init_petscdm(), and libMesh::EquationSystems::reinit_solutions().

{
  parallel_object_only();

#ifdef LIBMESH_ENABLE_CONSTRAINTS
  get_dof_map().create_dof_constraints(_mesh, this->time);
  user_constrain();
  get_dof_map().process_constraints(_mesh);
  if (libMesh::on_command_line ("--print-constraints"))
    get_dof_map().print_dof_constraints(libMesh::out);
#endif
  get_dof_map().prepare_send_list();
}

reinit_mesh()

void libMesh::System::reinit_mesh ( )

virtualinherited

Reinitializes the system with a new mesh.

Definition at line 289 of file system.C.

References libMesh::System::_basic_system_only, libMesh::System::init_data(), libMesh::System::n_vars(), and libMesh::System::user_initialization().

Referenced by libMesh::System::init(), libMesh::PetscPreconditioner< T >::set_hypre_ads_data(), and libMesh::PetscPreconditioner< T >::set_hypre_ams_data().

{
  parallel_object_only();

  // First initialize any required data:
  // either only the basic System data
  if (_basic_system_only)
    System::init_data();
  // or all the derived class' data too
  else
    this->init_data();

  // If no variables have been added to this system
  // don't do anything
  if (!this->n_vars())
    return;

  // Then call the user-provided initialization function
  this->user_initialization();

}

remove_matrix()

void libMesh::System::remove_matrix ( std::string_view  mat_name )

inherited

Removes the additional matrix mat_name from this system.

Definition at line 1076 of file system.C.

References libMesh::System::_matrices.

{
  parallel_object_only();  // Not strictly needed, but the only safe way to keep in sync

  if (const auto pos = _matrices.find(mat_name);
      pos != _matrices.end())
    _matrices.erase(pos); // erase()'d entries are destroyed
}

remove_vector()

void libMesh::System::remove_vector ( std::string_view  vec_name )

inherited

Removes the additional vector vec_name from this system.

Definition at line 861 of file system.C.

References libMesh::System::_vector_is_adjoint, libMesh::System::_vector_projections, libMesh::System::_vectors, and libMesh::libmesh_assert().

Referenced by libMesh::AdjointRefinementEstimator::estimate_error(), and libMesh::UnsteadySolver::integrate_adjoint_sensitivity().

{
  parallel_object_only();  // Not strictly needed, but the only safe way to keep in sync

  if (const auto pos = _vectors.find(vec_name);
      pos != _vectors.end())
    {
      _vectors.erase(pos);
      auto proj_it = _vector_projections.find(vec_name);
      libmesh_assert(proj_it != _vector_projections.end());
      _vector_projections.erase(proj_it);

      auto adj_it = _vector_is_adjoint.find(vec_name);
      libmesh_assert(adj_it != _vector_is_adjoint.end());
      _vector_is_adjoint.erase(adj_it);
    }
}

request_matrix() [1/2]

const SparseMatrix< Number > * libMesh::System::request_matrix ( std::string_view  mat_name ) const

inherited

Returns

A const pointer to this system's additional matrix named mat_name, or nullptr if no matrix by that name exists.

Definition at line 1087 of file system.C.

References libMesh::System::_matrices.

Referenced by libMesh::EigenSystem::has_matrix_A(), libMesh::EigenSystem::has_matrix_B(), libMesh::EigenSystem::has_precond_matrix(), libMesh::ImplicitSystem::sensitivity_solve(), libMesh::NewtonSolver::solve(), and libMesh::LinearImplicitSystem::solve().

{
  if (const auto pos = _matrices.find(mat_name);
      pos != _matrices.end())
    return pos->second.get();

  // Otherwise, mat_name does not exist
  return nullptr;
}

request_matrix() [2/2]

SparseMatrix< Number > * libMesh::System::request_matrix ( std::string_view  mat_name )

inherited

Returns

A writable pointer to this system's additional matrix named mat_name, or nullptr if no matrix by that name exists.

Definition at line 1099 of file system.C.

References libMesh::System::_matrices.

{
  if (auto pos = _matrices.find(mat_name);
      pos != _matrices.end())
    return pos->second.get();

  // Otherwise, mat_name does not exist
  return nullptr;
}

request_vector() [1/4]

const NumericVector< Number > * libMesh::System::request_vector ( std::string_view  vec_name ) const

inherited

Returns

A const pointer to the vector if this System has a vector associated with the given name, nullptr otherwise.

Definition at line 879 of file system.C.

References libMesh::System::_vectors, and libMesh::NumericVector< T >::get().

Referenced by libMesh::UniformRefinementEstimator::_estimate_error().

{
  if (const auto pos = _vectors.find(vec_name);
      pos != _vectors.end())
    return pos->second.get();

  // Otherwise, vec_name was not found
  return nullptr;
}

request_vector() [2/4]

NumericVector< Number > * libMesh::System::request_vector ( std::string_view  vec_name )

inherited

Returns

A pointer to the vector if this System has a vector associated with the given name, nullptr otherwise.

Definition at line 891 of file system.C.

References libMesh::System::_vectors, and libMesh::NumericVector< T >::get().

{
  if (auto pos = _vectors.find(vec_name);
      pos != _vectors.end())
    return pos->second.get();

  // Otherwise, vec_name was not found
  return nullptr;
}

request_vector() [3/4]

const NumericVector< Number > * libMesh::System::request_vector ( const unsigned int  vec_num ) const

inherited

Returns

A const pointer to this system's additional vector number vec_num (where the vectors are counted starting with 0), or nullptr if the system has no such vector.

Definition at line 903 of file system.C.

References libMesh::System::_vectors, and libMesh::System::vectors_begin().

{
  // If we don't have that many vectors, return nullptr
  if (vec_num >= _vectors.size())
    return nullptr;

  // Otherwise return a pointer to the vec_num'th vector
  auto it = vectors_begin();
  std::advance(it, vec_num);
  return it->second.get();
}

request_vector() [4/4]

NumericVector< Number > * libMesh::System::request_vector ( const unsigned int  vec_num )

inherited

Returns

A writable pointer to this system's additional vector number vec_num (where the vectors are counted starting with 0), or nullptr if the system has no such vector.

Definition at line 917 of file system.C.

References libMesh::System::_vectors, and libMesh::System::vectors_begin().

{
  // If we don't have that many vectors, return nullptr
  if (vec_num >= _vectors.size())
    return nullptr;

  // Otherwise return a pointer to the vec_num'th vector
  auto it = vectors_begin();
  std::advance(it, vec_num);
  return it->second.get();
}

reset_preevaluate_thetas_completed()

void libMesh::RBConstruction::reset_preevaluate_thetas_completed ( )

protected

Reset the _preevaluate_thetas_completed flag to false.

We can use this to force us to recalculate preevaluate thetas, in cases where that is necessary.

Definition at line 2865 of file rb_construction.C.

References _preevaluate_thetas_completed.

{
  _preevaluate_thetas_completed = false;
}

restrict_solve_to()

void libMesh::LinearImplicitSystem::restrict_solve_to ( const SystemSubset *  subset, const SubsetSolveMode  subset_solve_mode = SUBSET_ZERO  )

overridevirtualinherited

After calling this method, any solve will be limited to the given subset.

To disable this mode, call this method with subset being a nullptr.

Reimplemented from libMesh::System.

Definition at line 110 of file linear_implicit_system.C.

References libMesh::LinearImplicitSystem::_subset, libMesh::LinearImplicitSystem::_subset_solve_mode, and libMesh::SystemSubset::get_system().

Referenced by libMesh::LinearImplicitSystem::clear(), and main().

{
  _subset = subset;
  _subset_solve_mode = subset_solve_mode;

  if (subset != nullptr)
    libmesh_assert_equal_to (&subset->get_system(), this);
}

restrict_vectors()

void libMesh::System::restrict_vectors ( )

virtualinherited

Restrict vectors after the mesh has coarsened.

Definition at line 374 of file system.C.

References libMesh::System::_dof_map, libMesh::System::_solution_projection, libMesh::System::_vector_projections, libMesh::System::_vectors, libMesh::System::current_local_solution, libMesh::NumericVector< T >::get(), libMesh::GHOSTED, libMesh::System::n_dofs(), libMesh::System::n_local_dofs(), libMesh::PARALLEL, libMesh::System::project_vector(), libMesh::System::solution, and libMesh::System::vector_is_adjoint().

Referenced by libMesh::System::prolong_vectors(), and libMesh::EquationSystems::reinit_solutions().

{
  parallel_object_only();

#ifdef LIBMESH_ENABLE_AMR
  // Restrict the _vectors on the coarsened cells
  for (auto & [vec_name, vec] : _vectors)
    {
      NumericVector<Number> * v = vec.get();

      if (_vector_projections[vec_name])
        {
          this->project_vector (*v, this->vector_is_adjoint(vec_name));
        }
      else
        {
          const ParallelType type = vec->type();

          if (type == GHOSTED)
            {
#ifdef LIBMESH_ENABLE_GHOSTED
              vec->init (this->n_dofs(), this->n_local_dofs(),
                               _dof_map->get_send_list(), /*fast=*/false,
                               GHOSTED);
#else
              libmesh_error_msg("Cannot initialize ghosted vectors when they are not enabled.");
#endif
            }
          else
            vec->init (this->n_dofs(), this->n_local_dofs(), false, type);
        }
    }

  const std::vector<dof_id_type> & send_list = _dof_map->get_send_list ();

  // Restrict the solution on the coarsened cells
  if (_solution_projection)
    this->project_vector (*solution);
  // Or at least make sure the solution vector is the correct size
  else
    solution->init (this->n_dofs(), this->n_local_dofs(), true, PARALLEL);

#ifdef LIBMESH_ENABLE_GHOSTED
  current_local_solution->init(this->n_dofs(),
                               this->n_local_dofs(), send_list,
                               false, GHOSTED);
#else
  current_local_solution->init(this->n_dofs());
#endif

  if (_solution_projection)
    solution->localize (*current_local_solution, send_list);

#endif // LIBMESH_ENABLE_AMR
}

sensitivity_solve()

std::pair< unsigned int, Real > libMesh::ImplicitSystem::sensitivity_solve ( const ParameterVector &  parameters )

overridevirtualinherited

Assembles & solves the linear system(s) (dR/du)*u_p = -dR/dp, for those parameters contained within parameters.

Returns

A pair with the total number of linear iterations performed and the (sum of the) final residual norms

Reimplemented from libMesh::System.

Definition at line 140 of file implicit_system.C.

References libMesh::System::add_sensitivity_solution(), libMesh::System::assemble_before_solve, libMesh::ImplicitSystem::assemble_residual_derivatives(), libMesh::ImplicitSystem::assembly(), libMesh::SparseMatrix< T >::close(), libMesh::DofMap::enforce_constraints_exactly(), libMesh::System::get_dof_map(), libMesh::ImplicitSystem::get_linear_solve_parameters(), libMesh::ImplicitSystem::get_linear_solver(), libMesh::System::get_sensitivity_rhs(), libMesh::System::get_sensitivity_solution(), libMesh::make_range(), libMesh::ImplicitSystem::matrix, libMesh::System::request_matrix(), libMesh::ParameterVector::size(), and libMesh::LinearSolver< T >::solve().

Referenced by libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity(), and libMesh::ImplicitSystem::qoi_parameter_hessian().

{
  // Log how long the linear solve takes.
  LOG_SCOPE("sensitivity_solve()", "ImplicitSystem");

  // The forward system should now already be solved.
  // Now assemble the corresponding sensitivity system.

  if (this->assemble_before_solve)
    {
      // Build the Jacobian
      this->assembly(false, true);
      this->matrix->close();

      // Reset and build the RHS from the residual derivatives
      this->assemble_residual_derivatives(parameters_vec);
    }

  // The sensitivity problem is linear
  LinearSolver<Number> * solver = this->get_linear_solver();

  // Our iteration counts and residuals will be sums of the individual
  // results
  std::pair<unsigned int, Real> solver_params =
    this->get_linear_solve_parameters();
  std::pair<unsigned int, Real> totalrval = std::make_pair(0,0.0);

  // Solve the linear system.
  SparseMatrix<Number> * pc = this->request_matrix("Preconditioner");
  for (auto p : make_range(parameters_vec.size()))
    {
      std::pair<unsigned int, Real> rval =
        solver->solve (*matrix, pc,
                       this->add_sensitivity_solution(p),
                       this->get_sensitivity_rhs(p),
                       double(solver_params.second),
                       solver_params.first);

      totalrval.first  += rval.first;
      totalrval.second += rval.second;
    }

  // The linear solver may not have fit our constraints exactly
#ifdef LIBMESH_ENABLE_CONSTRAINTS
  for (auto p : make_range(parameters_vec.size()))
    this->get_dof_map().enforce_constraints_exactly
      (*this, &this->get_sensitivity_solution(p),
       /* homogeneous = */ true);
#endif

  return totalrval;
}

set_abs_training_tolerance()

void libMesh::RBConstruction::set_abs_training_tolerance ( Real  new_training_tolerance )

inline

Get/set the absolute tolerance for the basis training.

Definition at line 225 of file rb_construction.h.

References abs_training_tolerance.

Referenced by set_rb_construction_parameters().

  {this->abs_training_tolerance = new_training_tolerance; }

set_adjoint_already_solved()

void libMesh::System::set_adjoint_already_solved ( bool  setting )

inlineinherited

Setter for the adjoint_already_solved boolean.

Definition at line 417 of file system.h.

References libMesh::System::adjoint_already_solved.

Referenced by main().

  { adjoint_already_solved = setting;}

set_basic_system_only()

void libMesh::System::set_basic_system_only ( )

inlineinherited

Sets the system to be "basic only": i.e.

advanced system components such as ImplicitSystem matrices may not be initialized. This is useful for efficiency in certain utility programs that never use System::solve(). This method must be called after the System or derived class is created but before it is initialized; e.g. from within EquationSystems::read()

Definition at line 2465 of file system.h.

References libMesh::System::_basic_system_only.

Referenced by libMesh::EquationSystems::read().

{
  _basic_system_only = true;
}

set_context_solution_vec()

void libMesh::RBConstruction::set_context_solution_vec ( NumericVector< Number > &  vec )

protectedvirtual

Set current_local_solution = vec so that we can access vec from FEMContext during assembly.

Override in subclasses if different behavior is required.

Definition at line 962 of file rb_construction.C.

References libMesh::System::current_local_solution, libMesh::System::get_dof_map(), and libMesh::NumericVector< T >::localize().

{
  // Set current_local_solution = vec so that we can access
  // vec from DGFEMContext during assembly
  vec.localize
    (*current_local_solution, this->get_dof_map().get_send_list());
}

set_convergence_assertion_flag()

void libMesh::RBConstruction::set_convergence_assertion_flag

(

bool

flag

)

Setter for the flag determining if convergence should be checked after each solve.

Definition at line 2750 of file rb_construction.C.

References assert_convergence.

{
  assert_convergence = flag;
}

set_current_training_parameter_index()

void libMesh::RBConstruction::set_current_training_parameter_index ( unsigned int  index )

protected

Definition at line 2770 of file rb_construction.C.

References _current_training_parameter_index.

Referenced by compute_max_error_bound().

{
  _current_training_parameter_index = index;
}

set_deterministic_training_parameter_name()

void libMesh::RBConstructionBase< LinearImplicitSystem >::set_deterministic_training_parameter_name ( const std::string &  name )

inherited

In some cases we only want to allow discrete parameter values, instead of parameters that may take any value in a specified interval.

Here we provide a method to set the d Set the discrete values for parameter mu that are allowed in the training set. This must be called before the training set is generated. Set the name of the parameter that we will generate deterministic training parameters for. Defaults to "NONE".

set_energy_inner_product()

void libMesh::RBConstruction::set_energy_inner_product

(

const std::vector< Number > &

energy_inner_product_coeffs_in

)

Specify the coefficients of the A_q operators to be used in the energy inner-product.

Definition at line 438 of file rb_construction.C.

References energy_inner_product_coeffs, and use_energy_inner_product.

{
  use_energy_inner_product = true;
  energy_inner_product_coeffs = energy_inner_product_coeffs_in;
}

set_inner_product_assembly()

void libMesh::RBConstruction::set_inner_product_assembly

(

ElemAssembly &

inner_product_assembly_in

)

Set the rb_assembly_expansion object.

Definition at line 421 of file rb_construction.C.

References inner_product_assembly, and use_energy_inner_product.

Referenced by SimpleRBConstruction::init_data().

{
  use_energy_inner_product = false;
  inner_product_assembly = &inner_product_assembly_in;
}

set_Nmax()

void libMesh::RBConstruction::set_Nmax ( unsigned int  Nmax )

virtual

Definition at line 1703 of file rb_construction.C.

References Nmax.

Referenced by set_rb_construction_parameters().

{
  this->Nmax = Nmax_in;
}

set_normalize_rb_bound_in_greedy()

void libMesh::RBConstruction::set_normalize_rb_bound_in_greedy ( bool  normalize_rb_bound_in_greedy_in )

inline

Get/set the boolean to indicate if we normalize the RB error in the greedy.

Definition at line 232 of file rb_construction.h.

References normalize_rb_bound_in_greedy.

Referenced by set_rb_construction_parameters().

  {this->normalize_rb_bound_in_greedy = normalize_rb_bound_in_greedy_in; }

set_normalize_solution_snapshots()

void libMesh::RBConstructionBase< LinearImplicitSystem >::set_normalize_solution_snapshots ( bool  value )

inherited

Set the boolean option that indicates if we normalization solution snapshots or not.

Definition at line 147 of file rb_construction_base.C.

{
  _normalize_solution_snapshots = value;
}

set_parameters()

bool libMesh::RBParametrized::set_parameters ( const RBParameters &  params )

inherited

Set the current parameters to params The parameters are checked for validity; an error is thrown if the number of parameters or samples is different than expected.

We

Returns

a boolean true if the new parameters are within the min/max range, and false otherwise (but the parameters are set regardless). Enabling the "verbose_mode" flag will also print more details.

Definition at line 129 of file rb_parametrized.C.

References libMesh::RBParametrized::check_if_valid_params(), libMesh::RBParametrized::parameters, and libMesh::RBParametrized::parameters_initialized.

Referenced by libMesh::RBSCMConstruction::compute_SCM_bounds_on_training_set(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_interiors(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_nodes(), libMesh::RBEIMConstruction::enrich_eim_approximation_on_sides(), get_RB_error_bound(), SimpleRBEvaluation::get_stability_lower_bound(), libMesh::RBParametrized::initialize_parameters(), libMesh::RBSCMEvaluation::reload_current_parameters(), libMesh::RBSCMEvaluation::set_current_parameters_from_C_J(), and RBParametersTest::testRBParametrized().

{
  libmesh_error_msg_if(!parameters_initialized, "Error: parameters not initialized in RBParametrized::set_parameters");

  // Terminate if params has the wrong number of parameters or samples.
  // If the parameters are outside the min/max range, return false.
  const bool valid_params = check_if_valid_params(params);

  // Make a copy of params (default assignment operator just does memberwise copy, which is sufficient here)
  this->parameters = params;

  return valid_params;
}

set_params_from_training_set()

void libMesh::RBConstructionBase< LinearImplicitSystem >::set_params_from_training_set ( unsigned int  global_index )

protectedinherited

Set parameters to the RBParameters stored in index global_index of the global training set.

Definition at line 225 of file rb_construction_base.C.

Referenced by compute_max_error_bound(), preevaluate_thetas(), and train_reduced_basis_with_POD().

{
  set_parameters(get_params_from_training_set(global_index));
}

set_params_from_training_set_and_broadcast()

void libMesh::RBConstructionBase< LinearImplicitSystem >::set_params_from_training_set_and_broadcast ( unsigned int  global_index )

protectedvirtualinherited

Load the specified training parameter and then broadcast to all processors.

Definition at line 270 of file rb_construction_base.C.

{
  libmesh_error_msg_if(!_training_parameters_initialized,
                       "Error: training parameters must first be initialized.");

  processor_id_type root_id = 0;
  if ((this->get_first_local_training_index() <= global_index) &&
      (global_index < this->get_last_local_training_index()))
    {
      // Set parameters on only one processor
      set_params_from_training_set(global_index);

      // set root_id, only non-zero on one processor
      root_id = this->processor_id();
    }

  // broadcast
  this->comm().max(root_id);
  broadcast_parameters(root_id);
}

set_preevaluate_thetas_flag()

void libMesh::RBConstruction::set_preevaluate_thetas_flag

(

bool

flag

)

Definition at line 2760 of file rb_construction.C.

References _preevaluate_thetas_flag.

{
  _preevaluate_thetas_flag = flag;
}

set_project_with_constraints()

void libMesh::System::set_project_with_constraints ( bool  _project_with_constraints )

inlineinherited

Definition at line 1842 of file system.h.

References libMesh::System::project_with_constraints.

Referenced by libMesh::AdjointRefinementEstimator::estimate_error().

  {
    project_with_constraints = _project_with_constraints;
  }

set_qoi() [1/2]

void libMesh::System::set_qoi ( unsigned int  qoi_index, Number  qoi_value  )

inherited

Definition at line 2171 of file system.C.

References libMesh::System::_qoi, and libMesh::libmesh_assert().

Referenced by libMesh::ExplicitSystem::assemble_qoi(), libMesh::FEMSystem::assemble_qoi(), libMesh::Euler2Solver::integrate_qoi_timestep(), libMesh::TwostepTimeSolver::integrate_qoi_timestep(), and libMesh::EulerSolver::integrate_qoi_timestep().

