serif::solver Namespace Reference

Namespaces
namespace	admissibility
namespace	backend
namespace	continuation
namespace	detail
namespace	linear
namespace	normalization
namespace	preconditioner
namespace	seed

Classes
struct	Backtracking
	Shorten a Newton step until the complete residual merit decreases sufficiently. More...
struct	DirectionAssessment
	A modified direction must report whether it is usable and any measured merit slope. More...
struct	MetricEvaluation
	Separate the convergence norm from the line-search merit for custom metrics. More...
struct	Newton
	Accuracy and work limits for finding one equilibrium star. More...
struct	NewtonCheckpoint
	Own the accepted physical star and the small amount of Newton history needed to continue it. More...
struct	NewtonProgress
	Numerical continuation data at one accepted Newton boundary. More...
struct	NoModel
	Marker for algebraic tests and custom problems without a stellar description. More...
struct	NoObserver
	Disable observation without introducing runtime callback allocation. More...
struct	NormalizedResidualMetric
	Use the Euclidean norm of SERiF's normalized residual and half its square. More...
struct	SolveDiagnostics
	Bounded work totals; accepted physical snapshots are stored separately from diagnostics. More...
struct	SolveEvent
	Small event record suitable for logging without retaining every iterate. More...
class	SolveResult
	Own the last accepted physical state, matching frozen system, and descriptive model. More...
struct	SolverOptions
	Select numerical policies while keeping equilibrium physics in the model. More...
struct	SolveTiming
	Optional elapsed wall time, kept separately from numerical restart history. More...
struct	StandardNewton
	Ordinary Newton supplies every strategy phase; custom policies may replace individual phases explicitly. More...
class	TimedObserver
	Measure solves explicitly while retaining an ordinary progress or stop callback. More...
struct	TrialAssessment
	A numerical policy explicitly accepts its work or explains why this candidate must be rejected. More...

Typedefs
using	VolumeOptions = operators::backend::mfem::VolumeOptions

Enumerations
enum class	SolveStatus {   converged , initial_state_rejected , invalid_metric , preparation_failure ,   linear_solve_failure , globalization_failure , maximum_iterations , stopped }
	Numerical termination; invalid configuration and user callback errors remain exceptions. More...
enum class	EventKind {   initial , before_iteration , trial_rejected , before_accept ,   accepted , finished , resumed }
	Events expose immutable accepted/trial snapshots; callbacks never decide acceptance. More...
enum class	SolveControl { continue_solve , stop }
	A stop request is remembered until the next initial, resumed or accepted Newton boundary. More...

