TimeStepperSUNDIALS_IDA_impl< VectorType, SparseMatrixType, dim, LinearSolver > Class Template Reference

Implicit, adaptive time stepping with SUNDIALS IDA (variable-order BDF, order 1 to 5). More...

#include <sundials_ida.hh>

Inheritance diagram for TimeStepperSUNDIALS_IDA_impl< VectorType, SparseMatrixType, dim, LinearSolver >:

Public Types
using	NumberType = typename Base::NumberType
using	BlockVectorType = typename Base::BlockVectorType
using	IDAErrorDofCallback = std::function<void(const IDAErrorDofDiagnostics &)>

Public Member Functions
	TimeStepperSUNDIALS_IDA_impl (const ConfigTree &config, AbstractAssembler< VectorType, SparseMatrixType, dim > &assembler, OutputSession_impl< dim, VectorType > &data_out)
	Construct without mesh adaptation; see AbstractTimestepper.
	TimeStepperSUNDIALS_IDA_impl (const ConfigTree &config, AbstractAssembler< VectorType, SparseMatrixType, dim > &assembler, OutputSession_impl< dim, VectorType > &data_out, AbstractAdaptor< VectorType > &adaptor)
	Construct with a mesh adaptor; see AbstractTimestepper.
virtual void	run_segment (AbstractFlowingVariables< NumberType, VectorType > &initial_condition, const double t_start, const double t_stop) override
	Any derived class must implement this method to run the timestepping algorithm.
void	set_ida_error_dof_callback (IDAErrorDofCallback callback)
	Called with every report of /timestepping/implicit/ida_error_dof_diagnostics, in addition to the log.
Public Member Functions inherited from AbstractTimestepper< VectorType, SparseMatrixType, dim >
	AbstractTimestepper (const ConfigTree &config, AbstractAssembler< VectorType, SparseMatrixType, dim > &assembler, OutputSession_impl< dim, VectorType > &data_out, const bool implicit_stepper, const bool explicit_stepper)
	Construct without mesh adaptation.
	AbstractTimestepper (const ConfigTree &config, AbstractAssembler< VectorType, SparseMatrixType, dim > &assembler, OutputSession_impl< dim, VectorType > &data_out, AbstractAdaptor< VectorType > &adaptor, const bool implicit_stepper, const bool explicit_stepper)
	Construct with a mesh adaptor, which is called between steps (where the stepper supports it, see abstract_timestepper.hh).
void	run (AbstractFlowingVariables< NumberType, VectorType > &initial_condition, double t_start, const double t_stop)
	Evolve initial_condition from t_start to t_stop.
bool	is_implicit () const
	Whether this stepper reads /timestepping/implicit.
bool	is_explicit () const
	Whether this stepper reads /timestepping/explicit.

Additional Inherited Members
Protected Types inherited from AbstractTimestepper< VectorType, SparseMatrixType, dim >
using	VectorType
using	NumberType
using	SparseMatrixType
using	BlockVectorType
Protected Member Functions inherited from AbstractTimestepper< VectorType, SparseMatrixType, dim >
void	drain_output ()
void	finalize_output_after_failure (EmitFinalFrame &&emit_final_frame)
	Best-effort output cleanup while a timestepping failure is being reported.
void	require_no_adaptor (const std::string &stepper) const
	Throw unless the mesh adaptor is NoAdaptivity; for steppers that cannot adapt the mesh.
void	require_no_variables (const AbstractFlowingVariables< NumberType, VectorType > &state, const std::string &stepper) const
	Throw if the flow has variables; for steppers that integrate only the FE/FV unknowns.
Protected Attributes inherited from AbstractTimestepper< VectorType, SparseMatrixType, dim >
const ConfigTree	config
NoAdaptivity< VectorType >	adaptor_default
AbstractAssembler< VectorType, SparseMatrixType, dim > &	assembler
OutputSession_impl< dim, VectorType > &	data_out
AbstractAdaptor< VectorType > &	adaptor
ReportPort	log
const bool	m_is_implicit
const bool	m_is_explicit
double	output_dt
struct DiFfRG::AbstractTimestepper::ImplicitParameters	impl
struct DiFfRG::AbstractTimestepper::ExplicitParameters	expl
std::size_t	next_jacobian_build_id
TimestepperJacobianDiagnosticsState	jacobian_diagnostics_state
Static Protected Attributes inherited from AbstractTimestepper< VectorType, SparseMatrixType, dim >
static constexpr uint	dim

Detailed Description

template<typename VectorType, typename SparseMatrixType, uint dim, template< typename... > typename LinearSolver>
class DiFfRG::TimeStepperSUNDIALS_IDA_impl< VectorType, SparseMatrixType, dim, LinearSolver >

Implicit, adaptive time stepping with SUNDIALS IDA (variable-order BDF, order 1 to 5).