{
  libmesh_assert(qoi_index < _qoi.size());

  _qoi[qoi_index] = qoi_value;
}

set_qoi() [2/2]

void libMesh::System::set_qoi ( std::vector< Number >  new_qoi )

inherited

Definition at line 2192 of file system.C.

References libMesh::System::_qoi.

{
  libmesh_assert_equal_to(this->_qoi.size(), new_qoi.size());
  this->_qoi = std::move(new_qoi);
}

set_qoi_error_estimate()

void libMesh::System::set_qoi_error_estimate ( unsigned int  qoi_index, Number  qoi_error_estimate  )

inherited

Definition at line 2199 of file system.C.

References libMesh::System::_qoi_error_estimates, and libMesh::libmesh_assert().

Referenced by libMesh::Euler2Solver::integrate_adjoint_refinement_error_estimate(), libMesh::TwostepTimeSolver::integrate_adjoint_refinement_error_estimate(), and libMesh::EulerSolver::integrate_adjoint_refinement_error_estimate().

{
  libmesh_assert(qoi_index < _qoi_error_estimates.size());

  _qoi_error_estimates[qoi_index] = qoi_error_estimate;
}

set_quiet_mode()

void libMesh::RBConstructionBase< LinearImplicitSystem >::set_quiet_mode ( bool  quiet_mode_in )

inlineinherited

Set the quiet_mode flag.

If quiet == false then we print out a lot of extra information during the Offline stage.

Definition at line 100 of file rb_construction_base.h.

Referenced by set_rb_construction_parameters().

  { this->quiet_mode = quiet_mode_in; }

set_rb_assembly_expansion()

void libMesh::RBConstruction::set_rb_assembly_expansion

(

RBAssemblyExpansion &

rb_assembly_expansion_in

)

Set the rb_assembly_expansion object.

Definition at line 409 of file rb_construction.C.

References rb_assembly_expansion.

Referenced by SimpleRBConstruction::init_data().

{
  rb_assembly_expansion = &rb_assembly_expansion_in;
}

set_rb_construction_parameters()

void libMesh::RBConstruction::set_rb_construction_parameters	(	unsigned int	n_training_samples_in,
		bool	deterministic_training_in,
		int	training_parameters_random_seed_in,
		bool	quiet_mode_in,
		unsigned int	Nmax_in,
		Real	rel_training_tolerance_in,
		Real	abs_training_tolerance_in,
		bool	normalize_rb_error_bound_in_greedy_in,
		const std::string &	RB_training_type_in,
		const RBParameters &	mu_min_in,
		const RBParameters &	mu_max_in,
		const std::map< std::string, std::vector< Real >> &	discrete_parameter_values_in,
		const std::map< std::string, bool > &	log_scaling,
		std::map< std::string, std::vector< RBParameter >> *	training_sample_list = nullptr
	)

Set the state of this RBConstruction object based on the arguments to this function.

Definition at line 290 of file rb_construction.C.

References libMesh::RBParametrized::get_parameters_max(), libMesh::RBParametrized::get_parameters_min(), libMesh::RBParametrized::initialize_parameters(), libMesh::RBConstructionBase< LinearImplicitSystem >::initialize_training_parameters(), libMesh::RBConstructionBase< LinearImplicitSystem >::load_training_set(), set_abs_training_tolerance(), set_Nmax(), set_normalize_rb_bound_in_greedy(), libMesh::RBConstructionBase< LinearImplicitSystem >::set_quiet_mode(), set_RB_training_type(), set_rel_training_tolerance(), and libMesh::RBConstructionBase< LinearImplicitSystem >::set_training_random_seed().

Referenced by process_parameters_file().

{
  // Read in training_parameters_random_seed value.  This is used to
  // seed the RNG when picking the training parameters.  By default the
  // value is -1, which means use std::time to seed the RNG.
  set_training_random_seed(training_parameters_random_seed_in);

  // Set quiet mode
  set_quiet_mode(quiet_mode_in);

  // Initialize RB parameters
  set_Nmax(Nmax_in);

  set_rel_training_tolerance(rel_training_tolerance_in);
  set_abs_training_tolerance(abs_training_tolerance_in);

  set_normalize_rb_bound_in_greedy(normalize_rb_bound_in_greedy_in);

  set_RB_training_type(RB_training_type_in);

  // Initialize the parameter ranges and the parameters themselves
  initialize_parameters(mu_min_in, mu_max_in, discrete_parameter_values_in);

  bool updated_deterministic_training = deterministic_training_in;
  if (training_sample_list && (this->get_parameters_min().n_parameters() > 3))
    {
      // In this case we force deterministic_training to be false because
      // a) deterministic training samples are not currrently supported with
      //    more than 3 parameters, and
      // b) we will overwrite the training samples anyway in the call to
      //    load_training_set() below, so we do not want to generate an
      //    error due to deterministic training sample generation when
      //    the samples will be overwritten anyway.
      updated_deterministic_training = false;
    }

  initialize_training_parameters(this->get_parameters_min(),
                                 this->get_parameters_max(),
                                 n_training_samples_in,
                                 log_scaling_in,
                                 updated_deterministic_training); // use deterministic parameters

  if (training_sample_list)
    {
      // Note that we must call initialize_training_parameters() before
      // load_training_set() in order to initialize the parameter vectors.
      load_training_set(*training_sample_list);
    }
}

set_rb_evaluation()

void libMesh::RBConstruction::set_rb_evaluation

(

RBEvaluation &

rb_eval_in

)

Set the RBEvaluation object.

Definition at line 174 of file rb_construction.C.

References rb_eval.

Referenced by main().

{
  rb_eval = &rb_eval_in;
}

set_RB_training_type()

void libMesh::RBConstruction::set_RB_training_type

(

const std::string &

RB_training_type_in

)

Get/set the string that determines the training type.

Definition at line 1686 of file rb_construction.C.

References is_serial_training_type(), RB_training_type, and libMesh::RBConstructionBase< LinearImplicitSystem >::serial_training_set.

Referenced by set_rb_construction_parameters().

{
  this->RB_training_type = RB_training_type_in;

  if(is_serial_training_type(RB_training_type_in))
    {
      // We need to use a serial training set (so that the training
      // set is the same on all processes) if we're using POD
      this->serial_training_set = true;
    }
}

set_rel_training_tolerance()

void libMesh::RBConstruction::set_rel_training_tolerance ( Real  new_training_tolerance )

inline

Get/set the relative tolerance for the basis training.

Definition at line 218 of file rb_construction.h.

References rel_training_tolerance.

Referenced by set_rb_construction_parameters().

  {this->rel_training_tolerance = new_training_tolerance; }

set_training_parameter_values()

void libMesh::RBConstructionBase< LinearImplicitSystem >::set_training_parameter_values ( const std::string &  param_name, const std::vector< RBParameter > &  values  )

inherited

Overwrite the local training samples for param_name using values.

This assumes that values.size() matches get_local_n_training_samples().

Definition at line 456 of file rb_construction_base.C.

{
  libmesh_error_msg_if(!_training_parameters_initialized,
    "Training parameters must be initialized before calling set_training_parameter_values");
  libmesh_error_msg_if(values.size() != get_local_n_training_samples(),
    "Inconsistent sizes");

  // Copy the new data, overwriting the old data.
  auto & training_vector = libmesh_map_find(_training_parameters, param_name);
  training_vector = values;
}

set_training_random_seed()

void libMesh::RBConstructionBase< LinearImplicitSystem >::set_training_random_seed ( int  seed )

inherited

Set the seed that is used to randomly generate training parameters.

Definition at line 780 of file rb_construction_base.C.

Referenced by set_rb_construction_parameters().

{
  this->_training_parameters_random_seed = seed;
}

set_vector_as_adjoint()

void libMesh::System::set_vector_as_adjoint ( const std::string &  vec_name, int  qoi_num  )

inherited

Allows one to set the QoI index controlling whether the vector identified by vec_name represents a solution from the adjoint (qoi_num >= 0) or primal (qoi_num == -1) space.

This becomes significant if those spaces have differing heterogeneous Dirichlet constraints.

qoi_num == -2 can be used to indicate a vector which should not be affected by constraints during projection operations.

Definition at line 1147 of file system.C.

References libMesh::System::_vector_is_adjoint.

Referenced by libMesh::System::add_adjoint_solution(), and libMesh::System::add_weighted_sensitivity_adjoint_solution().

{
  parallel_object_only();  // Not strictly needed, but the only safe way to keep in sync

  // We reserve -1 for vectors which get primal constraints, -2 for
  // vectors which get no constraints
  libmesh_assert_greater_equal(qoi_num, -2);
  _vector_is_adjoint[vec_name] = qoi_num;
}

set_vector_preservation()

void libMesh::System::set_vector_preservation ( const std::string &  vec_name, bool  preserve  )

inherited

Allows one to set the boolean controlling whether the vector identified by vec_name should be "preserved": projected to new meshes, saved, etc.

Definition at line 1125 of file system.C.

References libMesh::System::_vector_projections.

Referenced by libMesh::AdjointRefinementEstimator::estimate_error(), and main().

{
  parallel_object_only();  // Not strictly needed, but the only safe way to keep in sync

  _vector_projections[vec_name] = preserve;
}

setup_static_condensation_preconditioner()

template<typename T >

template void libMesh::ImplicitSystem::setup_static_condensation_preconditioner ( T &  solver )

protectedinherited

Sets up the static condensation preconditioner for the supplied solver.

Definition at line 71 of file implicit_system.C.

References libMesh::ImplicitSystem::_sc_system_matrix, libMesh::StaticCondensation::get_preconditioner(), and libMesh::libmesh_assert().

Referenced by libMesh::LinearImplicitSystem::clear(), libMesh::NonlinearImplicitSystem::clear(), libMesh::LinearImplicitSystem::create_static_condensation(), libMesh::NonlinearImplicitSystem::create_static_condensation(), libMesh::LinearImplicitSystem::LinearImplicitSystem(), and libMesh::NonlinearImplicitSystem::NonlinearImplicitSystem().

{
  libmesh_assert(_sc_system_matrix);
  solver.attach_preconditioner(&_sc_system_matrix->get_preconditioner());
}

solve()

void libMesh::LinearImplicitSystem::solve ( )

overridevirtualinherited

Assembles & solves the linear system A*x=b.

Reimplemented from libMesh::ImplicitSystem.

Reimplemented in libMesh::FrequencySystem.

Definition at line 122 of file linear_implicit_system.C.

References libMesh::LinearImplicitSystem::_final_linear_residual, libMesh::LinearImplicitSystem::_n_linear_iterations, libMesh::LinearImplicitSystem::_shell_matrix, libMesh::LinearImplicitSystem::_subset, libMesh::LinearImplicitSystem::_subset_solve_mode, libMesh::LinearImplicitSystem::assemble(), libMesh::System::assemble_before_solve, libMesh::SystemSubset::dof_ids(), libMesh::ImplicitSystem::get_linear_solve_parameters(), libMesh::ImplicitSystem::linear_solver, libMesh::ImplicitSystem::matrix, libMesh::System::prefix(), libMesh::System::prefix_with_name(), libMesh::System::request_matrix(), libMesh::ExplicitSystem::rhs, libMesh::System::solution, and libMesh::System::update().

Referenced by assemble_and_solve(), main(), libMesh::ClawSystem::solve_conservation_law(), SystemsTest::testDofCouplingWithVarGroups(), PeriodicBCTest::testPeriodicBC(), DisjointNeighborTest::testTempJump(), and DisjointNeighborTest::testTempJumpRefine().

{
  if (this->assemble_before_solve)
    // Assemble the linear system
    this->assemble ();

  // If the linear solver hasn't been initialized, we do so here.
  if (this->prefix_with_name())
    linear_solver->init(this->prefix().c_str());
  else
    linear_solver->init();

  linear_solver->init_systems(*this);

  // Get the user-specified linear solver tolerance
  const auto [maxits, tol] = this->get_linear_solve_parameters();

  if (_subset != nullptr)
    linear_solver->restrict_solve_to(&_subset->dof_ids(),_subset_solve_mode);

  // Solve the linear system.  Several cases:
  std::pair<unsigned int, Real> rval = std::make_pair(0,0.0);
  if (_shell_matrix)
    // 1.) Shell matrix with or without user-supplied preconditioner.
    rval = linear_solver->solve(*_shell_matrix, this->request_matrix("Preconditioner"), *solution, *rhs, tol, maxits);
  else
    // 2.) No shell matrix, with or without user-supplied preconditioner
    rval = linear_solver->solve (*matrix, this->request_matrix("Preconditioner"), *solution, *rhs, tol, maxits);

  if (_subset != nullptr)
    linear_solver->restrict_solve_to(nullptr);

  // Store the number of linear iterations required to
  // solve and the final residual.
  _n_linear_iterations   = rval.first;
  _final_linear_residual = rval.second;

  // Update the system after the solve
  this->update();
}

solve_for_matrix_and_rhs()

void libMesh::RBConstruction::solve_for_matrix_and_rhs ( LinearSolver< Number > &  input_solver, SparseMatrix< Number > &  input_matrix, NumericVector< Number > &  input_rhs  )

virtual

Assembles & solves the linear system A*x=b for the specified matrix input_matrix and right-hand side rhs.

Definition at line 137 of file rb_construction.C.

References libMesh::LinearImplicitSystem::_final_linear_residual, libMesh::LinearImplicitSystem::_n_linear_iterations, libMesh::DofMap::enforce_constraints_exactly(), libMesh::Parameters::get(), libMesh::System::get_dof_map(), libMesh::System::get_equation_systems(), libMesh::LinearSolver< T >::init(), libMesh::EquationSystems::parameters, libMesh::Real, libMesh::System::solution, libMesh::LinearSolver< T >::solve(), and libMesh::System::update().

Referenced by compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), compute_residual_dual_norm_slow(), enrich_basis_from_rhs_terms(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

{
  // This is similar to LinearImplicitSysmte::solve()

  // Get a reference to the EquationSystems
  const EquationSystems & es =
    this->get_equation_systems();

  // If the linear solver hasn't been initialized, we do so here.
  input_solver.init();

  // Get the user-specifiied linear solver tolerance
  const double tol  =
    double(es.parameters.get<Real>("linear solver tolerance"));

  // Get the user-specified maximum # of linear solver iterations
  const unsigned int maxits =
    es.parameters.get<unsigned int>("linear solver maximum iterations");

  // It's good practice to clear the solution vector first since it can
  // affect convergence of iterative solvers
  solution->zero();

  // Solve the linear system.
  // Store the number of linear iterations required to
  // solve and the final residual.
  std::tie(_n_linear_iterations, _final_linear_residual) =
    input_solver.solve (input_matrix, *solution, input_rhs, tol, maxits);

  get_dof_map().enforce_constraints_exactly(*this);

  // Update the system after the solve
  this->update();
}

solve_for_unconstrained_dofs()

void libMesh::System::solve_for_unconstrained_dofs ( NumericVector< Number > &  vec, int  is_adjoint = -1  ) const

inherited

Definition at line 2102 of file system_projection.C.

References libMesh::DofMap::build_sparsity(), libMesh::DofMap::computed_sparsity_already(), libMesh::DofMapBase::end_dof(), libMesh::DofMapBase::first_dof(), libMesh::NumericVector< T >::get(), libMesh::DofMap::heterogenously_constrain_element_matrix_and_vector(), libMesh::DofMap::is_constrained_dof(), libMesh::NumericVector< T >::local_size(), libMesh::DofMap::n_dofs(), libMesh::DofMap::n_local_dofs(), libMesh::PARALLEL, libMesh::Real, libMesh::NumericVector< T >::size(), and libMesh::DofMap::update_sparsity_pattern().

{
  const DofMap & dof_map = this->get_dof_map();

  std::unique_ptr<SparseMatrix<Number>> mat =
    SparseMatrix<Number>::build(this->comm());

  std::unique_ptr<SparsityPattern::Build> sp;

  if (dof_map.computed_sparsity_already())
    dof_map.update_sparsity_pattern(*mat);
  else
    {
      mat->attach_dof_map(dof_map);
      sp = dof_map.build_sparsity(this->get_mesh());
      mat->attach_sparsity_pattern(*sp);
    }

  mat->init();

  libmesh_assert_equal_to(vec.size(), dof_map.n_dofs());
  libmesh_assert_equal_to(vec.local_size(), dof_map.n_local_dofs());

  std::unique_ptr<NumericVector<Number>> rhs =
    NumericVector<Number>::build(this->comm());

  rhs->init(dof_map.n_dofs(), dof_map.n_local_dofs(), false,
            PARALLEL);

  // Here we start with the unconstrained (and indeterminate) linear
  // system, K*u = f, where K is the identity matrix for constrained
  // DoFs and 0 elsewhere, and f is the current solution values for
  // constrained DoFs and 0 elsewhere.
  // We then apply the usual heterogeneous constraint matrix C and
  // offset h, where u = C*x + h,
  // to get C^T*K*C*x = C^T*f - C^T*K*h
  // - a constrained and no-longer-singular system that finds the
  // closest approximation for the unconstrained degrees of freedom.
  //
  // Here, though "closest" is in an algebraic sense; we're
  // effectively using a pseudoinverse that optimizes in a
  // discretization-dependent norm.  That only seems to give ~0.1%
  // excess error even in coarse unit test cases, but at some point it
  // might be reasonable to weight K and f properly.

  for (dof_id_type d : IntRange<dof_id_type>(dof_map.first_dof(),
                                             dof_map.end_dof()))
    {
      if (dof_map.is_constrained_dof(d))
        {
          DenseMatrix<Number> K(1,1);
          DenseVector<Number> F(1);
          std::vector<dof_id_type> dof_indices(1, d);
          K(0,0) = 1;
          F(0) = (*this->solution)(d);
          dof_map.heterogenously_constrain_element_matrix_and_vector
            (K, F, dof_indices, false, is_adjoint);
          mat->add_matrix(K, dof_indices);
          rhs->add_vector(F, dof_indices);
        }
    }

  std::unique_ptr<LinearSolver<Number>> linear_solver =
    LinearSolver<Number>::build(this->comm());

  linear_solver->solve(*mat, vec, *rhs,
                       double(this->get_equation_systems().parameters.get<Real>("linear solver tolerance")),
                       this->get_equation_systems().parameters.get<unsigned int>("linear solver maximum iterations"));
}

system()

sys_type& libMesh::RBConstruction::system ( )

inline

Returns

A reference to *this.

Definition at line 125 of file rb_construction.h.

{ return *this; }

system_type()

std::string libMesh::RBConstruction::system_type ( ) const

overridevirtual

Returns

A string indicating the type of the system.

Reimplemented from libMesh::LinearImplicitSystem.

Reimplemented in libMesh::TransientSystem< RBConstruction >.

Definition at line 132 of file rb_construction.C.

{
  return "RBConstruction";
}

train_reduced_basis()

Real libMesh::RBConstruction::train_reduced_basis ( const bool  resize_rb_eval_data = true )

virtual

Train the reduced basis.

This can use different approaches, e.g. Greedy or POD, which are chosen using the RB_training_type member variable.

In the case that we use Greedy training, this function returns the final maximum a posteriori error bound on the training set.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 1193 of file rb_construction.C.

References get_RB_training_type(), train_reduced_basis_with_greedy(), and train_reduced_basis_with_POD().

Referenced by main(), and libMesh::TransientRBConstruction::train_reduced_basis().

{
  if(get_RB_training_type() == "Greedy")
    {
      return train_reduced_basis_with_greedy(resize_rb_eval_data);
    }
  else if (get_RB_training_type() == "POD")
    {
      train_reduced_basis_with_POD();
      return 0.;
    }
  else
    {
      libmesh_error_msg("RB training type not recognized: " + get_RB_training_type());
    }

  return 0.;
}

train_reduced_basis_with_greedy()

Real libMesh::RBConstruction::train_reduced_basis_with_greedy

(

const bool

resize_rb_eval_data

)

Train the reduced basis using the "Greedy algorithm.".

Each stage of the Greedy algorithm involves solving the reduced basis over a large training set and selecting the parameter at which the reduced basis error bound is largest, then performing a truth_solve at that parameter and enriching the reduced basis with the corresponding snapshot.

resize_rb_eval_data is a boolean flag to indicate whether or not we call rb_eval->resize_data_structures(Nmax). True by default, but we may set it to false if, for example, we are continuing from a previous training run and don't want to clobber the existing rb_eval data.

Returns

The final maximum a posteriori error bound on the training set.

Definition at line 1212 of file rb_construction.C.

References check_if_zero_truth_solve(), compute_Fq_representor_innerprods(), compute_max_error_bound(), compute_output_dual_innerprods(), enrich_RB_space(), libMesh::RBEvaluation::get_n_basis_functions(), get_Nmax(), get_preevaluate_thetas_flag(), get_rb_evaluation(), libMesh::RBEvaluation::greedy_param_list, greedy_termination_test(), libMesh::RBParametrized::initialize_parameters(), libMesh::out, preevaluate_thetas(), libMesh::RBParametrized::print_parameters(), libMesh::Real, libMesh::RBEvaluation::resize_data_structures(), skip_residual_in_train_reduced_basis, truth_solve(), update_greedy_param_list(), update_residual_terms(), update_system(), and use_empty_rb_solve_in_greedy.

Referenced by train_reduced_basis().

