Reference documentation for deal.II version 9.6.0
\(\newcommand{\dealvcentcolon}{\mathrel{\mathop{:}}}\) \(\newcommand{\dealcoloneq}{\dealvcentcolon\mathrel{\mkern-1.2mu}=}\) \(\newcommand{\jump}[1]{\left[\!\left[ #1 \right]\!\right]}\) \(\newcommand{\average}[1]{\left\{\!\left\{ #1 \right\}\!\right\}}\)
Loading...
Searching...
No Matches
SUNDIALS::ARKode< VectorType > Class Template Reference

#include <deal.II/sundials/arkode.h>

Classes

class  AdditionalData
 

Public Member Functions

 ARKode (const AdditionalData &data=AdditionalData())
 
 ARKode (const AdditionalData &data, const MPI_Comm mpi_comm)
 
 ~ARKode ()
 
unsigned int solve_ode (VectorType &solution)
 
unsigned int solve_ode_incrementally (VectorType &solution, const double intermediate_time, const bool reset_solver=false)
 
void reset (const double t, const double h, const VectorType &y)
 
void * get_arkode_memory () const
 

Public Attributes

std::function< void(const double t, const VectorType &y, VectorType &explicit_f)> explicit_function
 
std::function< void(const double t, const VectorType &y, VectorType &res)> implicit_function
 
std::function< void(const double t, const VectorType &v, VectorType &Mv)> mass_times_vector
 
std::function< void(const double t)> mass_times_setup
 
std::function< void(const VectorType &v, VectorType &Jv, const double t, const VectorType &y, const VectorType &fy)> jacobian_times_vector
 
std::function< void(const double t, const VectorType &y, const VectorType &fy)> jacobian_times_setup
 
LinearSolveFunction< VectorType > solve_linearized_system
 
LinearSolveFunction< VectorType > solve_mass
 
std::function< void(const double t, const VectorType &y, const VectorType &fy, const VectorType &r, VectorType &z, const double gamma, const double tol, const int lr)> jacobian_preconditioner_solve
 
std::function< void(const double t, const VectorType &y, const VectorType &fy, const int jok, int &jcur, const double gamma)> jacobian_preconditioner_setup
 
std::function< void(const double t, const VectorType &r, VectorType &z, const double tol, const int lr)> mass_preconditioner_solve
 
std::function< void(const double t)> mass_preconditioner_setup
 
std::function< void(const double t, const VectorType &sol, const unsigned int step_number)> output_step
 
std::function< bool(const double t, VectorType &sol)> solver_should_restart
 
std::function< VectorType &()> get_local_tolerances
 
std::function< void(void *arkode_mem)> custom_setup
 

Private Member Functions

unsigned int do_evolve_time (VectorType &solution, ::DiscreteTime &time, const bool do_reset)
 
void setup_system_solver (const VectorType &solution)
 
void setup_mass_solver (const VectorType &solution)
 
void set_functions_to_trigger_an_assert ()
 

Static Private Member Functions

static ::ExceptionBaseExcFunctionNotProvided (std::string arg1)
 

Private Attributes

AdditionalData data
 
void * arkode_mem
 
SUNContext arkode_ctx
 
MPI_Comm mpi_communicator
 
double last_end_time
 
std::unique_ptr< internal::LinearSolverWrapper< VectorType > > linear_solver
 
std::unique_ptr< internal::LinearSolverWrapper< VectorType > > mass_solver
 
std::exception_ptr pending_exception
 

Detailed Description

template<typename VectorType = Vector<double>>
class SUNDIALS::ARKode< VectorType >

Interface to SUNDIALS additive Runge-Kutta methods (ARKode).

The class ARKode is a wrapper to SUNDIALS variable-step, embedded, additive Runge-Kutta solver which is a general purpose solver for systems of ordinary differential equations characterized by the presence of both fast and slow dynamics.

Fast dynamics are treated implicitly, and slow dynamics are treated explicitly, using nested families of implicit and explicit Runge-Kutta solvers.