IDA solves the system in residual form, \(F(t, y, \dot y) = 0\), with \(y = (u, v)\) and

\[ F_u = m(u, \dot u) + R(u, v)\,, \qquad F_v = \dot v + D(v, u)\,, \]

see abstract_timestepper.hh for the notation. Each step solves \(F = 0\) by Newton's method with the matrix \(\partial F/\partial y + \alpha\, \partial F/\partial \dot y\), where \(\alpha\) is set by IDA from the step size and the BDF order. The \(u\)- and \(v\)-blocks of this matrix are built separately, i.e. the couplings \(\partial F_u/\partial v\) and \(\partial F_v/\partial u\) are left out; they slow Newton's convergence but do not change the solution.

Components of \(u\) whose mass term does not depend on \(\dot u\) are algebraic constraints; they are excluded from the error test. At the start and after every restart (e.g. after mesh adaptation), IDA computes the algebraic components and the \(\dot u\) of the others from \(F = 0\). Output is written at exactly \(t_\text{start} + n\, \mathrm{output\_dt}\), interpolated by IDA.

Handles FE problems with or without variables, and pure-variables systems (dim == 0). Distributed linear algebra is supported. A callback that fails (non-finite state or residual, an exception from the model, a failed factorization or linear solve) makes IDA retry with a smaller step; the run aborts after 200 consecutive failures or if the time stops advancing.

Reads /timestepping/implicit, see abstract_timestepper.hh.

Template Parameters

VectorType	vector type of the assembler.
SparseMatrixType	sparse matrix type of the assembler.
dim	spatial dimension, 0 for a pure-variables system.
LinearSolver	solver for the Newton systems; see DefaultLinearSolver.

Member Typedef Documentation

BlockVectorType

template<typename VectorType , typename SparseMatrixType , uint dim, template< typename... > typename LinearSolver>

using BlockVectorType = typename Base::BlockVectorType

IDAErrorDofCallback

template<typename VectorType , typename SparseMatrixType , uint dim, template< typename... > typename LinearSolver>

using IDAErrorDofCallback = std::function<void(const IDAErrorDofDiagnostics &)>

NumberType

template<typename VectorType , typename SparseMatrixType , uint dim, template< typename... > typename LinearSolver>

using NumberType = typename Base::NumberType

Constructor & Destructor Documentation

TimeStepperSUNDIALS_IDA_impl() [1/2]

template<typename VectorType , typename SparseMatrixType , uint dim, template< typename... > typename LinearSolver>

TimeStepperSUNDIALS_IDA_impl ( const ConfigTree & config, AbstractAssembler< VectorType, SparseMatrixType, dim > & assembler, OutputSession_impl< dim, VectorType > & data_out )

inline

Construct without mesh adaptation; see AbstractTimestepper.

TimeStepperSUNDIALS_IDA_impl() [2/2]

template<typename VectorType , typename SparseMatrixType , uint dim, template< typename... > typename LinearSolver>

TimeStepperSUNDIALS_IDA_impl ( const ConfigTree & config, AbstractAssembler< VectorType, SparseMatrixType, dim > & assembler, OutputSession_impl< dim, VectorType > & data_out, AbstractAdaptor< VectorType > & adaptor )

inline

Construct with a mesh adaptor; see AbstractTimestepper.

Member Function Documentation

run_segment()

template<typename VectorType , typename SparseMatrixType , uint dim, template< typename... > typename LinearSolver>

virtual void run_segment ( AbstractFlowingVariables< NumberType, VectorType > & initial_condition, const double t_start, const double t_stop )

overridevirtual

Any derived class must implement this method to run the timestepping algorithm.

It must return the state at t_stop in initial_condition, because run() writes snapshots from it. Called once per segment, i.e. possibly several times per run() with consecutive intervals.

Parameters

initial_condition	The flowing variables object that contains the initial condition.
t_start	The start time of the segment.
t_stop	The end time of the segment.

Implements AbstractTimestepper< VectorType, SparseMatrixType, dim >.

set_ida_error_dof_callback()

template<typename VectorType , typename SparseMatrixType , uint dim, template< typename... > typename LinearSolver>

void set_ida_error_dof_callback ( IDAErrorDofCallback callback )

inline

Called with every report of /timestepping/implicit/ida_error_dof_diagnostics, in addition to the log.

---

The documentation for this class was generated from the following file:

• /home/runner/work/DiFfRG/DiFfRG/DiFfRG/include/DiFfRG/timestepping/sundials_ida.hh