{
  LOG_SCOPE("train_reduced_basis_with_greedy()", "RBConstruction");

  int count = 0;

  RBEvaluation & rbe = get_rb_evaluation();

  // initialize rbe's parameters
  rbe.initialize_parameters(*this);

  // possibly resize data structures according to Nmax
  if (resize_rb_eval_data)
    rbe.resize_data_structures(get_Nmax());

  // Clear the Greedy param list
  for (auto & plist : rbe.greedy_param_list)
    plist.clear();

  rbe.greedy_param_list.clear();

  Real training_greedy_error = 0.;


  // If we are continuing from a previous training run,
  // we might already be at the max number of basis functions.
  // If so, we can just return.
  if (rbe.get_n_basis_functions() >= get_Nmax())
    {
      libMesh::out << "Maximum number of basis functions reached: Nmax = "
                   << get_Nmax() << std::endl;
      return 0.;
    }

  // Optionally pre-evaluate the theta functions on the entire (local) training parameter set.
  if (get_preevaluate_thetas_flag())
    preevaluate_thetas();

  if(!skip_residual_in_train_reduced_basis)
    {
      // Compute the dual norms of the outputs if we haven't already done so.
      compute_output_dual_innerprods();

      // Compute the Fq Riesz representor dual norms if we haven't already done so.
      compute_Fq_representor_innerprods();
    }

  libMesh::out << std::endl << "---- Performing Greedy basis enrichment ----" << std::endl;
  Real initial_greedy_error = 0.;
  bool initial_greedy_error_initialized = false;
  while (true)
    {
      libMesh::out << std::endl << "---- Basis dimension: "
                   << rbe.get_n_basis_functions() << " ----" << std::endl;

      if (count > 0 || (count==0 && use_empty_rb_solve_in_greedy))
        {
          libMesh::out << "Performing RB solves on training set" << std::endl;
          training_greedy_error = compute_max_error_bound();

          libMesh::out << "Maximum error bound is " << training_greedy_error << std::endl << std::endl;

          // record the initial error
          if (!initial_greedy_error_initialized)
            {
              initial_greedy_error = training_greedy_error;
              initial_greedy_error_initialized = true;
            }

          // Break out of training phase if we have reached Nmax
          // or if the training_tolerance is satisfied.
          if (greedy_termination_test(training_greedy_error, initial_greedy_error, count))
            break;
        }

      libMesh::out << "Performing truth solve at parameter:" << std::endl;
      print_parameters();

      // Update the list of Greedily selected parameters
      this->update_greedy_param_list();

      // Perform an Offline truth solve for the current parameter
      truth_solve(-1);

      if (check_if_zero_truth_solve())
        {
          libMesh::out << "Zero basis function encountered hence ending basis enrichment" << std::endl;
          break;
        }

      // Add orthogonal part of the snapshot to the RB space
      libMesh::out << "Enriching the RB space" << std::endl;
      enrich_RB_space();

      update_system();

      // Check if we've reached Nmax now. We do this before calling
      // update_residual_terms() since we can skip that step if we've
      // already reached Nmax.
      if (rbe.get_n_basis_functions() >= this->get_Nmax())
      {
        libMesh::out << "Maximum number of basis functions reached: Nmax = "
                     << get_Nmax() << std::endl;
        break;
      }

      if(!skip_residual_in_train_reduced_basis)
        {
          update_residual_terms();
        }

      // Increment counter
      count++;
    }
  this->update_greedy_param_list();

  return training_greedy_error;
}

train_reduced_basis_with_POD()

void libMesh::RBConstruction::train_reduced_basis_with_POD

(

)

Train the reduced basis using Proper Orthogonal Decomposition (POD).

This is an alternative to train_reduced_basis(), which uses the RB greedy algorithm. In contrast to the RB greedy algorithm, POD requires us to perform truth solves at all training samples, which can be computationally intensive.

The main advantage of using POD is that it does not rely on the RB error indicator. The RB error indicator typically stagnates due to rounding error at approximately square-root of machine precision, since it involves taking the square-root of a sum of terms that cancel. This error indicator stagnation puts a limit on the accuracy level that can be achieved with the RB greedy algorithm, so for cases where we need higher accuracy, the POD approach is a good alternative.

Definition at line 1425 of file rb_construction.C.

References libMesh::RBConstructionBase< LinearImplicitSystem >::_normalize_solution_snapshots, libMesh::RBEvaluation::basis_functions, libMesh::NumericVector< T >::build(), libMesh::ParallelObject::comm(), delta_N, libMesh::DenseMatrix< T >::el(), libMesh::RBEvaluation::get_n_basis_functions(), libMesh::RBParametrized::get_n_params(), libMesh::RBConstructionBase< LinearImplicitSystem >::get_n_training_samples(), get_Nmax(), get_non_dirichlet_inner_product_matrix_if_avail(), get_rb_evaluation(), libMesh::RBParametrized::initialize_parameters(), libMesh::RBConstructionBase< LinearImplicitSystem >::inner_product_storage_vector, libMesh::libmesh_assert(), libMesh::libmesh_conj(), libMesh::System::n_dofs(), libMesh::System::n_local_dofs(), libMesh::TensorTools::norm(), libMesh::out, libMesh::PARALLEL, std::real(), libMesh::Real, recompute_all_residual_terms(), rel_training_tolerance, libMesh::RBEvaluation::resize_data_structures(), libMesh::RBConstructionBase< LinearImplicitSystem >::serial_training_set, libMesh::RBConstructionBase< LinearImplicitSystem >::set_params_from_training_set(), libMesh::System::solution, libMesh::DenseMatrix< T >::svd(), truth_solve(), update_system(), and libMesh::SparseMatrix< T >::vector_mult().

Referenced by train_reduced_basis().

{
  // We need to use the same training set on all processes so that
  // the truth solves below work correctly in parallel.
  libmesh_error_msg_if(!serial_training_set, "We must use a serial training set with POD");
  libmesh_error_msg_if(get_rb_evaluation().get_n_basis_functions() > 0, "Basis should not already be initialized");

  get_rb_evaluation().initialize_parameters(*this);
  get_rb_evaluation().resize_data_structures(get_Nmax());

  // Storage for the POD snapshots
  unsigned int n_snapshots = get_n_training_samples();

  if (get_n_params() == 0)
    {
      // In this case we should have generated an empty training set
      // so assert this
      libmesh_assert(n_snapshots == 0);

      // If we have no parameters, then we should do exactly one "truth solve"
      n_snapshots = 1;
    }

  std::vector<std::unique_ptr<NumericVector<Number>>> POD_snapshots(n_snapshots);
  for (unsigned int i=0; i<n_snapshots; i++)
    {
      POD_snapshots[i] = NumericVector<Number>::build(this->comm());
      POD_snapshots[i]->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
    }

  // We use the same training set on all processes
  libMesh::out << std::endl;
  for (unsigned int i=0; i<n_snapshots; i++)
    {
      if (get_n_params() > 0)
        {
          set_params_from_training_set(i);
        }

      libMesh::out << "Truth solve " << (i+1) << " of " << n_snapshots << std::endl;

      truth_solve(-1);

      *POD_snapshots[i] = *solution;
    }
  libMesh::out << std::endl;

  if (_normalize_solution_snapshots)
  {
    libMesh::out << "Normalizing solution snapshots" << std::endl;
    for (unsigned int i=0; i<n_snapshots; i++)
      {
        get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(
          *inner_product_storage_vector, *POD_snapshots[i]);
        Real norm = std::sqrt(std::real(POD_snapshots[i]->dot(*inner_product_storage_vector)));

        if (norm > 0.)
          POD_snapshots[i]->scale(1./norm);
      }
  }

  // Set up the "correlation matrix"
  DenseMatrix<Number> correlation_matrix(n_snapshots,n_snapshots);
  for (unsigned int i=0; i<n_snapshots; i++)
    {
      get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(
        *inner_product_storage_vector, *POD_snapshots[i]);

      for (unsigned int j=0; j<=i; j++)
        {
          Number inner_prod = (POD_snapshots[j]->dot(*inner_product_storage_vector));

          correlation_matrix(i,j) = inner_prod;
          if(i != j)
            {
              correlation_matrix(j,i) = libmesh_conj(inner_prod);
            }
        }
    }

  // compute SVD of correlation matrix
  DenseVector<Real> sigma( n_snapshots );
  DenseMatrix<Number> U( n_snapshots, n_snapshots );
  DenseMatrix<Number> VT( n_snapshots, n_snapshots );
  correlation_matrix.svd(sigma, U, VT );

  if (sigma(0) == 0.)
    return;

  // Add dominant vectors from the POD as basis functions.
  unsigned int j = 0;
  while (true)
    {
      if (j >= get_Nmax() || j >= n_snapshots)
        {
          libMesh::out << "Maximum number of basis functions (" << j << ") reached." << std::endl;
          break;
        }

      // The "energy" error in the POD approximation is determined by the first omitted
      // singular value, i.e. sigma(j). We normalize by sigma(0), which gives the total
      // "energy", in order to obtain a relative error.
      const Real rel_err = std::sqrt(sigma(j)) / std::sqrt(sigma(0));

      libMesh::out << "Number of basis functions: " << j
                   << ", POD error norm: " << rel_err << std::endl;

      if (rel_err < this->rel_training_tolerance)
        {
          libMesh::out << "Training tolerance reached." << std::endl;
          break;
        }

      std::unique_ptr< NumericVector<Number> > v = POD_snapshots[j]->zero_clone();
      for ( unsigned int i=0; i<n_snapshots; ++i )
        {
          v->add( U.el(i, j), *POD_snapshots[i] );
        }

      Real norm_v = std::sqrt(sigma(j));
      v->scale( 1./norm_v );

      get_rb_evaluation().basis_functions.emplace_back( std::move(v) );

      j++;
    }
  libMesh::out << std::endl;

  this->delta_N = get_rb_evaluation().get_n_basis_functions();
  update_system();

  // We now compute all terms required to evaluate the RB error indicator.
  // Unlike in the case of the RB Greedy algorithm, for the POD approach
  // we do not need this data in order to compute the basis. However, we
  // do need this data in order to evaluate error indicator quantities in
  // the Online stage, so we compute it now so that it can be saved in
  // the training data.
  recompute_all_residual_terms();
}

truth_assembly()

void libMesh::RBConstruction::truth_assembly ( )

protectedvirtual

Assemble the truth matrix and right-hand side for current_parameters.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 970 of file rb_construction.C.

References libMesh::SparseMatrix< T >::add(), libMesh::NumericVector< T >::add(), libMesh::NumericVector< T >::build(), libMesh::NumericVector< T >::close(), libMesh::SparseMatrix< T >::close(), libMesh::ParallelObject::comm(), get_Aq(), get_Fq(), libMesh::RBThetaExpansion::get_n_A_terms(), libMesh::RBThetaExpansion::get_n_F_terms(), libMesh::RBParametrized::get_parameters(), get_rb_theta_expansion(), libMesh::ImplicitSystem::matrix, libMesh::System::n_dofs(), libMesh::System::n_local_dofs(), libMesh::PARALLEL, libMesh::ExplicitSystem::rhs, libMesh::SparseMatrix< T >::zero(), and libMesh::NumericVector< T >::zero().

Referenced by compute_residual_dual_norm_slow(), enrich_basis_from_rhs_terms(), and truth_solve().

{
  LOG_SCOPE("truth_assembly()", "RBConstruction");

  const RBParameters & mu = get_parameters();

  this->matrix->zero();
  this->rhs->zero();

  this->matrix->close();
  this->rhs->close();

  {
    // We should have already assembled the matrices
    // and vectors in the affine expansion, so
    // just use them

    for (unsigned int q_a=0; q_a<get_rb_theta_expansion().get_n_A_terms(); q_a++)
      {
        matrix->add(get_rb_theta_expansion().eval_A_theta(q_a, mu), *get_Aq(q_a));
      }

    std::unique_ptr<NumericVector<Number>> temp_vec = NumericVector<Number>::build(this->comm());
    temp_vec->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
    for (unsigned int q_f=0; q_f<get_rb_theta_expansion().get_n_F_terms(); q_f++)
      {
        *temp_vec = *get_Fq(q_f);
        temp_vec->scale( get_rb_theta_expansion().eval_F_theta(q_f, mu) );
        rhs->add(*temp_vec);
      }
  }

  this->matrix->close();
  this->rhs->close();
}

truth_solve()

Real libMesh::RBConstruction::truth_solve ( int  plot_solution )

virtual

Perform a "truth" solve, i.e.

solve the finite element system at at the parameters currently set in the system. This is used extensively in training the reduced basis, since "truth snapshots" are employed as basis functions.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 1615 of file rb_construction.C.

References _untransformed_solution, assert_convergence, check_convergence(), libMesh::NumericVector< T >::dot(), libMesh::RBThetaExpansion::eval_output_theta(), extra_linear_solver, libMesh::System::get_equation_systems(), libMesh::LinearImplicitSystem::get_linear_solver(), libMesh::System::get_mesh(), libMesh::RBThetaExpansion::get_n_output_terms(), libMesh::RBThetaExpansion::get_n_outputs(), get_non_dirichlet_inner_product_matrix_if_avail(), get_output_vector(), libMesh::RBParametrized::get_parameters(), get_rb_theta_expansion(), libMesh::RBConstructionBase< LinearImplicitSystem >::inner_product_storage_vector, libMesh::libmesh_ignore(), libMesh::libmesh_real(), libMesh::ImplicitSystem::matrix, libMesh::System::name(), post_process_truth_solution(), libMesh::ExplicitSystem::rhs, libMesh::System::solution, solve_for_matrix_and_rhs(), store_untransformed_basis, truth_assembly(), truth_outputs, libMesh::SparseMatrix< T >::vector_mult(), and libMesh::ExodusII_IO::write_equation_systems().

Referenced by train_reduced_basis_with_greedy(), and train_reduced_basis_with_POD().

{
  LOG_SCOPE("truth_solve()", "RBConstruction");

  if(store_untransformed_basis && !_untransformed_solution)
  {
    _untransformed_solution = solution->zero_clone();
  }

  truth_assembly();

  // truth_assembly assembles into matrix and rhs, so use those for the solve
  if (extra_linear_solver)
    {
      // If extra_linear_solver has been initialized, then we use it for the
      // truth solves.
      solve_for_matrix_and_rhs(*extra_linear_solver, *matrix, *rhs);

      if (assert_convergence)
        check_convergence(*extra_linear_solver);
    }
  else
    {
      solve_for_matrix_and_rhs(*get_linear_solver(), *matrix, *rhs);

      if (assert_convergence)
        check_convergence(*get_linear_solver());
    }

  if (store_untransformed_basis)
    {
      *_untransformed_solution = *solution;
    }

  // Call user-defined post-processing routines on the truth solution.
  post_process_truth_solution();

  const RBParameters & mu = get_parameters();

  for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
    {
      truth_outputs[n] = 0.;
      for (unsigned int q_l=0; q_l<get_rb_theta_expansion().get_n_output_terms(n); q_l++)
        truth_outputs[n] += get_rb_theta_expansion().eval_output_theta(n, q_l, mu)*
          get_output_vector(n,q_l)->dot(*solution);
    }

#ifdef LIBMESH_HAVE_EXODUS_API
  if (plot_solution > 0)
    {
      ExodusII_IO exo_io(this->get_mesh());
      std::set<std::string> system_names = {this->name()};
      exo_io.write_equation_systems("truth.exo", this->get_equation_systems(), &system_names);
    }
#else
  libmesh_ignore(plot_solution);
#endif

  // Get the X norm of the truth solution
  // Useful for normalizing our true error data
  get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*inner_product_storage_vector, *solution);
  Number truth_X_norm = std::sqrt(inner_product_storage_vector->dot(*solution));

  return libmesh_real(truth_X_norm);
}

update()

void libMesh::System::update ( )

virtualinherited

Update the local values to reflect the solution on neighboring processors.

Reimplemented in SolidSystem.

Definition at line 498 of file system.C.

References libMesh::System::_dof_map, libMesh::System::current_local_solution, libMesh::libmesh_assert(), and libMesh::System::solution.

Referenced by libMesh::__libmesh_petsc_diff_solver_jacobian(), libMesh::__libmesh_petsc_diff_solver_residual(), libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::FEMSystem::assemble_qoi(), libMesh::FEMSystem::assemble_qoi_derivative(), libMesh::NonlinearImplicitSystem::assembly(), libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::NewmarkSolver::compute_initial_accel(), compute_stresses(), LinearElasticityWithContact::compute_stresses(), LinearElasticity::compute_stresses(), LargeDeformationElasticity::compute_stresses(), libMesh::Problem_Interface::computeF(), libMesh::Problem_Interface::computeJacobian(), libMesh::Problem_Interface::computePreconditioner(), libMesh::Nemesis_IO::copy_elemental_solution(), libMesh::ExodusII_IO::copy_elemental_solution(), libMesh::GMVIO::copy_nodal_solution(), libMesh::Nemesis_IO::copy_nodal_solution(), libMesh::ExodusII_IO::copy_nodal_solution(), libMesh::Nemesis_IO::copy_scalar_solution(), libMesh::ExodusII_IO::copy_scalar_solution(), DMlibMeshFunction(), DMlibMeshJacobian(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::CondensedEigenSystem::get_eigenpair(), libMesh::TransientRBConstruction::initialize_truth(), libMesh::Euler2Solver::integrate_adjoint_refinement_error_estimate(), libMesh::EulerSolver::integrate_adjoint_refinement_error_estimate(), libMesh::libmesh_petsc_snes_fd_residual(), libMesh::libmesh_petsc_snes_jacobian(), libMesh::libmesh_petsc_snes_mffd_residual(), libMesh::libmesh_petsc_snes_residual(), libMesh::libmesh_petsc_snes_residual_helper(), libMesh::NewtonSolver::line_search(), load_basis_function(), libMesh::TransientRBConstruction::load_rb_solution(), load_rb_solution(), main(), libMesh::FEMSystem::mesh_position_get(), HeatSystem::perturb_accumulate_residuals(), libMesh::ErrorVector::plot_error(), libMesh::FEMSystem::postprocess(), libMesh::ImplicitSystem::qoi_parameter_hessian(), libMesh::MemorySolutionHistory::retrieve(), libMesh::FileSolutionHistory::retrieve(), libMesh::NewtonSolver::solve(), libMesh::ExplicitSystem::solve(), libMesh::LinearImplicitSystem::solve(), libMesh::OptimizationSystem::solve(), libMesh::NonlinearImplicitSystem::solve(), libMesh::ClawSystem::solve_conservation_law(), solve_for_matrix_and_rhs(), libMesh::MeshFunctionSolutionTransfer::transfer(), libMesh::DirectSolutionTransfer::transfer(), and update_current_local_solution().

{
  parallel_object_only();

  libmesh_assert(solution->closed());

  const std::vector<dof_id_type> & send_list = _dof_map->get_send_list ();

  // Check sizes
  libmesh_assert_equal_to (current_local_solution->size(), solution->size());
  // More processors than elements => empty send_list
  //  libmesh_assert (!send_list.empty());
  libmesh_assert_less_equal (send_list.size(), solution->size());

  // Create current_local_solution from solution.  This will
  // put a local copy of solution into current_local_solution.
  // Only the necessary values (specified by the send_list)
  // are copied to minimize communication
  solution->localize (*current_local_solution, send_list);
}

update_global_solution() [1/2]

void libMesh::System::update_global_solution ( std::vector< Number > &  global_soln ) const

inherited

Fill the input vector global_soln so that it contains the global solution on all processors.

Requires communication with all other processors.

Definition at line 733 of file system.C.

References libMesh::System::solution.

Referenced by libMesh::ExactSolution::_compute_error(), libMesh::ExactErrorEstimator::estimate_error(), main(), and libMesh::InterMeshProjection::project_system_vectors().

{
  parallel_object_only();

  global_soln.resize (solution->size());

  solution->localize (global_soln);
}

update_global_solution() [2/2]

void libMesh::System::update_global_solution ( std::vector< Number > &  global_soln, const processor_id_type  dest_proc  ) const

inherited

Fill the input vector global_soln so that it contains the global solution on processor dest_proc.

Requires communication with all other processors.

Definition at line 744 of file system.C.

References libMesh::System::solution.

{
  parallel_object_only();

  global_soln.resize        (solution->size());

  solution->localize_to_one (global_soln, dest_proc);
}

update_greedy_param_list()

void libMesh::RBConstruction::update_greedy_param_list ( )

protected

Update the list of Greedily chosen parameters with current_parameters.

Definition at line 1602 of file rb_construction.C.

References libMesh::RBParametrized::get_parameters(), get_rb_evaluation(), and libMesh::RBEvaluation::greedy_param_list.

Referenced by train_reduced_basis_with_greedy().

{
  get_rb_evaluation().greedy_param_list.push_back( get_parameters() );
}

update_RB_system_matrices()

void libMesh::RBConstruction::update_RB_system_matrices ( )

protectedvirtual

Compute the reduced basis matrices for the current basis.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 1910 of file rb_construction.C.

References libMesh::NumericVector< T >::build(), libMesh::ParallelObject::comm(), compute_RB_inner_product, delta_N, libMesh::NumericVector< T >::dot(), libMesh::RBThetaExpansion::get_n_A_terms(), libMesh::RBEvaluation::get_n_basis_functions(), libMesh::RBThetaExpansion::get_n_F_terms(), libMesh::RBThetaExpansion::get_n_output_terms(), libMesh::RBThetaExpansion::get_n_outputs(), get_non_dirichlet_Aq_if_avail(), get_non_dirichlet_Fq_if_avail(), get_non_dirichlet_inner_product_matrix_if_avail(), get_output_vector(), get_rb_evaluation(), get_rb_theta_expansion(), libMesh::libmesh_conj(), libMesh::System::n_dofs(), libMesh::System::n_local_dofs(), libMesh::PARALLEL, libMesh::RBEvaluation::RB_Aq_vector, libMesh::RBEvaluation::RB_Fq_vector, libMesh::RBEvaluation::RB_inner_product_matrix, libMesh::RBEvaluation::RB_output_vectors, value, and libMesh::SparseMatrix< T >::vector_mult().