Citing directly from ARKode documentation:

ARKode solves ODE initial value problems (IVPs) in \(R^N\). These problems should be posed in explicit form as

\[ M\dot y = f_E(t, y) + f_I (t, y), \qquad y(t_0) = y_0. \]

Here, \(t\) is the independent variable (e.g. time), and the dependent variables are given by \(y \in R^N\), and we use notation \(\dot y\) to denote \(dy/dt\). \(M\) is a user-supplied nonsingular operator from \(R^N \to R^N\). This operator may depend on \(t\) but not on \(y\).

For standard systems of ordinary differential equations and for problems arising from the spatial semi-discretization of partial differential equations using finite difference or finite volume methods, \(M\) is typically the identity matrix, \(I\). For PDEs using a finite-element spatial semi-discretization \(M\) is typically a well-conditioned mass matrix.

The two right-hand side functions may be described as:

  • \(f_E(t, y)\): contains the "slow" time scale components of the system. This will be integrated using explicit methods.
  • \(f_I(t, y)\): contains the "fast" time scale components of the system. This will be integrated using implicit methods.

ARKode may be used to solve stiff, nonstiff and multi-rate problems. Roughly speaking, stiffness is characterized by the presence of at least one rapidly damped mode, whose time constant is small compared to the time scale of the solution itself. In the implicit/explicit (ImEx) splitting above, these stiff components should be included in the right-hand side function \(f_I (t, y)\).

For multi-rate problems, a user should provide both of the functions \(f_E\) and \(f_I\) that define the IVP system.

For nonstiff problems, only \(f_E\) should be provided, and \(f_I\) is assumed to be zero, i.e. the system reduces to the non-split IVP:

\[ M\dot y = f_E(t, y), \qquad y(t_0) = y_0. \]

In this scenario, the ARK methods reduce to classical explicit Runge-Kutta methods (ERK). For these classes of methods, ARKode allows orders of accuracy \(q = \{2, 3, 4, 5, 6, 8\}\), with embeddings of orders \(p = \{1, 2, 3, 4, 5, 7\}\). These default to the Heun-Euler-2-1-2, Bogacki-Shampine-4-2-3, Zonneveld-5-3-4, Cash-Karp-6-4-5, Verner-8-5-6 and Fehlberg-13-7-8 methods, respectively.

Finally, for stiff (linear or nonlinear) problems the user may provide only \(f_I\), implying that \(f_E = 0\), so that the system reduces to the non-split IVP

\[ M\dot y = f_I(t, y), \qquad y(t_0) = y_0. \]

Similarly to ERK methods, in this scenario the ARK methods reduce to classical diagonally-implicit Runge-Kutta methods (DIRK). For these classes of methods, ARKode allows orders of accuracy \(q = \{2, 3, 4, 5\}\), with embeddings of orders \(p = \{1, 2, 3, 4\}\). These default to the SDIRK-2-1-2, ARK-4-2-3 (implicit), SDIRK-5-3-4 and ARK-8-4-5 (implicit) methods, respectively.

For both DIRK and ARK methods, an implicit system of the form

\[ G(z_i) \dealcoloneq M z_i - h_n A^I_{i,i} f_I (t^I_{n,i}, z_i) - a_i = 0 \]

must be solved for each stage \(z_i , i = 1, \ldots, s\), where we have the data

\[ a_i \dealcoloneq M y_{n-1} + h_n \sum_{j=1}^{i-1} [ A^E_{i,j} f_E(t^E_{n,j}, z_j) + A^I_{i,j} f_I (t^I_{n,j}, z_j)] \]

for the ARK methods, or

\[ a_i \dealcoloneq M y_{n-1} + h_n \sum_{j=1}^{i-1} A^I_{i,j} f_I (t^I_{n,j}, z_j) \]

for the DIRK methods. Here \(A^I_{i,j}\) and \(A^E_{i,j}\) are the Butcher's tables for the chosen solver.