Functions
template<typename Family, typename Factory, typename State, typename Options = SolverOptions<>> requires std::copy_constructible<Options>
auto	continue_equilibrium (const Family &family, Factory configure, const State &initial_state, continuation::Options< Family > continuation_options, Options solver_options=continuation::default_solver_options())
	Continue moving-star equilibria through an explicitly declared physical family.
template<typename Family, typename Factory, typename State>
auto	continue_equilibrium (const Family &family, Factory configure, const State &initial_state)
	Start with the family's physics-owned default coordinates and bounded fallback route.
template<typename Family, typename Factory, typename State, typename Options = SolverOptions<>> requires std::copy_constructible<Options>
auto	continue_equilibrium (const Family &family, Factory configure, const State &initial_state, const Vector &displacement, continuation::Options< Family > continuation_options, Options solver_options=continuation::default_solver_options())
	Continue a fixed-geometry model while preserving its prescribed displacement.
template<typename Family, typename Factory, typename State>
auto	continue_equilibrium (const Family &family, Factory configure, const State &initial_state, const Vector &displacement)
	Use the same default route and absolute qualification tolerance for fixed geometry.
template<typename Configuration, typename Layout, typename Options = SolverOptions<>, typename Observer = NoObserver, utils::dependencies::IsDependencyContext Inputs = utils::dependencies::DependencyContext<>, typename ObserverFactory = utils::dependencies::NoGraphObservers> requires utils::dependencies::ContainsKey<discretization::blocks::surface_deformation::coefficients, typename Layout::Values>
auto	solve_equilibrium (Configuration configuration, model::StellarState< Layout > initial_state, Options options={}, Observer observer={}, VolumeOptions volume_options={}, Inputs prescribed=Inputs{}, ObserverFactory observer_factory={})
	Solve a moving star from an owning physical initial state.
template<typename Configuration, typename Layout, typename Options = SolverOptions<>, typename Observer = NoObserver, utils::dependencies::IsDependencyContext Inputs = utils::dependencies::DependencyContext<>, typename ObserverFactory = utils::dependencies::NoGraphObservers> requires (!utils::dependencies::ContainsKey<discretization::blocks::surface_deformation::coefficients, typename Layout::Values>)
auto	solve_equilibrium (Configuration configuration, model::StellarState< Layout > initial_state, const Vector &displacement, Options options={}, Observer observer={}, VolumeOptions volume_options={}, Inputs prescribed=Inputs{}, ObserverFactory observer_factory={})
	Solve a fixed-geometry model with prescribed displacement outside its unknown vector.
template<typename Configuration, typename Layout, typename Options = SolverOptions<>, typename Observer = NoObserver, utils::dependencies::IsDependencyContext Inputs = utils::dependencies::DependencyContext<>, typename ObserverFactory = utils::dependencies::NoGraphObservers> requires utils::dependencies::ContainsKey<discretization::blocks::surface_deformation::coefficients, typename Layout::Values>
auto	resume_equilibrium (Configuration configuration, NewtonCheckpoint< model::StellarState< Layout > > checkpoint, Options options={}, Observer observer={}, VolumeOptions volume_options={}, Inputs prescribed=Inputs{}, ObserverFactory observer_factory={})
	Continue a moving-star checkpoint using reconstructed physical and numerical configuration.
template<typename Configuration, typename Layout, typename Options = SolverOptions<>, typename Observer = NoObserver, utils::dependencies::IsDependencyContext Inputs = utils::dependencies::DependencyContext<>, typename ObserverFactory = utils::dependencies::NoGraphObservers> requires (!utils::dependencies::ContainsKey<discretization::blocks::surface_deformation::coefficients, typename Layout::Values>)
auto	resume_equilibrium (Configuration configuration, NewtonCheckpoint< model::StellarState< Layout > > checkpoint, const Vector &displacement, Options options={}, Observer observer={}, VolumeOptions volume_options={}, Inputs prescribed=Inputs{}, ObserverFactory observer_factory={})
	Continue a fixed-geometry checkpoint with its separately restored prescribed displacement.
template<typename Problem, typename State, typename Options = SolverOptions<>, typename Observer = NoObserver> requires detail::NewtonStrategyFor<Problem, State, Options>
auto	solve_newton (Problem &problem, State initial_state, Options options={}, Observer observer={})
	Solve a custom numerical problem using its native Vector, evaluate, prepare and solve_linear operations.
template<typename Problem, typename State, typename Options = SolverOptions<>, typename Observer = NoObserver> requires detail::NewtonStrategyFor<Problem, State, Options>
auto	resume_newton (Problem &problem, NewtonCheckpoint< State > checkpoint, Options options={}, Observer observer={})
	Resume the original convergence reference, correction reuse, and cumulative work budget.
double	next_backtracking_step (double rejected_step, double accepted_determinant, double rejected_determinant, bool inverted, const Backtracking &options) noexcept
	Skip only existing backtracking rungs when an inverted trial gives a finite boundary estimate.
std::optional< std::string >	validate_newton_progress (const NewtonProgress &progress, const std::size_t state_size)
	Reject incomplete or inconsistent imported history before preparing any runtime operators.