Referenced by libMesh::TransientRBConstruction::update_RB_system_matrices(), and update_system().

{
  LOG_SCOPE("update_RB_system_matrices()", "RBConstruction");

  unsigned int RB_size = get_rb_evaluation().get_n_basis_functions();

  std::unique_ptr<NumericVector<Number>> temp = NumericVector<Number>::build(this->comm());
  temp->init (this->n_dofs(), this->n_local_dofs(), false, PARALLEL);

  for (unsigned int q_f=0; q_f<get_rb_theta_expansion().get_n_F_terms(); q_f++)
    {
      for (unsigned int i=(RB_size-delta_N); i<RB_size; i++)
        {
          get_rb_evaluation().RB_Fq_vector[q_f](i) = get_non_dirichlet_Fq_if_avail(q_f)->dot(get_rb_evaluation().get_basis_function(i));
        }
    }

  for (unsigned int i=(RB_size-delta_N); i<RB_size; i++)
    {
      for (unsigned int n=0; n<get_rb_theta_expansion().get_n_outputs(); n++)
        for (unsigned int q_l=0; q_l<get_rb_theta_expansion().get_n_output_terms(n); q_l++)
          {
            get_rb_evaluation().RB_output_vectors[n][q_l](i) =
              get_output_vector(n,q_l)->dot(get_rb_evaluation().get_basis_function(i));
          }

      for (unsigned int j=0; j<RB_size; j++)
        {
          Number value = 0.;

          if (compute_RB_inner_product)
            {
              // Compute reduced inner_product_matrix
              temp->zero();
              get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*temp, get_rb_evaluation().get_basis_function(j));

              value = temp->dot( get_rb_evaluation().get_basis_function(i) );
              get_rb_evaluation().RB_inner_product_matrix(i,j) = value;
              if (i!=j)
                {
                  // The inner product matrix is assumed
                  // to be hermitian
                  get_rb_evaluation().RB_inner_product_matrix(j,i) = libmesh_conj(value);
                }
            }

          for (unsigned int q_a=0; q_a<get_rb_theta_expansion().get_n_A_terms(); q_a++)
            {
              // Compute reduced Aq matrix
              temp->zero();
              get_non_dirichlet_Aq_if_avail(q_a)->vector_mult(*temp, get_rb_evaluation().get_basis_function(j));

              value = (*temp).dot(get_rb_evaluation().get_basis_function(i));
              get_rb_evaluation().RB_Aq_vector[q_a](i,j) = value;

              if (i!=j)
                {
                  temp->zero();
                  get_non_dirichlet_Aq_if_avail(q_a)->vector_mult(*temp, get_rb_evaluation().get_basis_function(i));

                  value = (*temp).dot(get_rb_evaluation().get_basis_function(j));
                  get_rb_evaluation().RB_Aq_vector[q_a](j,i) = value;
                }
            }
        }
    }
}

update_residual_terms()

void libMesh::RBConstruction::update_residual_terms ( bool  compute_inner_products = true )

protectedvirtual

Compute the terms that are combined ‘online’ to determine the dual norm of the residual.

By default, inner product terms are also computed, but you can turn this feature off e.g. if you are already reading in that data from files.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 1979 of file rb_construction.C.

References _untransformed_basis_functions, libMesh::RBEvaluation::Aq_Aq_representor_innerprods, libMesh::RBEvaluation::Aq_representor, assert_convergence, libMesh::NumericVector< T >::build(), check_convergence(), libMesh::ParallelObject::comm(), delta_N, libMesh::LinearImplicitSystem::final_linear_residual(), libMesh::RBEvaluation::Fq_Aq_representor_innerprods, Fq_representor, get_Aq(), libMesh::RBThetaExpansion::get_n_A_terms(), libMesh::RBEvaluation::get_n_basis_functions(), libMesh::RBThetaExpansion::get_n_F_terms(), get_non_dirichlet_inner_product_matrix_if_avail(), get_rb_evaluation(), get_rb_theta_expansion(), libMesh::Utility::get_timestamp(), inner_product_matrix, inner_product_solver, libMesh::RBConstructionBase< LinearImplicitSystem >::inner_product_storage_vector, libMesh::RBConstructionBase< LinearImplicitSystem >::is_quiet(), libMesh::libmesh_assert(), libMesh::System::n_dofs(), libMesh::LinearImplicitSystem::n_linear_iterations(), libMesh::System::n_local_dofs(), libMesh::out, libMesh::PARALLEL, libMesh::ExplicitSystem::rhs, libMesh::NumericVector< T >::scale(), libMesh::System::solution, solve_for_matrix_and_rhs(), store_untransformed_basis, libMesh::SparseMatrix< T >::vector_mult(), and libMesh::NumericVector< T >::zero().

Referenced by recompute_all_residual_terms(), train_reduced_basis_with_greedy(), and libMesh::TransientRBConstruction::update_residual_terms().

{
  LOG_SCOPE("update_residual_terms()", "RBConstruction");

  libMesh::out << "Updating RB residual terms" << std::endl;

  unsigned int RB_size = get_rb_evaluation().get_n_basis_functions();

  if (store_untransformed_basis)
    {
      libmesh_assert_equal_to(_untransformed_basis_functions.size(), get_rb_evaluation().get_n_basis_functions());
    }

  for (unsigned int q_a=0; q_a<get_rb_theta_expansion().get_n_A_terms(); q_a++)
    {
      for (unsigned int i=(RB_size-delta_N); i<RB_size; i++)
        {
          // Initialize the vector in which we'll store the representor
          if (!get_rb_evaluation().Aq_representor[q_a][i])
            {
              get_rb_evaluation().Aq_representor[q_a][i] = NumericVector<Number>::build(this->comm());
              get_rb_evaluation().Aq_representor[q_a][i]->init(this->n_dofs(), this->n_local_dofs(), false, PARALLEL);
            }

          libmesh_assert(get_rb_evaluation().Aq_representor[q_a][i]->size()       == this->n_dofs()       &&
                         get_rb_evaluation().Aq_representor[q_a][i]->local_size() == this->n_local_dofs() );

          rhs->zero();
          if (!store_untransformed_basis)
            {
              get_Aq(q_a)->vector_mult(*rhs, get_rb_evaluation().get_basis_function(i));
            }
          else
            {
              get_Aq(q_a)->vector_mult(*rhs, *_untransformed_basis_functions[i]);
            }
          rhs->scale(-1.);

          if (!is_quiet())
            {
              libMesh::out << "Starting solve [q_a][i]=[" << q_a <<"]["<< i << "] in RBConstruction::update_residual_terms() at "
                           << Utility::get_timestamp() << std::endl;
            }

          solve_for_matrix_and_rhs(*inner_product_solver, *inner_product_matrix, *rhs);

          if (assert_convergence)
            check_convergence(*inner_product_solver);

          if (!is_quiet())
            {
              libMesh::out << "Finished solve [q_a][i]=[" << q_a <<"]["<< i << "] in RBConstruction::update_residual_terms() at "
                           << Utility::get_timestamp() << std::endl;
              libMesh::out << this->n_linear_iterations() << " iterations, final residual "
                           << this->final_linear_residual() << std::endl;
            }

          // Store the representor
          *get_rb_evaluation().Aq_representor[q_a][i] = *solution;
        }
    }

  // Now compute and store the inner products (if requested)
  if (compute_inner_products)
    {

      for (unsigned int q_f=0; q_f<get_rb_theta_expansion().get_n_F_terms(); q_f++)
        {
          get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*inner_product_storage_vector,*Fq_representor[q_f]);

          for (unsigned int q_a=0; q_a<get_rb_theta_expansion().get_n_A_terms(); q_a++)
            {
              for (unsigned int i=(RB_size-delta_N); i<RB_size; i++)
                {
                  get_rb_evaluation().Fq_Aq_representor_innerprods[q_f][q_a][i] =
                    inner_product_storage_vector->dot(*get_rb_evaluation().Aq_representor[q_a][i]);
                }
            }
        }

      unsigned int q=0;
      for (unsigned int q_a1=0; q_a1<get_rb_theta_expansion().get_n_A_terms(); q_a1++)
        {
          for (unsigned int q_a2=q_a1; q_a2<get_rb_theta_expansion().get_n_A_terms(); q_a2++)
            {
              for (unsigned int i=(RB_size-delta_N); i<RB_size; i++)
                {
                  for (unsigned int j=0; j<RB_size; j++)
                    {
                      get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*inner_product_storage_vector, *get_rb_evaluation().Aq_representor[q_a2][j]);
                      get_rb_evaluation().Aq_Aq_representor_innerprods[q][i][j] =
                        inner_product_storage_vector->dot(*get_rb_evaluation().Aq_representor[q_a1][i]);

                      if (i != j)
                        {
                          get_non_dirichlet_inner_product_matrix_if_avail()->vector_mult(*inner_product_storage_vector, *get_rb_evaluation().Aq_representor[q_a2][i]);
                          get_rb_evaluation().Aq_Aq_representor_innerprods[q][j][i] =
                            inner_product_storage_vector->dot(*get_rb_evaluation().Aq_representor[q_a1][j]);
                        }
                    }
                }
              q++;
            }
        }
    } // end if (compute_inner_products)
}

update_system()

void libMesh::RBConstruction::update_system ( )

protectedvirtual

Update the system after enriching the RB space; this calls a series of functions to update the system properly.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 1778 of file rb_construction.C.

References libMesh::out, and update_RB_system_matrices().

Referenced by enrich_basis_from_rhs_terms(), train_reduced_basis_with_greedy(), train_reduced_basis_with_POD(), and libMesh::TransientRBConstruction::update_system().

{
  libMesh::out << "Updating RB matrices" << std::endl;
  update_RB_system_matrices();
}

user_assembly()

void libMesh::System::user_assembly ( )

virtualinherited

Calls user's attached assembly function, or is overridden by the user in derived classes.

Definition at line 2104 of file system.C.

References libMesh::System::_assemble_system_function, libMesh::System::_assemble_system_object, libMesh::System::_equation_systems, libMesh::System::Assembly::assemble(), and libMesh::System::name().

Referenced by libMesh::System::assemble().

{
  // Call the user-provided assembly function,
  // if it was provided
  if (_assemble_system_function != nullptr)
    this->_assemble_system_function (_equation_systems, this->name());

  // ...or the user-provided assembly object.
  else if (_assemble_system_object != nullptr)
    this->_assemble_system_object->assemble();
}

user_constrain()

void libMesh::System::user_constrain ( )

virtualinherited

Calls user's attached constraint function, or is overridden by the user in derived classes.

Definition at line 2118 of file system.C.

References libMesh::System::_constrain_system_function, libMesh::System::_constrain_system_object, libMesh::System::_equation_systems, libMesh::System::Constraint::constrain(), and libMesh::System::name().

Referenced by libMesh::System::reinit_constraints().

{
  // Call the user-provided constraint function,
  // if it was provided
  if (_constrain_system_function!= nullptr)
    this->_constrain_system_function(_equation_systems, this->name());

  // ...or the user-provided constraint object.
  else if (_constrain_system_object != nullptr)
    this->_constrain_system_object->constrain();
}

user_initialization()

void libMesh::System::user_initialization ( )

virtualinherited

Calls user's attached initialization function, or is overridden by the user in derived classes.

Definition at line 2090 of file system.C.

References libMesh::System::_equation_systems, libMesh::System::_init_system_function, libMesh::System::_init_system_object, libMesh::System::Initialization::initialize(), and libMesh::System::name().

Referenced by libMesh::NewmarkSystem::initial_conditions(), and libMesh::System::reinit_mesh().

{
  // Call the user-provided initialization function,
  // if it was provided
  if (_init_system_function != nullptr)
    this->_init_system_function (_equation_systems, this->name());

  // ...or the user-provided initialization object.
  else if (_init_system_object != nullptr)
    this->_init_system_object->initialize();
}

user_QOI()

void libMesh::System::user_QOI ( const QoISet &  qoi_indices )

virtualinherited

Calls user's attached quantity of interest function, or is overridden by the user in derived classes.

Definition at line 2132 of file system.C.

References libMesh::System::_equation_systems, libMesh::System::_qoi_evaluate_function, libMesh::System::_qoi_evaluate_object, libMesh::System::name(), and libMesh::System::QOI::qoi().

Referenced by libMesh::System::assemble_qoi().

{
  // Call the user-provided quantity of interest function,
  // if it was provided
  if (_qoi_evaluate_function != nullptr)
    this->_qoi_evaluate_function(_equation_systems, this->name(), qoi_indices);

  // ...or the user-provided QOI function object.
  else if (_qoi_evaluate_object != nullptr)
    this->_qoi_evaluate_object->qoi(qoi_indices);
}

user_QOI_derivative()

void libMesh::System::user_QOI_derivative ( const QoISet &  qoi_indices = QoISet(), bool  include_liftfunc = true, bool  apply_constraints = true  )

virtualinherited

Calls user's attached quantity of interest derivative function, or is overridden by the user in derived classes.

Definition at line 2146 of file system.C.

References libMesh::System::_equation_systems, libMesh::System::_qoi_evaluate_derivative_function, libMesh::System::_qoi_evaluate_derivative_object, libMesh::System::name(), and libMesh::System::QOIDerivative::qoi_derivative().

Referenced by libMesh::System::assemble_qoi_derivative().

{
  // Call the user-provided quantity of interest derivative,
  // if it was provided
  if (_qoi_evaluate_derivative_function != nullptr)
    this->_qoi_evaluate_derivative_function
      (_equation_systems, this->name(), qoi_indices, include_liftfunc,
       apply_constraints);

  // ...or the user-provided QOI derivative function object.
  else if (_qoi_evaluate_derivative_object != nullptr)
    this->_qoi_evaluate_derivative_object->qoi_derivative
      (qoi_indices, include_liftfunc, apply_constraints);
}

variable()

const Variable & libMesh::System::variable ( unsigned int  var ) const

inherited

Return a constant reference to Variable var.

Definition at line 2699 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::variable().

Referenced by libMesh::ExactSolution::_compute_error(), libMesh::PetscDMWrapper::add_dofs_to_section(), libMesh::DifferentiableSystem::add_second_order_dot_vars(), libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::FirstOrderUnsteadySolver::compute_second_order_eqns(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::SubFunctor::find_dofs_to_send(), libMesh::DifferentiableSystem::have_first_order_scalar_vars(), libMesh::DifferentiableSystem::have_second_order_scalar_vars(), main(), libMesh::DifferentiablePhysics::nonlocal_mass_residual(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::SortAndCopy::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectVertices::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectEdges::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectSides::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectInteriors::operator()(), libMesh::System::point_gradient(), libMesh::System::point_hessian(), libMesh::System::point_value(), libMesh::System::read_parallel_data(), libMesh::System::read_SCALAR_dofs(), libMesh::System::read_serialized_vector(), libMesh::System::read_serialized_vectors(), libMesh::PetscPreconditioner< T >::set_petsc_aux_data(), libMesh::PetscDMWrapper::set_point_range_in_section(), SystemsTest::testFirstScalarNumber(), libMesh::System::write_header(), libMesh::Nemesis_IO_Helper::write_nodal_solution(), libMesh::System::write_parallel_data(), libMesh::System::write_serialized_vector(), and libMesh::System::write_serialized_vectors().

{
  return this->get_dof_map().variable(i);
}

variable_group()

const VariableGroup & libMesh::System::variable_group ( unsigned int  vg ) const

inherited

Return a constant reference to VariableGroup vg.

Definition at line 2704 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::variable_group().

Referenced by libMesh::FEMSystem::assembly(), and libMesh::System::get_info().

{
  return this->get_dof_map().variable_group(vg);
}

variable_name()

const std::string & libMesh::System::variable_name ( const unsigned int  i ) const

inherited

Returns

The name of variable i.

Definition at line 2674 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::variable_name().

Referenced by libMesh::PetscDMWrapper::build_section(), libMesh::ExactErrorEstimator::estimate_error(), libMesh::ExactSolution::ExactSolution(), libMesh::EquationSystems::find_variable_numbers(), main(), output_norms(), libMesh::petsc_auto_fieldsplit(), libMesh::InterMeshProjection::project_system_vectors(), MeshInputTest::testExodusWriteElementDataFromDiscontinuousNodalData(), libMesh::System::write_header(), and libMesh::Nemesis_IO_Helper::write_nodal_solution().

{
  return this->get_dof_map().variable_name(i);
}

variable_number()

unsigned int libMesh::System::variable_number ( std::string_view  var ) const

inherited

Returns

The variable number associated with the user-specified variable named var.

Definition at line 1398 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::variable_number().

Referenced by libMesh::ExactSolution::_compute_error(), alternative_fe_assembly(), LinearElasticity::assemble(), AssembleOptimization::assemble_A_and_F(), assemble_divgrad(), assemble_elasticity(), assemble_graddiv(), assemble_matrix_and_rhs(), assemble_shell(), assemble_stokes(), compute_enriched_soln(), compute_stresses(), LinearElasticityWithContact::compute_stresses(), LinearElasticity::compute_stresses(), LargeDeformationElasticity::compute_stresses(), libMesh::Nemesis_IO::copy_elemental_solution(), libMesh::ExodusII_IO::copy_elemental_solution(), libMesh::GMVIO::copy_nodal_solution(), libMesh::Nemesis_IO::copy_nodal_solution(), libMesh::ExodusII_IO::copy_nodal_solution(), libMesh::ExactErrorEstimator::estimate_error(), fe_assembly(), libMesh::ExactErrorEstimator::find_squared_element_error(), CoupledSystemQoI::init_context(), libMesh::HDGProblem::jacobian(), LargeDeformationElasticity::jacobian(), line_print(), main(), LinearElasticityWithContact::move_mesh(), libMesh::System::read_header(), libMesh::HDGProblem::residual(), LargeDeformationElasticity::residual(), LinearElasticityWithContact::residual_and_jacobian(), OverlappingAlgebraicGhostingTest::run_ghosting_test(), OverlappingCouplingGhostingTest::run_sparsity_pattern_test(), OverlappingTestBase::setup_coupling_matrix(), libMesh::DTKAdapter::update_variable_values(), libMesh::System::variable_scalar_number(), libMesh::EnsightIO::write_scalar_ascii(), and libMesh::EnsightIO::write_vector_ascii().

{
  return this->get_dof_map().variable_number(var);
}

variable_scalar_number() [1/2]

unsigned int libMesh::System::variable_scalar_number ( std::string_view  var, unsigned int  component  ) const

inlineinherited

Returns

An index, starting from 0 for the first component of the first variable, and incrementing for each component of each (potentially vector-valued) variable in the system in order. For systems with only scalar-valued variables, this will be the same as variable_number(var)

Irony: currently our only non-scalar-valued variable type is SCALAR.

Definition at line 2474 of file system.h.

References libMesh::System::variable_number().

Referenced by libMesh::Nemesis_IO::copy_scalar_solution(), libMesh::ExodusII_IO::copy_scalar_solution(), libMesh::ExactErrorEstimator::find_squared_element_error(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectVertices::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectEdges::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectSides::operator()(), and libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectInteriors::operator()().

{
  return variable_scalar_number(this->variable_number(var), component);
}

variable_scalar_number() [2/2]

unsigned int libMesh::System::variable_scalar_number ( unsigned int  var_num, unsigned int  component  ) const

inherited

Returns

An index, starting from 0 for the first component of the first variable, and incrementing for each component of each (potentially vector-valued) variable in the system in order. For systems with only scalar-valued variables, this will be the same as var_num

Irony: currently our only non-scalar-valued variable type is SCALAR.

Definition at line 2710 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::variable_scalar_number().

{
  return this->get_dof_map().variable_scalar_number(var_num, component);
}

variable_type() [1/2]

const FEType & libMesh::System::variable_type ( const unsigned int  i ) const

inherited

Returns

The finite element type variable number i.

Definition at line 2716 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::variable_type().

Referenced by libMesh::ExactSolution::_compute_error(), alternative_fe_assembly(), assemble(), libMesh::ClawSystem::assemble_advection_matrices(), libMesh::ClawSystem::assemble_avg_coupling_matrices(), libMesh::ClawSystem::assemble_boundary_condition_matrices(), assemble_ellipticdg(), libMesh::ClawSystem::assemble_jump_coupling_matrix(), libMesh::ClawSystem::assemble_mass_matrix(), assemble_shell(), assemble_stokes(), libMesh::FEMContext::attach_quadrature_rules(), libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::Nemesis_IO::copy_elemental_solution(), libMesh::ExodusII_IO::copy_elemental_solution(), libMesh::GMVIO::copy_nodal_solution(), libMesh::DGFEMContext::DGFEMContext(), fe_assembly(), libMesh::FEMContext::find_hardest_fe_type(), libMesh::EquationSystems::find_variable_numbers(), form_functionA(), form_functionB(), form_matrixA(), libMesh::FEMSystem::init_context(), libMesh::FEMContext::init_internal_data(), RationalMapTest< elem_type >::setUp(), FETestBase< order, family, elem_type, N_x >::setUp(), EquationSystemsTest::testBadVarNames(), libMesh::Nemesis_IO_Helper::write_element_values(), libMesh::System::write_header(), libMesh::Nemesis_IO_Helper::write_nodal_solution(), libMesh::EnsightIO::write_scalar_ascii(), and libMesh::EnsightIO::write_vector_ascii().