If \(f_I(t,y)\) depends nonlinearly on \(y\) then the systems above correspond to a nonlinear system of equations; if \(f_I (t, y)\) depends linearly on \(y\) then this is a linear system of equations. By specifying the flag implicit_function_is_linear, ARKode takes some shortcuts that allow a faster solution process.

For systems of either type, ARKode allows a choice of solution strategy. The default solver choice is a variant of Newton's method,

\[ z_i^{m+1} = z_i^m +\delta^{m+1}, \]

where \(m\) is the Newton step index, and the Newton update \(\delta^{m+1}\) requires the solution of the linear Newton system

\[ N(z_i^m) \delta^{m+1} = -G(z_i^m), \]

where

\[ N \dealcoloneq M - \gamma J, \quad J \dealcoloneq \frac{\partial f_I}{\partial y}, \qquad \gamma\dealcoloneq h_n A^I_{i,i}. \]

As an alternate to Newton's method, ARKode may solve for each stage \(z_i ,i = 1, \ldots , s\) using an Anderson-accelerated fixed point iteration

\[ z_i^{m+1} = g(z_i^{m}), m=0,1,\ldots. \]

Unlike with Newton's method, this option does not require the solution of a linear system at each iteration, instead opting for solution of a low-dimensional least-squares solution to construct the nonlinear update.

Finally, if the user specifies implicit_function_is_linear, i.e., \(f_I(t, y)\) depends linearly on \(y\), and if the Newton-based nonlinear solver is chosen, then the system will be solved using only a single Newton iteration. Notice that in order for the Newton solver to be used, then jacobian_times_vector() should be supplied. If it is not supplied then only the fixed-point iteration will be supported, and the implicit_function_is_linear setting is ignored.

The optimal solver (Newton vs fixed-point) is highly problem-dependent. Since fixed-point solvers do not require the solution of any linear systems, each iteration may be significantly less costly than their Newton counterparts. However, this can come at the cost of slower convergence (or even divergence) in comparison with Newton-like methods. These fixed-point solvers do allow for user specification of the Anderson-accelerated subspace size, \(m_k\). While the required amount of solver memory grows proportionately to \(m_k N\), larger values of \(m_k\) may result in faster convergence.

This improvement may be significant even for "small" values, e.g. \(1 \leq m_k \leq 5\), and convergence may not improve (or even deteriorate) for larger values of \(m_k\). While ARKode uses a Newton-based iteration as its default solver due to its increased robustness on very stiff problems, it is highly recommended that users also consider the fixed-point solver for their cases when attempting a new problem.

For either the Newton or fixed-point solvers, it is well-known that both the efficiency and robustness of the algorithm intimately depends on the choice of a good initial guess. In ARKode, the initial guess for either nonlinear solution method is a predicted value \(z_i(0)\) that is computed explicitly from the previously-computed data (e.g. \(y_{n-2}, y_{n-1}\), and \(z_j\) where \(j < i\)). Additional information on the specific predictor algorithms implemented in ARKode is provided in ARKode documentation.

The user has to provide the implementation of at least one (or both) of the following std::functions:

If the mass matrix is different from the identity, the user should supply

If the use of a Newton method is desired, then the user should also supply jacobian_times_vector(). jacobian_times_setup() is optional.

Note
Although SUNDIALS can provide a difference quotient approximation of the Jacobian, this is currently not supported through this wrapper.

A SUNDIALS default solver (SPGMR) is used to solve the linear systems. To use a custom linear solver for the mass matrix and/or Jacobian, set:

To use a custom preconditioner with either a default or custom linear solver, set:

Also the following functions could be rewritten. By default they do nothing, or are not required.

To produce output at fixed steps, set the function

Any other custom settings of the ARKODE object can be specified in

To provide a simple example, consider the harmonic oscillator problem:

\[ \begin{split} u'' & = -k^2 u \\ u (0) & = 0 \\ u'(0) & = k \end{split} \]

We write it in terms of a first order ode:

\[ \begin{matrix} y_0' & = y_1 \\ y_1' & = - k^2 y_0 \end{matrix} \]

That is \(y' = A y\) where

\[ A \dealcoloneq \begin{pmatrix} 0 & 1 \\ -k^2 &0 \end{pmatrix} \]

and \(y(0)=(0, k)^T\).