Typedef Documentation

VolumeOptions

using serif::solver::VolumeOptions = operators::backend::mfem::VolumeOptions

Enumeration Type Documentation

EventKind

enum class serif::solver::EventKind

strong

Events expose immutable accepted/trial snapshots; callbacks never decide acceptance.

Enumerator
initial
before_iteration
trial_rejected
before_accept
accepted
finished
resumed

SolveControl

enum class serif::solver::SolveControl

strong

A stop request is remembered until the next initial, resumed or accepted Newton boundary.

Enumerator
continue_solve
stop

SolveStatus

enum class serif::solver::SolveStatus

strong

Numerical termination; invalid configuration and user callback errors remain exceptions.

Enumerator
converged
initial_state_rejected
invalid_metric
preparation_failure
linear_solve_failure
globalization_failure
maximum_iterations
stopped

Function Documentation

continue_equilibrium() [1/4]

template<typename Family, typename Factory, typename State>

auto serif::solver::continue_equilibrium ( const Family & family, Factory configure, const State & initial_state )

nodiscard

Start with the family's physics-owned default coordinates and bounded fallback route.

continue_equilibrium() [2/4]

template<typename Family, typename Factory, typename State>

auto serif::solver::continue_equilibrium ( const Family & family, Factory configure, const State & initial_state, const Vector & displacement )

nodiscard

Use the same default route and absolute qualification tolerance for fixed geometry.

continue_equilibrium() [3/4]

template<typename Family, typename Factory, typename State, typename Options = SolverOptions<>>
requires std::copy_constructible<Options>

auto serif::solver::continue_equilibrium ( const Family & family, Factory configure, const State & initial_state, const Vector & displacement, continuation::Options< Family > continuation_options, Options solver_options = continuation::default_solver_options() )

nodiscard

Continue a fixed-geometry model while preserving its prescribed displacement.

continue_equilibrium() [4/4]

template<typename Family, typename Factory, typename State, typename Options = SolverOptions<>>
requires std::copy_constructible<Options>

auto serif::solver::continue_equilibrium ( const Family & family, Factory configure, const State & initial_state, continuation::Options< Family > continuation_options, Options solver_options = continuation::default_solver_options() )

nodiscard

Continue moving-star equilibria through an explicitly declared physical family.

Parameters

configure

Own or borrow a callable that returns a configured discretization for its supplied model. It must preserve the physical coordinate layout and finite-element reference of transferred states; fresh evaluators are implicit.

Note

The explicit factory supports move-only resource policies without inventing a mutable clone/context hierarchy.

next_backtracking_step()

double serif::solver::next_backtracking_step ( double rejected_step, double accepted_determinant, double rejected_determinant, bool inverted, const Backtracking & options )

nodiscardnoexcept

Skip only existing backtracking rungs when an inverted trial gives a finite boundary estimate.

resume_equilibrium() [1/2]

template<typename Configuration, typename Layout, typename Options = SolverOptions<>, typename Observer = NoObserver, utils::dependencies::IsDependencyContext Inputs = utils::dependencies::DependencyContext<>, typename ObserverFactory = utils::dependencies::NoGraphObservers>
requires (!utils::dependencies::ContainsKey<discretization::blocks::surface_deformation::coefficients, typename Layout::Values>)

auto serif::solver::resume_equilibrium ( Configuration configuration, NewtonCheckpoint< model::StellarState< Layout > > checkpoint, const Vector & displacement, Options options = {}, Observer observer = {}, VolumeOptions volume_options = {}, Inputs prescribed = Inputs{}, ObserverFactory observer_factory = {} )

nodiscard

Continue a fixed-geometry checkpoint with its separately restored prescribed displacement.