{
  return this->get_dof_map().variable_type(i);
}

variable_type() [2/2]

const FEType & libMesh::System::variable_type ( std::string_view  var ) const

inherited

Returns

The finite element type for variable var.

Definition at line 2721 of file system.C.

References libMesh::System::get_dof_map(), and libMesh::DofMap::variable_type().

{
  return this->get_dof_map().variable_type(var);
}

vector_is_adjoint()

int libMesh::System::vector_is_adjoint ( std::string_view  vec_name ) const

inherited

Returns

The integer describing whether the vector identified by vec_name represents a solution from an adjoint (non-negative) or the primal (-1) space.

Definition at line 1160 of file system.C.

References libMesh::System::_vector_is_adjoint, and libMesh::libmesh_assert().

Referenced by libMesh::InterMeshProjection::project_system_vectors(), and libMesh::System::restrict_vectors().

{
  const auto it = _vector_is_adjoint.find(vec_name);
  libmesh_assert(it != _vector_is_adjoint.end());
  return it->second;
}

vector_name() [1/2]

const std::string & libMesh::System::vector_name ( const unsigned int  vec_num ) const

inherited

Returns

The name of this system's additional vector number vec_num (where the vectors are counted starting with 0).

Definition at line 971 of file system.C.

References libMesh::System::_vectors, and libMesh::System::vectors_begin().

Referenced by libMesh::AdjointRefinementEstimator::estimate_error(), and main().

{
  // If we don't have that many vectors, throw an error
  libmesh_assert_less(vec_num, _vectors.size());

  // Otherwise return a reference to the vec_num'th vector name
  auto it = vectors_begin();
  std::advance(it, vec_num);
  return it->first;
}

vector_name() [2/2]

const std::string & libMesh::System::vector_name ( const NumericVector< Number > &  vec_reference ) const

inherited

Returns

The name of a system vector, given a reference to that vector

Definition at line 982 of file system.C.

References libMesh::System::_vectors, libMesh::NumericVector< T >::get(), libMesh::libmesh_assert(), libMesh::System::vectors_begin(), and libMesh::System::vectors_end().

{
  // Linear search for a vector whose pointer matches vec_reference
  auto it = std::find_if(vectors_begin(), vectors_end(),
                         [&vec_reference](const decltype(_vectors)::value_type & pr)
                         { return &vec_reference == pr.second.get(); });

  // Before returning, make sure we didn't loop till the end and not find any match
  libmesh_assert (it != vectors_end());

  // Return the string associated with the current vector
  return it->first;
}

vector_preservation()

bool libMesh::System::vector_preservation ( std::string_view  vec_name ) const

inherited

Returns

The boolean describing whether the vector identified by vec_name should be "preserved": projected to new meshes, saved, etc.

Definition at line 1135 of file system.C.

References libMesh::System::_vector_projections.

Referenced by libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::MemoryHistoryData::store_vectors(), SystemsTest::testAddVectorProjChange(), SystemsTest::testAddVectorTypeChange(), and SystemsTest::testPostInitAddVectorTypeChange().

{
  if (auto it = _vector_projections.find(vec_name);
      it != _vector_projections.end())
    return it->second;

  // vec_name was not in the map, return false
  return false;
}

vectors_begin() [1/2]

System::vectors_iterator libMesh::System::vectors_begin ( )

inlineinherited

Beginning of vectors container.

Definition at line 2505 of file system.h.

References libMesh::System::_vectors.

Referenced by libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::System::get_vector(), libMesh::InterMeshProjection::project_system_vectors(), libMesh::System::request_vector(), libMesh::MemoryHistoryData::store_vectors(), and libMesh::System::vector_name().

{
  return _vectors.begin();
}

vectors_begin() [2/2]

System::const_vectors_iterator libMesh::System::vectors_begin ( ) const

inlineinherited

Beginning of vectors container.

Definition at line 2511 of file system.h.

References libMesh::System::_vectors.

{
  return _vectors.begin();
}

vectors_end() [1/2]

System::vectors_iterator libMesh::System::vectors_end ( )

inlineinherited

End of vectors container.

Definition at line 2517 of file system.h.

References libMesh::System::_vectors.

Referenced by libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::InterMeshProjection::project_system_vectors(), libMesh::MemoryHistoryData::store_vectors(), and libMesh::System::vector_name().

{
  return _vectors.end();
}

vectors_end() [2/2]

System::const_vectors_iterator libMesh::System::vectors_end ( ) const

inlineinherited

End of vectors container.

Definition at line 2523 of file system.h.

References libMesh::System::_vectors.

{
  return _vectors.end();
}

weighted_sensitivity_adjoint_solve()

std::pair< unsigned int, Real > libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve ( const ParameterVector &  parameters, const ParameterVector &  weights, const QoISet &  qoi_indices = QoISet()  )

overridevirtualinherited

Assembles & solves the linear system(s) (dR/du)^T*z_w = sum(w_p*(d^2q/dudp - d^2R/dudp*z)), for those parameters p contained within parameters, weighted by the values w_p found within weights.

Assumes that adjoint_solve has already calculated z for each qoi in qoi_indices.

Returns

A pair with the total number of linear iterations performed and the (sum of the) final residual norms

Reimplemented from libMesh::System.

Definition at line 247 of file implicit_system.C.

References libMesh::System::add_weighted_sensitivity_adjoint_solution(), libMesh::ExplicitSystem::assemble_qoi_derivative(), libMesh::ImplicitSystem::assembly(), libMesh::NumericVector< T >::close(), libMesh::SparseMatrix< T >::close(), libMesh::ParameterVector::deep_copy(), libMesh::DofMap::enforce_constraints_exactly(), libMesh::System::get_adjoint_rhs(), libMesh::System::get_adjoint_solution(), libMesh::System::get_dof_map(), libMesh::ImplicitSystem::get_linear_solve_parameters(), libMesh::ImplicitSystem::get_linear_solver(), libMesh::SparseMatrix< T >::get_transpose(), libMesh::System::get_weighted_sensitivity_adjoint_solution(), libMesh::DofMap::has_adjoint_dirichlet_boundaries(), libMesh::QoISet::has_index(), libMesh::libmesh_assert(), libMesh::make_range(), libMesh::ImplicitSystem::matrix, libMesh::System::n_qois(), libMesh::Real, libMesh::ExplicitSystem::rhs, libMesh::LinearSolver< T >::solve(), libMesh::TOLERANCE, libMesh::ParameterVector::value_copy(), libMesh::SparseMatrix< T >::vector_mult_add(), and libMesh::NumericVector< T >::zero_clone().

Referenced by libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product().

{
  // Log how long the linear solve takes.
  LOG_SCOPE("weighted_sensitivity_adjoint_solve()", "ImplicitSystem");

  // We currently get partial derivatives via central differencing
  const Real delta_p = TOLERANCE;

  ParameterVector & parameters_vec =
    const_cast<ParameterVector &>(parameters_in);

  // The forward system should now already be solved.
  // The adjoint system should now already be solved.
  // Now we're assembling a weighted sum of adjoint-adjoint systems:
  //
  // dR/du (u, sum_l(w_l*z^l)) = sum_l(w_l*(Q''_ul - R''_ul (u, z)))

  // FIXME: The derivation here does not yet take adjoint boundary
  // conditions into account.
#ifdef LIBMESH_ENABLE_DIRICHLET
  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      libmesh_assert(!this->get_dof_map().has_adjoint_dirichlet_boundaries(i));
#endif

  // We'll assemble the rhs first, because the R'' term will require
  // perturbing the jacobian

  // We'll use temporary rhs vectors, because we haven't (yet) found
  // any good reasons why users might want to save these:

  std::vector<std::unique_ptr<NumericVector<Number>>> temprhs(this->n_qois());
  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      temprhs[i] = this->rhs->zero_clone();

  // We approximate the _l partial derivatives via a central
  // differencing perturbation in the w_l direction:
  //
  // sum_l(w_l*v_l) ~= (v(p + dp*w_l*e_l) - v(p - dp*w_l*e_l))/(2*dp)

  // PETSc doesn't implement SGEMX, so neither does NumericVector,
  // so we want to avoid calculating f -= R'*z.  We'll thus evaluate
  // the above equation by first adding -v(p+dp...), then multiplying
  // the intermediate result vectors by -1, then adding -v(p-dp...),
  // then finally dividing by 2*dp.

  ParameterVector oldparameters, parameterperturbation;
  parameters_vec.deep_copy(oldparameters);
  weights.deep_copy(parameterperturbation);
  parameterperturbation *= delta_p;
  parameters_vec += parameterperturbation;

  this->assembly(false, true);
  this->matrix->close();

  // Take the discrete adjoint, so that we can calculate R_u(u,z) with
  // a matrix-vector product of R_u and z.
  matrix->get_transpose(*matrix);

  this->assemble_qoi_derivative(qoi_indices,
                                /* include_liftfunc = */ false,
                                /* apply_constraints = */ true);
  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      {
        this->get_adjoint_rhs(i).close();
        *(temprhs[i]) -= this->get_adjoint_rhs(i);
        this->matrix->vector_mult_add(*(temprhs[i]), this->get_adjoint_solution(i));
        *(temprhs[i]) *= -1.0;
      }

  oldparameters.value_copy(parameters_vec);
  parameterperturbation *= -1.0;
  parameters_vec += parameterperturbation;

  this->assembly(false, true);
  this->matrix->close();
  matrix->get_transpose(*matrix);

  this->assemble_qoi_derivative(qoi_indices,
                                /* include_liftfunc = */ false,
                                /* apply_constraints = */ true);
  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      {
        this->get_adjoint_rhs(i).close();
        *(temprhs[i]) -= this->get_adjoint_rhs(i);
        this->matrix->vector_mult_add(*(temprhs[i]), this->get_adjoint_solution(i));
        *(temprhs[i]) /= (2.0*delta_p);
      }

  // Finally, assemble the jacobian at the non-perturbed parameter
  // values.  Ignore assemble_before_solve; if we had a good
  // non-perturbed matrix before we've already overwritten it.
  oldparameters.value_copy(parameters_vec);

  // if (this->assemble_before_solve)
  {
    // Build the Jacobian
    this->assembly(false, true);
    this->matrix->close();

    // Take the discrete adjoint
    matrix->get_transpose(*matrix);
  }

  // The weighted adjoint-adjoint problem is linear
  LinearSolver<Number> * solver = this->get_linear_solver();

  // Our iteration counts and residuals will be sums of the individual
  // results
  std::pair<unsigned int, Real> solver_params =
    this->get_linear_solve_parameters();
  std::pair<unsigned int, Real> totalrval = std::make_pair(0,0.0);

  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      {
        const std::pair<unsigned int, Real> rval =
          solver->solve (*matrix, this->add_weighted_sensitivity_adjoint_solution(i),
                         *(temprhs[i]),
                         double(solver_params.second),
                         solver_params.first);

        totalrval.first  += rval.first;
        totalrval.second += rval.second;
      }

  // The linear solver may not have fit our constraints exactly
#ifdef LIBMESH_ENABLE_CONSTRAINTS
  for (auto i : make_range(this->n_qois()))
    if (qoi_indices.has_index(i))
      this->get_dof_map().enforce_constraints_exactly
        (*this, &this->get_weighted_sensitivity_adjoint_solution(i),
         /* homogeneous = */ true);
#endif

  return totalrval;
}

weighted_sensitivity_solve()

std::pair< unsigned int, Real > libMesh::ImplicitSystem::weighted_sensitivity_solve ( const ParameterVector &  parameters, const ParameterVector &  weights  )

overridevirtualinherited

Assembles & solves the linear system(s) (dR/du)*u_w = sum(w_p*-dR/dp), for those parameters p contained within parameters weighted by the values w_p found within weights.

Returns

A pair with the total number of linear iterations performed and the (sum of the) final residual norms

Reimplemented from libMesh::System.

Definition at line 393 of file implicit_system.C.

References libMesh::System::add_weighted_sensitivity_solution(), libMesh::ImplicitSystem::assembly(), libMesh::NumericVector< T >::clone(), libMesh::NumericVector< T >::close(), libMesh::SparseMatrix< T >::close(), libMesh::ParameterVector::deep_copy(), libMesh::DofMap::enforce_constraints_exactly(), libMesh::System::get_dof_map(), libMesh::ImplicitSystem::get_linear_solve_parameters(), libMesh::ImplicitSystem::get_linear_solver(), libMesh::System::get_weighted_sensitivity_solution(), libMesh::ImplicitSystem::matrix, libMesh::Real, libMesh::ExplicitSystem::rhs, libMesh::LinearSolver< T >::solve(), libMesh::TOLERANCE, and libMesh::ParameterVector::value_copy().

Referenced by libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product().

{
  // Log how long the linear solve takes.
  LOG_SCOPE("weighted_sensitivity_solve()", "ImplicitSystem");

  // We currently get partial derivatives via central differencing
  const Real delta_p = TOLERANCE;

  ParameterVector & parameters_vec =
    const_cast<ParameterVector &>(parameters_in);

  // The forward system should now already be solved.

  // Now we're assembling a weighted sum of sensitivity systems:
  //
  // dR/du (u, v)(sum(w_l*u'_l)) = -sum_l(w_l*R'_l (u, v)) forall v

  // We'll assemble the rhs first, because the R' term will require
  // perturbing the system, and some applications may not be able to
  // assemble a perturbed residual without simultaneously constructing
  // a perturbed jacobian.

  // We approximate the _l partial derivatives via a central
  // differencing perturbation in the w_l direction:
  //
  // sum_l(w_l*v_l) ~= (v(p + dp*w_l*e_l) - v(p - dp*w_l*e_l))/(2*dp)

  ParameterVector oldparameters, parameterperturbation;
  parameters_vec.deep_copy(oldparameters);
  weights.deep_copy(parameterperturbation);
  parameterperturbation *= delta_p;
  parameters_vec += parameterperturbation;

  this->assembly(true, false, true);
  this->rhs->close();

  std::unique_ptr<NumericVector<Number>> temprhs = this->rhs->clone();

  oldparameters.value_copy(parameters_vec);
  parameterperturbation *= -1.0;
  parameters_vec += parameterperturbation;

  this->assembly(true, false, true);
  this->rhs->close();

  *temprhs -= *(this->rhs);
  *temprhs /= (2.0*delta_p);

  // Finally, assemble the jacobian at the non-perturbed parameter
  // values
  oldparameters.value_copy(parameters_vec);

  // Build the Jacobian
  this->assembly(false, true);
  this->matrix->close();

  // The weighted sensitivity problem is linear
  LinearSolver<Number> * solver = this->get_linear_solver();

  std::pair<unsigned int, Real> solver_params =
    this->get_linear_solve_parameters();

  const std::pair<unsigned int, Real> rval =
    solver->solve (*matrix, this->add_weighted_sensitivity_solution(),
                   *temprhs,
                   double(solver_params.second),
                   solver_params.first);

  // The linear solver may not have fit our constraints exactly
#ifdef LIBMESH_ENABLE_CONSTRAINTS
  this->get_dof_map().enforce_constraints_exactly
    (*this, &this->get_weighted_sensitivity_solution(),
     /* homogeneous = */ true);
#endif

  return rval;
}

write_header()

void libMesh::System::write_header ( Xdr &  io, std::string_view  version, const bool  write_additional_data  ) const

inherited

Writes the basic data header for this System.

This method implements the output of a System object, embedded in the output of an EquationSystems<T_sys>. This warrants some documentation. The output of this part consists of 5 sections:

for this system

5.) The number of variables in the system (unsigned int)

for each variable in the system

6.) The name of the variable (string)

6.1.) subdomain where the variable lives

7.) Combined in an FEType:

• The approximation order(s) of the variable (Order Enum, cast to int/s)
• The finite element family/ies of the variable (FEFamily Enum, cast to int/s)

end variable loop

8.) The number of additional vectors (unsigned int),

for each additional vector in the system object

9.) the name of the additional vector (string)

end system

Definition at line 1120 of file system_io.C.

References libMesh::System::_vector_projections, libMesh::System::_vectors, libMesh::Variable::active_subdomains(), libMesh::Xdr::data(), libMesh::FEType::family, libMesh::System::get_mesh(), libMesh::FEType::inf_map, libMesh::libmesh_assert(), libMesh::make_range(), libMesh::System::n_vars(), libMesh::System::n_vectors(), libMesh::System::name(), libMesh::FEType::order, libMesh::ParallelObject::processor_id(), libMesh::FEType::radial_family, libMesh::FEType::radial_order, libMesh::System::variable(), libMesh::System::variable_name(), libMesh::System::variable_type(), and libMesh::Xdr::writing().

Referenced by libMesh::RBEvaluation::write_out_vectors().

{
  libmesh_assert (io.writing());


  // Only write the header information
  // if we are processor 0.
  if (this->get_mesh().processor_id() != 0)
    return;

  std::string comment;

  // 5.)
  // Write the number of variables in the system

  {
    // set up the comment
    comment = "# No. of Variables in System \"";
    comment += this->name();
    comment += "\"";

    unsigned int nv = this->n_vars();
    io.data (nv, comment);
  }


  for (auto var : make_range(this->n_vars()))
    {
      // 6.)
      // Write the name of the var-th variable
      {
        // set up the comment
        comment  = "#   Name, Variable No. ";
        comment += std::to_string(var);
        comment += ", System \"";
        comment += this->name();
        comment += "\"";

        std::string var_name = this->variable_name(var);
        io.data (var_name, comment);
      }

      // 6.1.) Variable subdomains
      {
        // set up the comment
        comment  = "#     Subdomains, Variable \"";
        comment += this->variable_name(var);
        comment += "\", System \"";
        comment += this->name();
        comment += "\"";

        const std::set<subdomain_id_type> & domains = this->variable(var).active_subdomains();
        std::vector<subdomain_id_type> domain_array;
        domain_array.assign(domains.begin(), domains.end());
        io.data (domain_array, comment);
      }

      // 7.)
      // Write the approximation order of the var-th variable
      // in this system
      {
        // set up the comment
        comment = "#     Approximation Order, Variable \"";
        comment += this->variable_name(var);
        comment += "\", System \"";
        comment += this->name();
        comment += "\"";

        int order = static_cast<int>(this->variable_type(var).order);
        io.data (order, comment);
      }


#ifdef LIBMESH_ENABLE_INFINITE_ELEMENTS

      // do the same for radial_order
      {
        comment = "#     Radial Approximation Order, Variable \"";
        comment += this->variable_name(var);
        comment += "\", System \"";
        comment += this->name();
        comment += "\"";

        int rad_order = static_cast<int>(this->variable_type(var).radial_order);
        io.data (rad_order, comment);
      }

#endif

      // Write the Finite Element type of the var-th variable
      // in this System
      {
        // set up the comment
        comment = "#     FE Family, Variable \"";
        comment += this->variable_name(var);
        comment += "\", System \"";
        comment += this->name();
        comment += "\"";

        const FEType & type = this->variable_type(var);
        int fam = static_cast<int>(type.family);
        io.data (fam, comment);

#ifdef LIBMESH_ENABLE_INFINITE_ELEMENTS

        comment = "#     Radial FE Family, Variable \"";
        comment += this->variable_name(var);
        comment += "\", System \"";
        comment += this->name();
        comment += "\"";

        int radial_fam = static_cast<int>(type.radial_family);
        io.data (radial_fam, comment);

        comment = "#     Infinite Mapping Type, Variable \"";
        comment += this->variable_name(var);
        comment += "\", System \"";
        comment += this->name();
        comment += "\"";

        int i_map = static_cast<int>(type.inf_map);
        io.data (i_map, comment);
#endif
      }
    } // end of the variable loop

  // 8.)
  // Write the number of additional vectors in the System.
  // If write_additional_data==false, then write zero for
  // the number of additional vectors.
  {
    {
      // set up the comment
      comment = "# No. of Additional Vectors, System \"";
      comment += this->name();
      comment += "\"";

      unsigned int nvecs = write_additional_data ? this->n_vectors () : 0;
      io.data (nvecs, comment);
    }

    if (write_additional_data)
      {
        unsigned int cnt=0;
        for (const auto & [vec_name, vec] : _vectors)
          {
            // 9.)
            // write the name of the cnt-th additional vector
            const std::string dth_vector = std::to_string(cnt++)+"th vector";
            comment =  "# Name of " + dth_vector;
            std::string nonconst_vec_name = vec_name; // Stupid XDR API

            io.data (nonconst_vec_name, comment);
            int vec_projection = _vector_projections.at(vec_name);
            comment = "# Whether to do projections for " + dth_vector;
            io.data (vec_projection, comment);
            int vec_type = vec->type();
            comment = "# Parallel type of " + dth_vector;
            io.data (vec_type, comment);
          }
      }
  }
}

write_parallel_data()

void libMesh::System::write_parallel_data ( Xdr &  io, const bool  write_additional_data  ) const

inherited

Writes additional data, namely vectors, for this System.

This method may safely be called on a distributed-memory mesh. This method will create an individual file for each processor in the simulation where the local solution components for that processor will be stored.

This method implements the output of the vectors contained in this System object, embedded in the output of an EquationSystems<T_sys>.

9.) The global solution vector, re-ordered to be node-major (More on this later.)

for each additional vector in the object

10.) The global additional vector, re-ordered to be node-major (More on this later.)

Note that the actual IO is handled through the Xdr class (to be renamed later?) which provides a uniform interface to both the XDR (eXternal Data Representation) interface and standard ASCII output. Thus this one section of code will read XDR or ASCII files with no changes.

Definition at line 1321 of file system_io.C.

References libMesh::System::_vectors, libMesh::Xdr::data(), libMesh::FEType::family, libMesh::System::get_dof_map(), libMesh::System::get_mesh(), libMesh::DofObject::invalid_id, libMesh::libmesh_assert(), libMesh::make_range(), libMesh::ParallelObject::n_processors(), libMesh::System::n_vars(), libMesh::System::name(), libMesh::System::number(), libMesh::ParallelObject::processor_id(), libMesh::SCALAR, libMesh::DofMap::SCALAR_dof_indices(), libMesh::System::solution, libMesh::Variable::type(), libMesh::System::variable(), and libMesh::Xdr::writing().

{
  // PerfLog pl("IO Performance",false);
  // pl.push("write_parallel_data");
  // std::size_t total_written_size = 0;

  std::string comment;

  libmesh_assert (io.writing());

  std::vector<Number> io_buffer; io_buffer.reserve(this->solution->local_size());

  // build the ordered nodes and element maps.
  // when writing/reading parallel files we need to iterate
  // over our nodes/elements in order of increasing global id().
  // however, this is not guaranteed to be ordering we obtain
  // by using the node_iterators/element_iterators directly.
  // so build a set, sorted by id(), that provides the ordering.
  // further, for memory economy build the set but then transfer
  // its contents to vectors, which will be sorted.
  std::vector<const DofObject *> ordered_nodes, ordered_elements;
  {
    std::set<const DofObject *, CompareDofObjectsByID>
      ordered_nodes_set (this->get_mesh().local_nodes_begin(),
                         this->get_mesh().local_nodes_end());

    ordered_nodes.insert(ordered_nodes.end(),
                         ordered_nodes_set.begin(),
                         ordered_nodes_set.end());
  }
  {
    std::set<const DofObject *, CompareDofObjectsByID>
      ordered_elements_set (this->get_mesh().local_elements_begin(),
                            this->get_mesh().local_elements_end());

    ordered_elements.insert(ordered_elements.end(),
                            ordered_elements_set.begin(),
                            ordered_elements_set.end());
  }

  const unsigned int sys_num = this->number();
  const unsigned int nv      = this->n_vars();

  // Loop over each non-SCALAR variable and each node, and write out the value.
  for (unsigned int var=0; var<nv; var++)
    if (this->variable(var).type().family != SCALAR)
      {
        // First write the node DOF values
        for (const auto & node : ordered_nodes)
          for (auto comp : make_range(node->n_comp(sys_num,var)))
            {
              libmesh_assert_not_equal_to (node->dof_number(sys_num, var, comp),
                                           DofObject::invalid_id);

              io_buffer.push_back((*this->solution)(node->dof_number(sys_num, var, comp)));
            }

        // Then write the element DOF values
        for (const auto & elem : ordered_elements)
          for (auto comp : make_range(elem->n_comp(sys_num,var)))
            {
              libmesh_assert_not_equal_to (elem->dof_number(sys_num, var, comp),
                                           DofObject::invalid_id);

              io_buffer.push_back((*this->solution)(elem->dof_number(sys_num, var, comp)));
            }
      }

  // Finally, write the SCALAR data on the last processor
  for (auto var : make_range(this->n_vars()))
    if (this->variable(var).type().family == SCALAR)
      {
        if (this->processor_id() == (this->n_processors()-1))
          {
            const DofMap & dof_map = this->get_dof_map();
            std::vector<dof_id_type> SCALAR_dofs;
            dof_map.SCALAR_dof_indices(SCALAR_dofs, var);

            for (auto dof : SCALAR_dofs)
              io_buffer.push_back((*this->solution)(dof));
          }
      }

  // 9.)
  //
  // Actually write the reordered solution vector
  // for the ith system to disk

  // set up the comment
  {
    comment = "# System \"";
    comment += this->name();
    comment += "\" Solution Vector";
  }

  io.data (io_buffer, comment);

  // total_written_size += io_buffer.size();

  // Only write additional vectors if wanted
  if (write_additional_data)
    {
      for (auto & [vec_name, vec] : _vectors)
        {
          io_buffer.clear();
          io_buffer.reserve(vec->local_size());

          // Loop over each non-SCALAR variable and each node, and write out the value.
          for (unsigned int var=0; var<nv; var++)
            if (this->variable(var).type().family != SCALAR)
              {
                // First write the node DOF values
                for (const auto & node : ordered_nodes)
                  for (auto comp : make_range(node->n_comp(sys_num,var)))
                    {
                      libmesh_assert_not_equal_to (node->dof_number(sys_num, var, comp),
                                                   DofObject::invalid_id);

                      io_buffer.push_back((*vec)(node->dof_number(sys_num, var, comp)));
                    }

                // Then write the element DOF values
                for (const auto & elem : ordered_elements)
                  for (auto comp : make_range(elem->n_comp(sys_num,var)))
                    {
                      libmesh_assert_not_equal_to (elem->dof_number(sys_num, var, comp),
                                                   DofObject::invalid_id);

                      io_buffer.push_back((*vec)(elem->dof_number(sys_num, var, comp)));
                    }
              }

          // Finally, write the SCALAR data on the last processor
          for (auto var : make_range(this->n_vars()))
            if (this->variable(var).type().family == SCALAR)
              {
                if (this->processor_id() == (this->n_processors()-1))
                  {
                    const DofMap & dof_map = this->get_dof_map();
                    std::vector<dof_id_type> SCALAR_dofs;
                    dof_map.SCALAR_dof_indices(SCALAR_dofs, var);

                    for (auto dof : SCALAR_dofs)
                      io_buffer.push_back((*vec)(dof));
                  }
              }

          // 10.)
          //
          // Actually write the reordered additional vector
          // for this system to disk

          // set up the comment
          {
            comment = "# System \"";
            comment += this->name();
            comment += "\" Additional Vector \"";
            comment += vec_name;
            comment += "\"";
          }

          io.data (io_buffer, comment);

          // total_written_size += io_buffer.size();
        }
    }

  // const Real
  //   dt   = pl.get_elapsed_time(),
  //   rate = total_written_size*sizeof(Number)/dt;

  // libMesh::err << "Write " << total_written_size << " \"Number\" values\n"
  //     << " Elapsed time = " << dt << '\n'
  //     << " Rate = " << rate/1.e6 << "(MB/sec)\n\n";

  // pl.pop("write_parallel_data");
}

write_parameter_data_to_files()

void libMesh::RBParametrized::write_parameter_data_to_files ( const std::string &  continuous_param_file_name, const std::string &  discrete_param_file_name, const bool  write_binary_data  )

inherited

Write out the parameter ranges to files.

Definition at line 185 of file rb_parametrized.C.

References libMesh::RBParametrized::write_discrete_parameter_values_to_file(), and libMesh::RBParametrized::write_parameter_ranges_to_file().

Referenced by libMesh::RBSCMEvaluation::legacy_write_offline_data_to_files(), and libMesh::RBEvaluation::legacy_write_offline_data_to_files().

{
  write_parameter_ranges_to_file(continuous_param_file_name, write_binary_data);
  write_discrete_parameter_values_to_file(discrete_param_file_name, write_binary_data);
}

write_riesz_representors_to_files()

void libMesh::RBConstruction::write_riesz_representors_to_files ( const std::string &  riesz_representors_dir, const bool  write_binary_residual_representors  )

virtual

Write out all the Riesz representor data to files.

residual_representors_dir specifies which directory to write to. write_binary_residual_representors specifies whether we write binary or ASCII data.

Reimplemented in libMesh::TransientRBConstruction.

Definition at line 2540 of file rb_construction.C.

References libMesh::RBEvaluation::Aq_representor, TIMPI::Communicator::barrier(), libMesh::ParallelObject::comm(), libMesh::ENCODE, Fq_representor, get_delta_N(), libMesh::RBEvaluation::get_n_basis_functions(), get_rb_evaluation(), libMesh::index_range(), libMesh::libmesh_assert(), libMesh::Utility::mkdir(), libMesh::out, libMesh::ParallelObject::processor_id(), libMesh::System::solution, libMesh::WRITE, and libMesh::System::write_serialized_data().

{
  LOG_SCOPE("write_riesz_representors_to_files()", "RBConstruction");

  // Write out Riesz representors. These are useful to have when restarting,
  // so you don't have to recompute them all over again.

  // First we write out the Fq representors, these are independent of an RBEvaluation object.
  libMesh::out << "Writing out the Fq_representors..." << std::endl;

  std::ostringstream file_name;
  const std::string riesz_representor_suffix = (write_binary_residual_representors ? ".xdr" : ".dat");
  struct stat stat_info;

  // Residual representors written out to their own separate directory
  if ( this->processor_id() == 0)
    if ( Utility::mkdir(riesz_representors_dir.c_str()) != 0)
      libMesh::out << "Skipping creating residual_representors directory: " << strerror(errno) << std::endl;

  for (auto i : index_range(Fq_representor))
    {
      if (Fq_representor[i])
        {
          file_name.str(""); // reset filename
          file_name << riesz_representors_dir << "/Fq_representor" << i << riesz_representor_suffix;

          // Check to see if file exists, if so, don't overwrite it, we assume it was
          // there from a previous call to this function.  Note: if stat returns zero
          // it means it successfully got the file attributes (and therefore the file
          // exists).  Because of the following factors:
          // 1.) write_serialized_data takes longer for proc 0 than it does for others,
          //     so processors can get out of sync.
          // 2.) The constructor for Xdr opens a (0 length) file on *all* processors,
          // there are typically hundreds of 0-length files created during this loop,
          // and that screws up checking for the existence of files.  One way to stay
          // in sync is to but a barrier at each iteration of the loop -- not sure how
          // bad this will affect performance, but it can't be much worse than serialized
          // I/O already is :)
          int stat_result = stat(file_name.str().c_str(), &stat_info);

          if ( (stat_result != 0) ||     // file definitely doesn't already exist
               (stat_info.st_size == 0)) // file exists, but has zero length (can happen if another proc already opened it!)
            {
              // No need to copy!
              // *solution = *(Fq_representor[i]);
              // std::swap doesn't work on pointers
              //std::swap(solution.get(), Fq_representor[i]);
              Fq_representor[i]->swap(*solution);

              Xdr fqr_data(file_name.str(),
                           write_binary_residual_representors ? ENCODE : WRITE);

              write_serialized_data(fqr_data, false);

              // Synchronize before moving on
              this->comm().barrier();

              // Swap back.
              Fq_representor[i]->swap(*solution);

              // TODO: bzip the resulting file?  See $LIBMESH_DIR/src/mesh/unstructured_mesh.C
              // for the system call, be sure to do it only on one processor, etc.
            }
        }
    }


  // Next, write out the Aq representors associated with rb_eval.
  libMesh::out << "Writing out the Aq_representors..." << std::endl;

  const unsigned int jstop  = get_rb_evaluation().get_n_basis_functions();
  const unsigned int jstart = jstop-get_delta_N();

  RBEvaluation & rbe = get_rb_evaluation();
  for (auto i : index_range(rbe.Aq_representor))
    for (unsigned int j=jstart; j<jstop; ++j)
      {
        libMesh::out << "Writing out Aq_representor[" << i << "][" << j << "]..." << std::endl;
        libmesh_assert(get_rb_evaluation().Aq_representor[i][j]);

        file_name.str(""); // reset filename
        file_name << riesz_representors_dir
                  << "/Aq_representor" << i << "_" << j << riesz_representor_suffix;

        {
          // No need to copy! Use swap instead.
          // *solution = *(Aq_representor[i][j]);
          rbe.Aq_representor[i][j]->swap(*solution);

          Xdr aqr_data(file_name.str(),
                       write_binary_residual_representors ? ENCODE : WRITE);

          write_serialized_data(aqr_data, false);

          // Synchronize before moving on
          this->comm().barrier();

          // Swap back.
          rbe.Aq_representor[i][j]->swap(*solution);

          // TODO: bzip the resulting file?  See $LIBMESH_DIR/src/mesh/unstructured_mesh.C
          // for the system call, be sure to do it only on one processor, etc.
        }
      }
}

write_serialized_data()

void libMesh::System::write_serialized_data ( Xdr &  io, const bool  write_additional_data = true  ) const

inherited

Writes additional data, namely vectors, for this System.

This method may safely be called on a distributed-memory mesh.

This method implements the output of the vectors contained in this System object, embedded in the output of an EquationSystems<T_sys>.

9.) The global solution vector, re-ordered to be node-major (More on this later.)

for each additional vector in the object

10.) The global additional vector, re-ordered to be node-major (More on this later.)

Definition at line 1521 of file system_io.C.

References libMesh::System::_vectors, libMesh::Xdr::comment(), libMesh::System::name(), libMesh::ParallelObject::processor_id(), libMesh::System::solution, and libMesh::System::write_serialized_vector().

Referenced by libMesh::TransientRBConstruction::write_riesz_representors_to_files(), and write_riesz_representors_to_files().

{
  parallel_object_only();
  std::string comment;

  // PerfLog pl("IO Performance",false);
  // pl.push("write_serialized_data");
  // std::size_t total_written_size = 0;

  // total_written_size +=
  this->write_serialized_vector(io, *this->solution);

  // set up the comment
  if (this->processor_id() == 0)
    {
      comment = "# System \"";
      comment += this->name();
      comment += "\" Solution Vector";

      io.comment (comment);
    }

  // Only write additional vectors if wanted
  if (write_additional_data)
    {
      for (auto & pair : this->_vectors)
        {
          // total_written_size +=
          this->write_serialized_vector(io, *pair.second);

          // set up the comment
          if (this->processor_id() == 0)
            {
              comment = "# System \"";
              comment += this->name();
              comment += "\" Additional Vector \"";
              comment += pair.first;
              comment += "\"";
              io.comment (comment);
            }
        }
    }

  // const Real
  //   dt   = pl.get_elapsed_time(),
  //   rate = total_written_size*sizeof(Number)/dt;

  // libMesh::out << "Write " << total_written_size << " \"Number\" values\n"
  //     << " Elapsed time = " << dt << '\n'
  //     << " Rate = " << rate/1.e6 << "(MB/sec)\n\n";

  // pl.pop("write_serialized_data");




  // // test the new method
  // {
  //   std::vector<std::string> names;
  //   std::vector<NumericVector<Number> *> vectors_to_write;

  //   names.push_back("Solution Vector");
  //   vectors_to_write.push_back(this->solution.get());

  //   // Only write additional vectors if wanted
  //   if (write_additional_data)
  //     {
  // std::map<std::string, NumericVector<Number> *>::const_iterator
  //   pos = _vectors.begin();

  // for (; pos != this->_vectors.end(); ++pos)
  //   {
  //     names.push_back("Additional Vector " + pos->first);
  //     vectors_to_write.push_back(pos->second);
  //   }
  //     }

  //   total_written_size =
  //     this->write_serialized_vectors (io, names, vectors_to_write);

  //   const Real
  //     dt2   = pl.get_elapsed_time(),
  //     rate2 = total_written_size*sizeof(Number)/(dt2-dt);

  //   libMesh::out << "Write (new) " << total_written_size << " \"Number\" values\n"
  //       << " Elapsed time = " << (dt2-dt) << '\n'
  //       << " Rate = " << rate2/1.e6 << "(MB/sec)\n\n";

  // }
}

write_serialized_vectors()

std::size_t libMesh::System::write_serialized_vectors ( Xdr &  io, const std::vector< const NumericVector< Number > *> &  vectors  ) const

inherited

Serialize & write a number of identically distributed vectors.

This method allows for optimization for the multiple vector case by only communicating the metadata once.

Definition at line 2114 of file system_io.C.

References libMesh::Xdr::data(), libMesh::System::get_mesh(), libMesh::libmesh_assert(), libMesh::make_range(), libMesh::MeshTools::n_elem(), libMesh::MeshBase::n_elem(), n_nodes, libMesh::MeshBase::n_nodes(), libMesh::System::n_vars(), libMesh::ParallelObject::processor_id(), libMesh::SCALAR, libMesh::System::variable(), libMesh::System::write_SCALAR_dofs(), libMesh::System::write_serialized_blocked_dof_objects(), and libMesh::Xdr::writing().

Referenced by libMesh::RBEvaluation::write_out_vectors().

{
  parallel_object_only();

  libmesh_assert (io.writing());

  // Cache these - they are not free!
  const dof_id_type
    n_nodes       = this->get_mesh().n_nodes(),
    n_elem        = this->get_mesh().n_elem();

  std::size_t written_length = 0;

  if (this->processor_id() == 0)
    {
      unsigned int
        n_vec    = cast_int<unsigned int>(vectors.size());
      dof_id_type
        vec_size = vectors.empty() ? 0 : vectors[0]->size();
      // Set the number of vectors
      io.data(n_vec, "# number of vectors");
      // Set the buffer size
      io.data(vec_size, "# vector length");
    }

  //---------------------------------
  // Collect the values for all nodes
  written_length +=
    this->write_serialized_blocked_dof_objects (vectors,
                                                n_nodes,
                                                this->get_mesh().local_nodes_begin(),
                                                this->get_mesh().local_nodes_end(),
                                                io);

  //------------------------------------
  // Collect the values for all elements
  written_length +=
    this->write_serialized_blocked_dof_objects (vectors,
                                                n_elem,
                                                this->get_mesh().local_elements_begin(),
                                                this->get_mesh().local_elements_end(),
                                                io);

  //-------------------------------------------
  // Finally loop over all the SCALAR variables
  for (const NumericVector<Number> * vec : vectors)
    for (auto var : make_range(this->n_vars()))
      if (this->variable(var).type().family == SCALAR)
        {
          libmesh_assert_not_equal_to (vec, 0);

          written_length +=
            this->write_SCALAR_dofs (*vec, var, io);
        }

  return written_length;
}

zero_constrained_dofs_on_vector()

void libMesh::RBConstruction::zero_constrained_dofs_on_vector

(

NumericVector< Number > &

vector

)

const

It is sometimes useful to be able to zero vector entries that correspond to constrained dofs.

Definition at line 444 of file rb_construction.C.

References libMesh::NumericVector< T >::close(), libMesh::DofMapBase::end_dof(), libMesh::DofMapBase::first_dof(), libMesh::System::get_dof_map(), libMesh::DofMap::is_constrained_dof(), and libMesh::NumericVector< T >::set().

{
#ifdef LIBMESH_ENABLE_CONSTRAINTS
  const DofMap & dof_map = get_dof_map();

  for (dof_id_type i=dof_map.first_dof(); i<dof_map.end_dof(); i++)
    {
      if (get_dof_map().is_constrained_dof(i))
        {
          vector.set(i, 0.);
        }
    }
#endif

  vector.close();
}

zero_variable()

void libMesh::System::zero_variable ( NumericVector< Number > &  v, unsigned int  var_num  ) const

inherited

Zeroes all dofs in v that correspond to variable number var_num.

Definition at line 1450 of file system.C.

References libMesh::System::get_mesh(), mesh, libMesh::System::n_vars(), libMesh::System::number(), and libMesh::NumericVector< T >::set().

{
  /* Make sure the call makes sense.  */
  libmesh_assert_less (var_num, this->n_vars());

  /* Get a reference to the mesh.  */
  const MeshBase & mesh = this->get_mesh();

  /* Check which system we are.  */
  const unsigned int sys_num = this->number();

  // Loop over nodes.
  for (const auto & node : mesh.local_node_ptr_range())
    {
      unsigned int n_comp = node->n_comp(sys_num,var_num);
      for (unsigned int i=0; i<n_comp; i++)
        {
          const dof_id_type index = node->dof_number(sys_num,var_num,i);
          v.set(index,0.0);
        }
    }

  // Loop over elements.
  for (const auto & elem : mesh.active_local_element_ptr_range())
    {
      unsigned int n_comp = elem->n_comp(sys_num,var_num);
      for (unsigned int i=0; i<n_comp; i++)
        {
          const dof_id_type index = elem->dof_number(sys_num,var_num,i);
          v.set(index,0.0);
        }
    }
}

Member Data Documentation

_communicator

const Parallel::Communicator& libMesh::ParallelObject::_communicator

protectedinherited

Definition at line 120 of file parallel_object.h.

Referenced by libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::StaticCondensation::close(), libMesh::ParallelObject::comm(), libMesh::CondensedEigenSystem::initialize_condensed_matrices(), libMesh::ParallelObject::n_processors(), libMesh::ParallelObject::operator=(), libMesh::ParallelObject::processor_id(), libMesh::BoundaryInfo::regenerate_id_sets(), libMesh::StaticCondensationDofMap::reinit(), and libMesh::BoundaryInfo::synchronize_global_id_set().

_counts [1/2]

ReferenceCounter::Counts libMesh::ReferenceCounter::_counts

staticprotectedinherited

Actually holds the data.

Definition at line 124 of file reference_counter.h.

Referenced by libMesh::ReferenceCounter::get_info().

_counts [2/2]

ReferenceCounter::Counts libMesh::ReferenceCounter::_counts

staticprotectedinherited

Actually holds the data.

Definition at line 124 of file reference_counter.h.

Referenced by libMesh::ReferenceCounter::get_info().

_current_training_parameter_index

unsigned int libMesh::RBConstruction::_current_training_parameter_index

private

The current training parameter index during reduced basis training.

Definition at line 1011 of file rb_construction.h.

Referenced by get_current_training_parameter_index(), and set_current_training_parameter_index().

_enable_print_counter [1/2]

bool libMesh::ReferenceCounter::_enable_print_counter = true

staticprotectedinherited

Flag to control whether reference count information is printed when print_info is called.

Definition at line 143 of file reference_counter.h.

Referenced by libMesh::ReferenceCounter::disable_print_counter_info(), libMesh::ReferenceCounter::enable_print_counter_info(), and libMesh::ReferenceCounter::print_info().