Note

The optional final observer factory measures graph work separately from physics and solver history. It is rebuilt for this solve and is not stored in the checkpoint.

resume_equilibrium() [2/2]

template<typename Configuration, typename Layout, typename Options = SolverOptions<>, typename Observer = NoObserver, utils::dependencies::IsDependencyContext Inputs = utils::dependencies::DependencyContext<>, typename ObserverFactory = utils::dependencies::NoGraphObservers>
requires utils::dependencies::ContainsKey<discretization::blocks::surface_deformation::coefficients, typename Layout::Values>

auto serif::solver::resume_equilibrium ( Configuration configuration, NewtonCheckpoint< model::StellarState< Layout > > checkpoint, Options options = {}, Observer observer = {}, VolumeOptions volume_options = {}, Inputs prescribed = Inputs{}, ObserverFactory observer_factory = {} )

nodiscard

Continue a moving-star checkpoint using reconstructed physical and numerical configuration.

Note

The optional final observer factory measures graph work separately from physics and solver history. It is rebuilt for this solve and is not stored in the checkpoint.

resume_newton()

template<typename Problem, typename State, typename Options = SolverOptions<>, typename Observer = NoObserver>
requires detail::NewtonStrategyFor<Problem, State, Options>

auto serif::solver::resume_newton ( Problem & problem, NewtonCheckpoint< State > checkpoint, Options options = {}, Observer observer = {} )

nodiscard

Resume the original convergence reference, correction reuse, and cumulative work budget.

solve_equilibrium() [1/2]

template<typename Configuration, typename Layout, typename Options = SolverOptions<>, typename Observer = NoObserver, utils::dependencies::IsDependencyContext Inputs = utils::dependencies::DependencyContext<>, typename ObserverFactory = utils::dependencies::NoGraphObservers>
requires (!utils::dependencies::ContainsKey<discretization::blocks::surface_deformation::coefficients, typename Layout::Values>)

auto serif::solver::solve_equilibrium ( Configuration configuration, model::StellarState< Layout > initial_state, const Vector & displacement, Options options = {}, Observer observer = {}, VolumeOptions volume_options = {}, Inputs prescribed = Inputs{}, ObserverFactory observer_factory = {} )

nodiscard

Solve a fixed-geometry model with prescribed displacement outside its unknown vector.

Note

The optional final observer factory measures graph work separately from physics and solver history. It is rebuilt for this solve and is not stored in the checkpoint.

solve_equilibrium() [2/2]

template<typename Configuration, typename Layout, typename Options = SolverOptions<>, typename Observer = NoObserver, utils::dependencies::IsDependencyContext Inputs = utils::dependencies::DependencyContext<>, typename ObserverFactory = utils::dependencies::NoGraphObservers>
requires utils::dependencies::ContainsKey<discretization::blocks::surface_deformation::coefficients, typename Layout::Values>

auto serif::solver::solve_equilibrium ( Configuration configuration, model::StellarState< Layout > initial_state, Options options = {}, Observer observer = {}, VolumeOptions volume_options = {}, Inputs prescribed = Inputs{}, ObserverFactory observer_factory = {} )

nodiscard

Solve a moving star from an owning physical initial state.

Note

The model defines equilibrium; its evaluator and reusable cache are constructed internally. The returned result owns the last accepted physical star even when the solve does not converge. Initialize the backend's MPI runtime before using a recipe containing Hypre AMG. The optional final observer factory measures graph work separately from physics and solver history.

solve_newton()

template<typename Problem, typename State, typename Options = SolverOptions<>, typename Observer = NoObserver>
requires detail::NewtonStrategyFor<Problem, State, Options>

auto serif::solver::solve_newton ( Problem & problem, State initial_state, Options options = {}, Observer observer = {} )

nodiscard

Solve a custom numerical problem using its native Vector, evaluate, prepare and solve_linear operations.

Note

Strategies here are executable numerical methods; solve_equilibrium uses the same strategies.

validate_newton_progress()

std::optional< std::string > serif::solver::validate_newton_progress ( const NewtonProgress & progress, const std::size_t state_size )

inlinenodiscard

Reject incomplete or inconsistent imported history before preparing any runtime operators.