_enable_print_counter [2/2]

bool libMesh::ReferenceCounter::_enable_print_counter = true

staticprotectedinherited

Flag to control whether reference count information is printed when print_info is called.

Definition at line 143 of file reference_counter.h.

Referenced by libMesh::ReferenceCounter::disable_print_counter_info(), libMesh::ReferenceCounter::enable_print_counter_info(), and libMesh::ReferenceCounter::print_info().

_evaluated_thetas

std::vector<std::vector<Number> > libMesh::RBConstruction::_evaluated_thetas

private

Storage of evaluated theta functions at a set of parameters.

This can be used to store all of our theta functions at training samples instead of re-evaluating the same values repeatedly during training.

Definition at line 1018 of file rb_construction.h.

Referenced by get_evaluated_thetas(), and preevaluate_thetas().

_final_linear_residual

Real libMesh::LinearImplicitSystem::_final_linear_residual

protectedinherited

The final residual for the linear system Ax=b.

Definition at line 197 of file linear_implicit_system.h.

Referenced by libMesh::LinearImplicitSystem::final_linear_residual(), libMesh::FrequencySystem::solve(), libMesh::LinearImplicitSystem::solve(), and solve_for_matrix_and_rhs().

_mutex [1/2]

Threads::spin_mutex libMesh::ReferenceCounter::_mutex

staticprotectedinherited

Mutual exclusion object to enable thread-safe reference counting.

Definition at line 137 of file reference_counter.h.

_mutex [2/2]

Threads::spin_mutex libMesh::ReferenceCounter::_mutex

staticprotectedinherited

Mutual exclusion object to enable thread-safe reference counting.

Definition at line 137 of file reference_counter.h.

_n_linear_iterations

unsigned int libMesh::LinearImplicitSystem::_n_linear_iterations

protectedinherited

The number of linear iterations required to solve the linear system Ax=b.

Definition at line 192 of file linear_implicit_system.h.

Referenced by libMesh::LinearImplicitSystem::n_linear_iterations(), libMesh::FrequencySystem::solve(), libMesh::LinearImplicitSystem::solve(), and solve_for_matrix_and_rhs().

_n_objects [1/2]

Threads::atomic< unsigned int > libMesh::ReferenceCounter::_n_objects

staticprotectedinherited

The number of objects.

Print the reference count information when the number returns to 0.

Definition at line 132 of file reference_counter.h.

Referenced by libMesh::ReferenceCounter::n_objects(), libMesh::ReferenceCounter::ReferenceCounter(), and libMesh::ReferenceCounter::~ReferenceCounter().

_n_objects [2/2]

Threads::atomic< unsigned int > libMesh::ReferenceCounter::_n_objects

staticprotectedinherited

The number of objects.

Print the reference count information when the number returns to 0.

Definition at line 132 of file reference_counter.h.

Referenced by libMesh::ReferenceCounter::n_objects(), libMesh::ReferenceCounter::ReferenceCounter(), and libMesh::ReferenceCounter::~ReferenceCounter().

_normalize_solution_snapshots

bool libMesh::RBConstructionBase< LinearImplicitSystem >::_normalize_solution_snapshots

protectedinherited

Set this boolean to true if we want to normalize solution snapshots used in training to have norm of 1.

This is relevant if snapshots have differing magnitudes and we want to approximate them all with equal accuracy.

Definition at line 281 of file rb_construction_base.h.

Referenced by train_reduced_basis_with_POD().

_preevaluate_thetas_completed

bool libMesh::RBConstruction::_preevaluate_thetas_completed

private

Flag to indicate if the preevaluate_thetas function has been called, since this allows us to avoid calling preevaluate_thetas more than once, which is typically unnecessary.

Definition at line 1006 of file rb_construction.h.

Referenced by preevaluate_thetas(), and reset_preevaluate_thetas_completed().

_preevaluate_thetas_flag

bool libMesh::RBConstruction::_preevaluate_thetas_flag

private

Flag to indicate if we preevaluate the theta functions.

Definition at line 999 of file rb_construction.h.

Referenced by get_preevaluate_thetas_flag(), and set_preevaluate_thetas_flag().

_shell_matrix

ShellMatrix<Number>* libMesh::LinearImplicitSystem::_shell_matrix

protectedinherited

User supplies shell matrix or nullptr if no shell matrix is used.

Definition at line 202 of file linear_implicit_system.h.

Referenced by libMesh::LinearImplicitSystem::attach_shell_matrix(), libMesh::LinearImplicitSystem::get_shell_matrix(), and libMesh::LinearImplicitSystem::solve().

_subset

const SystemSubset* libMesh::LinearImplicitSystem::_subset

protectedinherited

The current subset on which to solve (or nullptr if none).

Definition at line 207 of file linear_implicit_system.h.

Referenced by libMesh::LinearImplicitSystem::restrict_solve_to(), and libMesh::LinearImplicitSystem::solve().

_subset_solve_mode

SubsetSolveMode libMesh::LinearImplicitSystem::_subset_solve_mode

protectedinherited

If restrict-solve-to-subset mode is active, this member decides what happens with the dofs outside the subset.

Definition at line 213 of file linear_implicit_system.h.

Referenced by libMesh::LinearImplicitSystem::restrict_solve_to(), and libMesh::LinearImplicitSystem::solve().

_untransformed_basis_functions

std::vector<std::unique_ptr<NumericVector<Number> > > libMesh::RBConstruction::_untransformed_basis_functions

private

In cases where we have dof transformations such as a change of coordinates at some nodes we need to store an extra set of basis functions which have not had dof transformations applied to them.

These vectors are required in order to compute the residual in the error indicator.

Definition at line 988 of file rb_construction.h.

Referenced by compute_residual_dual_norm_slow(), enrich_RB_space(), and update_residual_terms().

_untransformed_solution

std::unique_ptr<NumericVector<Number> > libMesh::RBConstruction::_untransformed_solution

private

We also store a copy of the untransformed solution in order to create _untransformed_basis_functions.

Definition at line 994 of file rb_construction.h.

Referenced by enrich_RB_space(), and truth_solve().

abs_training_tolerance

Real libMesh::RBConstruction::abs_training_tolerance

private

Definition at line 965 of file rb_construction.h.

Referenced by get_abs_training_tolerance(), get_RB_error_bound(), greedy_termination_test(), process_parameters_file(), and set_abs_training_tolerance().

Aq_vector

std::vector<std::unique_ptr<SparseMatrix<Number> > > libMesh::RBConstruction::Aq_vector

private

Vector storing the Q_a matrices from the affine expansion.

Definition at line 936 of file rb_construction.h.

Referenced by allocate_data_structures(), clear(), and get_Aq().

assemble_before_solve

bool libMesh::System::assemble_before_solve

inherited

Flag which tells the system to whether or not to call the user assembly function during each call to solve().

By default, every call to solve() begins with a call to the user assemble, so this flag is true. (For explicit systems, "solving" the system occurs during the assembly step, so this flag is always true for explicit systems.)

You will only want to set this to false if you need direct control over when the system is assembled, and are willing to track the state of its assembly yourself. An example of such a case is an implicit system with multiple right hand sides. In this instance, a single assembly would likely be followed with multiple calls to solve.

The frequency system and Newmark system have their own versions of this flag, called _finished_assemble, which might be able to be replaced with this more general concept.

Definition at line 1609 of file system.h.

Referenced by libMesh::ImplicitSystem::adjoint_solve(), libMesh::ClawSystem::ClawSystem(), libMesh::ImplicitSystem::disable_cache(), libMesh::System::disable_cache(), main(), RBConstruction(), libMesh::RBSCMConstruction::RBSCMConstruction(), libMesh::ImplicitSystem::sensitivity_solve(), libMesh::EigenSystem::solve(), libMesh::CondensedEigenSystem::solve(), and libMesh::LinearImplicitSystem::solve().

assert_convergence

bool libMesh::RBConstruction::assert_convergence

protected

A boolean flag to indicate whether to check for proper convergence after each solve.

Definition at line 891 of file rb_construction.h.

Referenced by compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), enrich_basis_from_rhs_terms(), get_convergence_assertion_flag(), set_convergence_assertion_flag(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

compute_RB_inner_product

bool libMesh::RBConstruction::compute_RB_inner_product

Boolean flag to indicate whether we compute the RB_inner_product_matrix.

This is false by default in RBConstruction since (in the default implementation) the RB inner-product matrix will just be the identity. But we may need the inner-product matrix subclasses.

Definition at line 617 of file rb_construction.h.

Referenced by libMesh::TransientRBConstruction::TransientRBConstruction(), and update_RB_system_matrices().

current_local_solution

std::unique_ptr<NumericVector<Number> > libMesh::System::current_local_solution

inherited

All the values I need to compute my contribution to the simulation at hand.

Think of this as the current solution with any ghost values needed from other processors. This vector is necessarily larger than the solution vector in the case of a parallel simulation. The update() member is used to synchronize the contents of the solution and current_local_solution vectors.

Definition at line 1667 of file system.h.

Referenced by libMesh::__libmesh_petsc_diff_solver_jacobian(), libMesh::__libmesh_petsc_diff_solver_residual(), libMesh::UniformRefinementEstimator::_estimate_error(), alternative_fe_assembly(), libMesh::VariationalSmootherSystem::assembly(), libMesh::NonlinearImplicitSystem::assembly(), libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::EquationSystems::build_parallel_solution_vector(), libMesh::System::clear(), libMesh::Problem_Interface::computeF(), libMesh::Problem_Interface::computeJacobian(), libMesh::Problem_Interface::computePreconditioner(), libMesh::System::current_solution(), DMlibMeshFunction(), DMlibMeshJacobian(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::ExactErrorEstimator::estimate_error(), fe_assembly(), libMesh::StaticCondensation::init(), libMesh::System::init_data(), libMesh::libmesh_petsc_snes_fd_residual(), libMesh::libmesh_petsc_snes_jacobian(), libMesh::libmesh_petsc_snes_mffd_residual(), libMesh::libmesh_petsc_snes_residual(), libMesh::libmesh_petsc_snes_residual_helper(), libMesh::System::point_gradient(), libMesh::System::point_hessian(), libMesh::System::point_value(), libMesh::FEMContext::pre_fe_reinit(), libMesh::System::re_update(), libMesh::System::reinit(), libMesh::System::restrict_vectors(), OverlappingAlgebraicGhostingTest::run_ghosting_test(), OverlappingCouplingGhostingTest::run_sparsity_pattern_test(), SolidSystem::save_initial_mesh(), set_context_solution_vec(), setup(), MeshFunctionTest::test_bad_gradient_var_with_out_of_mesh_value(), MeshFunctionTest::test_bad_hessian_var_with_out_of_mesh_value(), MeshFunctionTest::test_subdomain_id_sets(), MeshInputTest::testCopyElementVectorImpl(), libMesh::BoundaryVolumeSolutionTransfer::transfer_boundary_volume(), libMesh::TransientRBConstruction::truth_assembly(), libMesh::TransientRBConstruction::truth_solve(), libMesh::System::update(), libMesh::Nemesis_IO_Helper::write_element_values(), and libMesh::Nemesis_IO_Helper::write_nodal_solution().

delta_N

unsigned int libMesh::RBConstruction::delta_N

protected

The number of basis functions that we add at each greedy step.

This defaults to 1 in the steady case.

Definition at line 878 of file rb_construction.h.

Referenced by get_delta_N(), libMesh::TransientRBConstruction::process_parameters_file(), recompute_all_residual_terms(), libMesh::TransientRBConstruction::set_delta_N(), train_reduced_basis_with_POD(), libMesh::TransientRBConstruction::update_RB_initial_condition_all_N(), libMesh::TransientRBConstruction::update_RB_system_matrices(), update_RB_system_matrices(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

energy_inner_product_coeffs

std::vector<Number> libMesh::RBConstruction::energy_inner_product_coeffs

private

We may optionally want to use the "energy inner-product" rather than the inner-product assembly specified in inner_product_assembly.

In this case the inner-product will be defined by sum_q^Q k_q * A_q. Here we provide the k_q values that will be used. (Note that a true "energy-inner product" would obtain the k_q from the theta_q's, but this is different for each parameter choice so we just provide a fixed set of k_q's here to ensure that the inner-product is parameter independent)

Definition at line 931 of file rb_construction.h.

Referenced by assemble_inner_product_matrix(), and set_energy_inner_product().

exit_on_repeated_greedy_parameters

bool libMesh::RBConstruction::exit_on_repeated_greedy_parameters

Boolean flag to indicate whether we exit the greedy if we select the same parameters twice in a row.

In some problems this indicates that the greedy has "saturated" typically due to numerical rounding effects.

Definition at line 594 of file rb_construction.h.

Referenced by greedy_termination_test(), and libMesh::TransientRBConstruction::TransientRBConstruction().

extra_linear_solver

LinearSolver<Number>* libMesh::RBConstruction::extra_linear_solver

Also, we store a pointer to an extra linear solver.

This can be useful if we want to pass in the linear solver from somewhere else. For example, if a solver is already primed elsewhere then it can be more efficient to use that solver.

Definition at line 543 of file rb_construction.h.

Referenced by enrich_basis_from_rhs_terms(), and truth_solve().

extra_quadrature_order

int libMesh::System::extra_quadrature_order

inherited

A member int that can be employed to indicate increased or reduced quadrature order.

Note

For FEMSystem users, by default, when calling the user-defined residual functions, the FEMSystem will first set up an appropriate FEType::default_quadrature_rule() object for performing the integration. This rule will integrate elements of order up to 2*p+1 exactly (where p is the sum of the base FEType and local p refinement levels), but if additional (or reduced) quadrature accuracy is desired then this extra_quadrature_order (default 0) will be added.

Definition at line 1640 of file system.h.

Referenced by CurlCurlSystem::init_data(), and set_system_parameters().

Fq_representor

std::vector<std::unique_ptr<NumericVector<Number> > > libMesh::RBConstruction::Fq_representor

Vector storing the residual representors associated with the right-hand side.

These are basis independent and hence stored here, whereas the Aq_representors are stored in RBEvaluation

Definition at line 568 of file rb_construction.h.

Referenced by allocate_data_structures(), clear(), compute_Fq_representor_innerprods(), read_riesz_representors_from_files(), libMesh::TransientRBConstruction::update_residual_terms(), update_residual_terms(), and write_riesz_representors_to_files().

Fq_representor_innerprods

std::vector<Number> libMesh::RBConstruction::Fq_representor_innerprods

Vectors storing the residual representor inner products to be used in computing the residuals online.

We store the Fq representor norms here because they are independent of a reduced basis space. The basis dependent representors are stored in RBEvaluation.

Definition at line 577 of file rb_construction.h.

Referenced by allocate_data_structures(), and compute_Fq_representor_innerprods().

Fq_representor_innerprods_computed

bool libMesh::RBConstruction::Fq_representor_innerprods_computed

A boolean flag to indicate whether or not the Fq representor norms have already been computed — used to make sure that we don't recompute them unnecessarily.

Definition at line 657 of file rb_construction.h.

Referenced by clear(), compute_Fq_representor_innerprods(), read_riesz_representors_from_files(), and recompute_all_residual_terms().

Fq_vector

std::vector<std::unique_ptr<NumericVector<Number> > > libMesh::RBConstruction::Fq_vector

private

Vector storing the Q_f vectors in the affine decomposition of the right-hand side.

Definition at line 942 of file rb_construction.h.

Referenced by allocate_data_structures(), clear(), and get_Fq().

impose_internal_fluxes

bool libMesh::RBConstruction::impose_internal_fluxes

Boolean flag to indicate whether we impose "fluxes" (i.e.

element boundary contributions to the weak form) on internal element boundaries in the assembly routines.

Definition at line 601 of file rb_construction.h.

Referenced by add_scaled_matrix_and_vector().

inner_product_assembly

ElemAssembly* libMesh::RBConstruction::inner_product_assembly

private

Pointer to inner product assembly.

Definition at line 913 of file rb_construction.h.

Referenced by assemble_inner_product_matrix(), get_inner_product_assembly(), and set_inner_product_assembly().

inner_product_matrix

std::unique_ptr<SparseMatrix<Number> > libMesh::RBConstruction::inner_product_matrix

The inner product matrix.

Definition at line 548 of file rb_construction.h.

Referenced by allocate_data_structures(), assemble_misc_matrices(), compute_Fq_representor_innerprods(), compute_residual_dual_norm_slow(), get_inner_product_matrix(), get_matrix_for_output_dual_solves(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

inner_product_solver

std::unique_ptr<LinearSolver<Number> > libMesh::RBConstruction::inner_product_solver

We store an extra linear solver object which we can optionally use for solving all systems in which the system matrix is set to inner_product_matrix.

Definition at line 535 of file rb_construction.h.

Referenced by compute_Fq_representor_innerprods(), compute_residual_dual_norm_slow(), initialize_rb_construction(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

inner_product_storage_vector

std::unique_ptr<NumericVector<Number> > libMesh::RBConstructionBase< LinearImplicitSystem >::inner_product_storage_vector

protectedinherited

We keep an extra temporary vector that is useful for performing inner products (avoids unnecessary memory allocation/deallocation).

Definition at line 288 of file rb_construction_base.h.

Referenced by libMesh::TransientRBConstruction::add_IC_to_RB_space(), compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), compute_residual_dual_norm_slow(), libMesh::TransientRBConstruction::enrich_RB_space(), enrich_RB_space(), libMesh::TransientRBConstruction::set_error_temporal_data(), train_reduced_basis_with_POD(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::TransientRBConstruction::update_residual_terms(), and update_residual_terms().

linear_solver

std::unique_ptr<LinearSolver<Number> > libMesh::ImplicitSystem::linear_solver

mutableinherited

This class handles all the details of interfacing with various linear algebra packages like PETSc or LASPACK.

This is a public member for backwards compatibility reasons, but in general it's better to use get_linear_solver() to access this member, since that function will also handle initialization if it hasn't already been taken care of.

Definition at line 328 of file implicit_system.h.

Referenced by libMesh::LinearImplicitSystem::clear(), compute_output_dual_innerprods(), libMesh::ContinuationSystem::continuation_solve(), libMesh::ContinuationSystem::ContinuationSystem(), libMesh::LinearImplicitSystem::create_static_condensation(), libMesh::ImplicitSystem::get_linear_solver(), libMesh::LinearImplicitSystem::get_linear_solver(), libMesh::LinearImplicitSystem::init_data(), libMesh::LinearImplicitSystem::LinearImplicitSystem(), libMesh::LinearImplicitSystem::reinit(), libMesh::FrequencySystem::solve(), libMesh::LinearImplicitSystem::solve(), libMesh::ClawSystem::solve_conservation_law(), libMesh::ContinuationSystem::solve_tangent(), and libMesh::TransientRBConstruction::truth_solve().

matrix

SparseMatrix<Number>* libMesh::ImplicitSystem::matrix

inherited

The system matrix.

Implicit systems are characterized by the need to solve the linear system Ax=b. This is the system matrix A.

Public access to this member variable will be deprecated in the future! Use get_system_matrix() instead.

Definition at line 311 of file implicit_system.h.

Referenced by libMesh::__libmesh_petsc_diff_solver_jacobian(), add_M_C_K_helmholtz(), libMesh::ImplicitSystem::add_matrices(), libMesh::ImplicitSystem::adjoint_solve(), libMesh::ImplicitSystem::assemble(), assemble_func(), assemble_temperature_jump(), libMesh::FEMSystem::assembly(), libMesh::LinearImplicitSystem::assembly(), libMesh::NonlinearImplicitSystem::assembly(), libMesh::ImplicitSystem::clear(), libMesh::NewmarkSystem::compute_matrix(), compute_residual_dual_norm_slow(), libMesh::ContinuationSystem::continuation_solve(), libMesh::ImplicitSystem::create_static_condensation_system_matrix(), DMCreateMatrix_libMesh(), DMlibMeshJacobian(), enrich_basis_from_rhs_terms(), fill_dirichlet_bc(), libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity(), libMesh::ImplicitSystem::get_system_matrix(), main(), periodic_bc_test_poisson(), libMesh::ImplicitSystem::qoi_parameter_hessian(), libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product(), libMesh::ImplicitSystem::sensitivity_solve(), libMesh::NewtonSolver::solve(), libMesh::PetscDiffSolver::solve(), libMesh::NoxNonlinearSolver< Number >::solve(), libMesh::EigenTimeSolver::solve(), libMesh::FrequencySystem::solve(), libMesh::LinearImplicitSystem::solve(), libMesh::NonlinearImplicitSystem::solve(), libMesh::ClawSystem::solve_conservation_law(), libMesh::ContinuationSystem::solve_tangent(), libMesh::TransientRBConstruction::truth_assembly(), truth_assembly(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve(), and libMesh::ImplicitSystem::weighted_sensitivity_solve().

Nmax

unsigned int libMesh::RBConstruction::Nmax

protected

Maximum number of reduced basis functions we are willing to use.

Definition at line 872 of file rb_construction.h.

Referenced by libMesh::TransientRBConstruction::allocate_data_structures(), get_Nmax(), print_info(), process_parameters_file(), and set_Nmax().

non_dirichlet_Aq_vector

std::vector<std::unique_ptr<SparseMatrix<Number> > > libMesh::RBConstruction::non_dirichlet_Aq_vector

private

We may also need a second set of matrices/vectors that do not have the Dirichlet boundary conditions enforced.

Definition at line 955 of file rb_construction.h.

Referenced by allocate_data_structures(), clear(), and get_non_dirichlet_Aq().

non_dirichlet_Fq_vector

std::vector<std::unique_ptr<NumericVector<Number> > > libMesh::RBConstruction::non_dirichlet_Fq_vector

private

Definition at line 956 of file rb_construction.h.

Referenced by allocate_data_structures(), clear(), and get_non_dirichlet_Fq().

non_dirichlet_inner_product_matrix

std::unique_ptr<SparseMatrix<Number> > libMesh::RBConstruction::non_dirichlet_inner_product_matrix

private

Definition at line 958 of file rb_construction.h.

Referenced by allocate_data_structures(), assemble_misc_matrices(), and get_non_dirichlet_inner_product_matrix().

non_dirichlet_outputs_vector

std::vector<std::vector<std::unique_ptr<NumericVector<Number> > > > libMesh::RBConstruction::non_dirichlet_outputs_vector

private

Definition at line 957 of file rb_construction.h.

Referenced by allocate_data_structures(), clear(), and get_non_dirichlet_output_vector().

normalize_rb_bound_in_greedy

bool libMesh::RBConstruction::normalize_rb_bound_in_greedy

private

This boolean indicates if we normalize the RB error in the greedy using RBEvaluation::get_error_bound_normalization().

Definition at line 971 of file rb_construction.h.

Referenced by get_normalize_rb_bound_in_greedy(), get_RB_error_bound(), and set_normalize_rb_bound_in_greedy().

output_dual_innerprods

std::vector<std::vector<Number > > libMesh::RBConstruction::output_dual_innerprods

The vector storing the dual norm inner product terms for each output.

Definition at line 560 of file rb_construction.h.

Referenced by allocate_data_structures(), and compute_output_dual_innerprods().

output_dual_innerprods_computed

bool libMesh::RBConstruction::output_dual_innerprods_computed

protected

A boolean flag to indicate whether or not the output dual norms have already been computed — used to make sure that we don't recompute them unnecessarily.

Definition at line 885 of file rb_construction.h.

Referenced by compute_output_dual_innerprods().

outputs_vector

std::vector<std::vector<std::unique_ptr<NumericVector<Number> > > > libMesh::RBConstruction::outputs_vector

private

The libMesh vectors that define the output functionals.

Each row corresponds to the affine expansion of an output.

Definition at line 948 of file rb_construction.h.

Referenced by allocate_data_structures(), clear(), and get_output_vector().

parameters

Parameters libMesh::System::parameters

inherited

Parameters for the system. If a parameter is not provided, it should be retrieved from the EquationSystems.

Definition at line 1588 of file system.h.

Referenced by libMesh::ImplicitSystem::get_linear_solve_parameters(), libMesh::FrequencySystem::init_data(), libMesh::NonlinearImplicitSystem::set_solver_parameters(), libMesh::EigenSystem::solve(), libMesh::CondensedEigenSystem::solve(), and libMesh::EigenSystem::solve_helper().

quiet_mode

bool libMesh::RBConstructionBase< LinearImplicitSystem >::quiet_mode

protectedinherited

Flag to indicate whether we print out extra information during the Offline stage.

Definition at line 265 of file rb_construction_base.h.

Referenced by process_parameters_file().

rb_assembly_expansion

RBAssemblyExpansion* libMesh::RBConstruction::rb_assembly_expansion

private

This member holds the (parameter independent) assembly functors that define the "affine expansion" of the PDE that we are solving.

Definition at line 908 of file rb_construction.h.

Referenced by add_scaled_Aq(), assemble_all_output_vectors(), assemble_Aq_matrix(), assemble_Fq_vector(), assemble_inner_product_matrix(), get_rb_assembly_expansion(), and set_rb_assembly_expansion().

rb_eval

RBEvaluation* libMesh::RBConstruction::rb_eval

private

The current RBEvaluation object we are using to perform the Evaluation stage of the reduced basis method.

Definition at line 902 of file rb_construction.h.

Referenced by get_rb_evaluation(), is_rb_eval_initialized(), and set_rb_evaluation().

RB_training_type

std::string libMesh::RBConstruction::RB_training_type

private

This string indicates the type of training that we will use.

Options are:

• Greedy: Reduced basis greedy algorithm
• POD: Proper Orthogonal Decomposition

Definition at line 979 of file rb_construction.h.

Referenced by get_RB_training_type(), and set_RB_training_type().

rel_training_tolerance

Real libMesh::RBConstruction::rel_training_tolerance

private

Relative and absolute tolerances for training reduced basis using the Greedy scheme.

Definition at line 964 of file rb_construction.h.

Referenced by get_rel_training_tolerance(), greedy_termination_test(), process_parameters_file(), set_rel_training_tolerance(), and train_reduced_basis_with_POD().

rhs

NumericVector<Number>* libMesh::ExplicitSystem::rhs

inherited

The system matrix.

Implicit systems are characterized by the need to solve the linear system Ax=b. This is the right-hand-side vector b.

Definition at line 124 of file explicit_system.h.

Referenced by libMesh::__libmesh_petsc_diff_solver_residual(), add_M_C_K_helmholtz(), libMesh::ExplicitSystem::add_system_rhs(), assemble(), libMesh::ImplicitSystem::assemble(), LinearElasticity::assemble(), assemble_1D(), assemble_biharmonic(), libMesh::AdvectionSystem::assemble_claw_rhs(), assemble_divgrad(), assemble_elasticity(), assemble_ellipticdg(), assemble_func(), assemble_graddiv(), assemble_laplace(), assemble_matrix_and_rhs(), assemble_poisson(), libMesh::ImplicitSystem::assemble_residual_derivatives(), assemble_shell(), assemble_stokes(), assemble_temperature_jump(), assemble_wave(), libMesh::FEMSystem::assembly(), libMesh::VariationalSmootherSystem::assembly(), libMesh::LinearImplicitSystem::assembly(), libMesh::NonlinearImplicitSystem::assembly(), assembly_with_dg_fem_context(), compute_Fq_representor_innerprods(), compute_output_dual_innerprods(), compute_residual_dual_norm_slow(), libMesh::Problem_Interface::computeF(), libMesh::ContinuationSystem::continuation_solve(), DMlibMeshFunction(), enrich_basis_from_rhs_terms(), fill_dirichlet_bc(), libMesh::ImplicitSystem::forward_qoi_parameter_sensitivity(), libMesh::NewtonSolver::line_search(), periodic_bc_test_poisson(), HeatSystem::perturb_accumulate_residuals(), libMesh::ImplicitSystem::qoi_parameter_hessian(), libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product(), libMesh::NewtonSolver::solve(), libMesh::PetscDiffSolver::solve(), libMesh::FrequencySystem::solve(), libMesh::LinearImplicitSystem::solve(), libMesh::NonlinearImplicitSystem::solve(), libMesh::ContinuationSystem::solve_tangent(), libMesh::TransientRBConstruction::truth_assembly(), truth_assembly(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::TransientRBConstruction::update_residual_terms(), update_residual_terms(), libMesh::NewmarkSystem::update_rhs(), libMesh::ImplicitSystem::weighted_sensitivity_adjoint_solve(), and libMesh::ImplicitSystem::weighted_sensitivity_solve().

serial_training_set

bool libMesh::RBConstructionBase< LinearImplicitSystem >::serial_training_set

protectedinherited

This boolean flag indicates whether or not the training set should be the same on all processors.

By default it is false, but in the case of the Empirical Interpolation Method (RBEIMConstruction), for example, we need the training set to be identical on all processors.

Definition at line 273 of file rb_construction_base.h.

Referenced by compute_max_error_bound(), set_RB_training_type(), and train_reduced_basis_with_POD().

skip_degenerate_sides

bool libMesh::RBConstruction::skip_degenerate_sides

In some cases meshes are intentionally created with degenerate sides as a way to represent, say, triangles using a hex-only mesh.

In this situation we should detect and skip any degenerate sides in order to prevent zero or negative element Jacobian errors.

Definition at line 609 of file rb_construction.h.

Referenced by add_scaled_matrix_and_vector().

skip_residual_in_train_reduced_basis

bool libMesh::RBConstruction::skip_residual_in_train_reduced_basis

Boolean flag to indicate if we skip residual calculations in train_reduced_basis.

This should only be used in special cases, e.g. when we know a priori that we want exactly one basis function and hence we do not need the residual based error indicator.

Definition at line 586 of file rb_construction.h.

Referenced by train_reduced_basis_with_greedy().

solution

std::unique_ptr<NumericVector<Number> > libMesh::System::solution

inherited

Data structure to hold solution values.

Definition at line 1655 of file system.h.

Referenced by libMesh::__libmesh_petsc_diff_solver_jacobian(), libMesh::__libmesh_petsc_diff_solver_residual(), libMesh::ExactSolution::_compute_error(), libMesh::UniformRefinementEstimator::_estimate_error(), libMesh::TransientRBConstruction::add_IC_to_RB_space(), libMesh::NewmarkSolver::advance_timestep(), libMesh::AdaptiveTimeSolver::advance_timestep(), libMesh::UnsteadySolver::advance_timestep(), libMesh::ContinuationSystem::apply_predictor(), libMesh::TransientRBConstruction::assemble_affine_expansion(), libMesh::FEMSystem::assembly(), libMesh::VariationalSmootherSystem::assembly(), libMesh::LinearImplicitSystem::assembly(), libMesh::EquationSystems::build_parallel_elemental_solution_vector(), libMesh::EquationSystems::build_parallel_solution_vector(), check_if_zero_truth_solve(), libMesh::System::clear(), libMesh::System::compare(), compute_enriched_soln(), compute_Fq_representor_innerprods(), libMesh::NewmarkSolver::compute_initial_accel(), compute_output_dual_innerprods(), compute_residual_dual_norm_slow(), compute_stresses(), LinearElasticityWithContact::compute_stresses(), LinearElasticity::compute_stresses(), LargeDeformationElasticity::compute_stresses(), libMesh::Problem_Interface::computeF(), libMesh::Problem_Interface::computeJacobian(), libMesh::Problem_Interface::computePreconditioner(), libMesh::ContinuationSystem::continuation_solve(), libMesh::Nemesis_IO::copy_elemental_solution(), libMesh::ExodusII_IO::copy_elemental_solution(), libMesh::GMVIO::copy_nodal_solution(), libMesh::Nemesis_IO::copy_nodal_solution(), libMesh::ExodusII_IO::copy_nodal_solution(), libMesh::Nemesis_IO::copy_scalar_solution(), libMesh::ExodusII_IO::copy_scalar_solution(), create_wrapped_function(), DMCreateGlobalVector_libMesh(), DMlibMeshFunction(), DMlibMeshJacobian(), libMesh::UnsteadySolver::du(), enrich_RB_space(), libMesh::PatchRecoveryErrorEstimator::estimate_error(), libMesh::WeightedPatchRecoveryErrorEstimator::estimate_error(), libMesh::JumpErrorEstimator::estimate_error(), libMesh::AdjointRefinementEstimator::estimate_error(), libMesh::AdjointResidualErrorEstimator::estimate_error(), libMesh::ExactErrorEstimator::estimate_error(), libMesh::SmoothnessEstimator::estimate_smoothness(), libMesh::RBSCMConstruction::evaluate_stability_constant(), libMesh::EigenSystem::get_eigenpair(), libMesh::CondensedEigenSystem::get_eigenpair(), LinearElasticityWithContact::get_least_and_max_gap_function(), libMesh::VariationalSmootherSystem::init_data(), libMesh::System::init_data(), libMesh::ContinuationSystem::initialize_tangent(), libMesh::TransientRBConstruction::initialize_truth(), libMesh::libmesh_petsc_snes_fd_residual(), libMesh::libmesh_petsc_snes_jacobian(), libMesh::libmesh_petsc_snes_mffd_residual(), libMesh::libmesh_petsc_snes_residual(), libMesh::libmesh_petsc_snes_residual_helper(), load_basis_function(), libMesh::TransientRBConstruction::load_rb_solution(), load_rb_solution(), main(), libMesh::DofMap::max_constraint_error(), libMesh::FEMSystem::mesh_position_get(), libMesh::ErrorVector::plot_error(), print_basis_function_orthogonality(), libMesh::InterMeshProjection::project_system_vectors(), libMesh::ImplicitSystem::qoi_parameter_hessian(), libMesh::ImplicitSystem::qoi_parameter_hessian_vector_product(), libMesh::System::re_update(), libMesh::System::read_parallel_data(), libMesh::TransientRBConstruction::read_riesz_representors_from_files(), read_riesz_representors_from_files(), libMesh::System::read_serialized_data(), MeshFunctionTest::read_variable_info_from_output_data(), libMesh::System::reinit(), libMesh::System::restrict_vectors(), libMesh::MemoryHistoryData::retrieve_vectors(), OverlappingAlgebraicGhostingTest::run_ghosting_test(), OverlappingCouplingGhostingTest::run_sparsity_pattern_test(), libMesh::ContinuationSystem::save_current_solution(), libMesh::TransientRBConstruction::set_error_temporal_data(), setup(), WriteVecAndScalar::setupTests(), libMesh::TwostepTimeSolver::solve(), libMesh::NewtonSolver::solve(), libMesh::PetscDiffSolver::solve(), libMesh::FrequencySystem::solve(), libMesh::LinearImplicitSystem::solve(), libMesh::NonlinearImplicitSystem::solve(), libMesh::ClawSystem::solve_conservation_law(), solve_for_matrix_and_rhs(), libMesh::ContinuationSystem::solve_tangent(), libMesh::MemoryHistoryData::store_vectors(), ConstraintOperatorTest::test1DCoarseningOperator(), MeshfunctionDFEM::test_mesh_function_dfem(), MeshfunctionDFEM::test_mesh_function_dfem_grad(), MeshFunctionTest::test_p_level(), ProjectSolutionTest::test_partial_project_solution(), SystemsTest::testBoundaryProjectCube(), ConstraintOperatorTest::testCoreform(), SystemsTest::testDofCouplingWithVarGroups(), MeshInputTest::testExodusWriteElementDataFromDiscontinuousNodalData(), SystemsTest::testPostInitAddVector(), SystemsTest::testProjectCubeWithMeshFunction(), MeshInputTest::testProjectionRegression(), WriteVecAndScalar::testSolution(), train_reduced_basis_with_POD(), libMesh::MeshFunctionSolutionTransfer::transfer(), libMesh::DirectSolutionTransfer::transfer(), libMesh::MeshfreeSolutionTransfer::transfer(), libMesh::BoundaryVolumeSolutionTransfer::transfer_boundary_volume(), libMesh::BoundaryVolumeSolutionTransfer::transfer_volume_boundary(), libMesh::TransientRBConstruction::truth_solve(), truth_solve(), libMesh::System::update(), update_current_local_solution(), libMesh::System::update_global_solution(), libMesh::TransientRBConstruction::update_RB_initial_condition_all_N(), libMesh::TransientRBConstruction::update_residual_terms(), update_residual_terms(), libMesh::ContinuationSystem::update_solution(), libMesh::NewmarkSystem::update_u_v_a(), libMesh::DTKAdapter::update_variable_values(), libMesh::System::write_parallel_data(), libMesh::TransientRBConstruction::write_riesz_representors_to_files(), write_riesz_representors_to_files(), and libMesh::System::write_serialized_data().

store_dirichlet_operators

bool libMesh::RBConstruction::store_dirichlet_operators

Boolean flag to indicate whether we store affine operator matrices and vectors with constraints enforced.

This is true by default, but we can choose to skip it in some cases when all we need are operators without constraints enforced (e.g. when we only need to compute inner-product-type terms).

Definition at line 626 of file rb_construction.h.

Referenced by allocate_data_structures(), assemble_all_affine_operators(), assemble_all_affine_vectors(), assemble_all_output_vectors(), assemble_misc_matrices(), clear(), get_all_matrices(), get_all_vectors(), get_Aq(), get_Fq(), get_inner_product_matrix(), and get_output_vectors().

store_non_dirichlet_operators

bool libMesh::RBConstruction::store_non_dirichlet_operators

Boolean flag to indicate whether we store a second copy of each affine operator and vector which does not have Dirichlet bcs enforced.

Definition at line 633 of file rb_construction.h.

Referenced by libMesh::TransientRBConstruction::allocate_data_structures(), allocate_data_structures(), libMesh::TransientRBConstruction::assemble_all_affine_operators(), assemble_all_affine_operators(), assemble_all_affine_vectors(), assemble_all_output_vectors(), libMesh::TransientRBConstruction::assemble_misc_matrices(), assemble_misc_matrices(), libMesh::TransientRBConstruction::clear(), clear(), libMesh::TransientRBConstruction::get_all_matrices(), get_all_matrices(), get_all_vectors(), get_non_dirichlet_Aq(), get_non_dirichlet_Aq_if_avail(), get_non_dirichlet_Fq(), get_non_dirichlet_Fq_if_avail(), get_non_dirichlet_inner_product_matrix(), get_non_dirichlet_inner_product_matrix_if_avail(), libMesh::TransientRBConstruction::get_non_dirichlet_M_q(), and get_output_vectors().

store_untransformed_basis

bool libMesh::RBConstruction::store_untransformed_basis

Boolean flag to indicate whether we store a second copy of the basis without constraints or dof transformations applied to it.

This is necessary when we have dof transformations and need to calculate the residual R(U) = C^T F - C^T A C U, since we need to evaluate R(U) using the untransformed basis U rather than C U to avoid "double applying" dof transformations in C.

Definition at line 643 of file rb_construction.h.

Referenced by compute_residual_dual_norm_slow(), enrich_RB_space(), truth_solve(), and update_residual_terms().

time

Real libMesh::System::time

inherited

For time-dependent problems, this is the time t at the beginning of the current timestep.

Note

For DifferentiableSystem users: do not access this time during an assembly! Use the DiffContext::time value instead to get correct results.

Definition at line 1677 of file system.h.

Referenced by libMesh::AdaptiveTimeSolver::adjoint_advance_timestep(), libMesh::UnsteadySolver::adjoint_advance_timestep(), libMesh::TwostepTimeSolver::adjoint_solve(), libMesh::AdaptiveTimeSolver::advance_timestep(), libMesh::UnsteadySolver::advance_timestep(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::SubProjector::construct_projection(), HeatSystem::element_qoi(), fill_dirichlet_bc(), libMesh::ExactErrorEstimator::find_squared_element_error(), initialize(), libMesh::Euler2Solver::integrate_adjoint_refinement_error_estimate(), libMesh::EulerSolver::integrate_adjoint_refinement_error_estimate(), libMesh::UnsteadySolver::integrate_adjoint_sensitivity(), libMesh::Euler2Solver::integrate_qoi_timestep(), libMesh::EulerSolver::integrate_qoi_timestep(), main(), libMesh::WeightedPatchRecoveryErrorEstimator::EstimateError::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectVertices::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectEdges::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectSides::operator()(), libMesh::GenericProjector< FFunctor, GFunctor, FValue, ProjectionAction >::ProjectInteriors::operator()(), libMesh::System::reinit_constraints(), libMesh::UnsteadySolver::retrieve_timestep(), and libMesh::TwostepTimeSolver::solve().

training_error_bounds

std::vector<Real> libMesh::RBConstruction::training_error_bounds

Vector storing the values of the error bound for each parameter in the training set — the parameter giving the largest error bound is chosen for the next snapshot in the Greedy basis training.

Definition at line 528 of file rb_construction.h.

Referenced by compute_max_error_bound().

truth_outputs

std::vector<Number > libMesh::RBConstruction::truth_outputs

Vector storing the truth output values from the most recent truth solve.

Definition at line 554 of file rb_construction.h.

Referenced by allocate_data_structures(), and truth_solve().

use_empty_rb_solve_in_greedy

bool libMesh::RBConstruction::use_empty_rb_solve_in_greedy

A boolean flag to indicate whether or not we initialize the Greedy algorithm by performing rb_solves on the training set with an "empty" (i.e.

N=0) reduced basis space.

Definition at line 650 of file rb_construction.h.

Referenced by train_reduced_basis_with_greedy().

use_energy_inner_product

bool libMesh::RBConstruction::use_energy_inner_product

private

Boolean to indicate whether we're using the energy inner-product.

If this is false then we use inner_product_assembly instead.

Definition at line 919 of file rb_construction.h.

Referenced by assemble_inner_product_matrix(), get_inner_product_assembly(), set_energy_inner_product(), and set_inner_product_assembly().

use_fixed_solution

bool libMesh::System::use_fixed_solution

inherited

A boolean to be set to true by systems using elem_fixed_solution, for optional use by e.g.

stabilized methods. False by default.

Note

For FEMSystem users, if this variable is set to true, it must be before init_data() is called.

Definition at line 1625 of file system.h.

Referenced by libMesh::EulerSolver::_general_residual(), libMesh::Euler2Solver::_general_residual(), libMesh::SteadySolver::_general_residual(), libMesh::NewmarkSolver::_general_residual(), libMesh::DifferentiableSystem::clear(), libMesh::DiffContext::DiffContext(), and libMesh::FEMContext::pre_fe_reinit().

verbose_mode

bool libMesh::RBParametrized::verbose_mode

inherited

Public boolean to toggle verbose mode.

Definition at line 181 of file rb_parametrized.h.

Referenced by libMesh::RBParametrized::check_if_valid_params().

zero_out_matrix_and_rhs

bool libMesh::ImplicitSystem::zero_out_matrix_and_rhs

inherited

By default, the system will zero out the matrix and the right hand side.

If this flag is false, it is the responsibility of the client code to take care of setting these to zero before assembly begins

Definition at line 318 of file implicit_system.h.

Referenced by libMesh::ImplicitSystem::assemble().

---

The documentation for this class was generated from the following files:

• include/reduced_basis/rb_construction.h
• src/reduced_basis/rb_construction.C