The exact solution is \(y_0(t) = \sin(k t)\), \(y_1(t) = y_0'(t) = k \cos(k *t)\), \(y_1'(t) = -k^2 \sin(k t)\).

A minimal implementation, using only explicit RK methods, is given by the following code snippet:

using VectorType = Vector<double>;
const double k = 1.0;
ode.explicit_function = [k] (const double / * time * /,
const VectorType &y,
VectorType &ydot)
{
ydot[0] = y[1];
ydot[1] = -k*k*y[0];
};
y[1] = k;
ode.solve_ode(y);
std::function< void(const double t, const VectorType &y, VectorType &explicit_f)> explicit_function
Definition arkode.h:593
unsigned int solve_ode(VectorType &solution)
Definition arkode.cc:114

In practice, you would of course at least want to set the end time up to which the time integrator should run. In the example code shown here, it is taken from the default value defined by ARKode::AdditionalData.

Definition at line 328 of file arkode.h.

Constructor & Destructor Documentation

◆ ARKode() [1/2]

template<typename VectorType >
SUNDIALS::ARKode< VectorType >::ARKode ( const AdditionalData & data = AdditionalData())

Constructor, with class parameters set by the AdditionalData object.

Parameters
dataARKode configuration data
Note
With SUNDIALS 6 and later this constructor sets up logging objects to only work on the present processor (i.e., results are only communicated over MPI_COMM_SELF).

Definition at line 53 of file arkode.cc.

◆ ARKode() [2/2]

template<typename VectorType >
SUNDIALS::ARKode< VectorType >::ARKode ( const AdditionalData & data,
const MPI_Comm mpi_comm )

Constructor.

Parameters
dataARKode configuration data
mpi_commMPI Communicator over which logging operations are computed. Only used in SUNDIALS 6 and newer.

Definition at line 59 of file arkode.cc.

◆ ~ARKode()

template<typename VectorType >
SUNDIALS::ARKode< VectorType >::~ARKode ( )

Destructor.

Definition at line 97 of file arkode.cc.

Member Function Documentation

◆ solve_ode()

template<typename VectorType >
unsigned int SUNDIALS::ARKode< VectorType >::solve_ode ( VectorType & solution)

Integrate the initial value problem. This function returns the final number of computed steps.

Parameters
solutionOn input, this vector contains the initial condition. On output, it contains the solution at the final time.

Definition at line 114 of file arkode.cc.

◆ solve_ode_incrementally()

template<typename VectorType >
unsigned int SUNDIALS::ARKode< VectorType >::solve_ode_incrementally ( VectorType & solution,
const double intermediate_time,
const bool reset_solver = false )

Integrate the initial value problem. Compared to the function above, this function allows to specify an intermediate_time for the next solution. Repeated calls of this function must use monotonously increasing values for intermediate_time. The last solution state is saved internally along with the intermediate_time and will be reused as initial condition for the next call.

Users may find this function useful when integrating ARKode into an outer time loop of their own, especially when output_step() is too restrictive.

Note
intermediate_time may be larger than AdditionalData::final_time, which is ignored by this function.
Parameters
solutionThe final solution. If the solver restarts, either because it is the first ever solve or the flag reset_solver is set, the vector is also used as initial condition.
intermediate_timeThe time for the incremental solution step. Must be greater than the last time that was used in a previous call to this function.
reset_solverOptional flag to recreate all internal objects which may be desirable for spatial adaptivity methods. If set to true, reset() is called before solving the ODE, which sets solution as initial condition. This will not reset the stored time from previous calls to this function.

Definition at line 127 of file arkode.cc.

◆ reset()

template<typename VectorType >
void SUNDIALS::ARKode< VectorType >::reset ( const double t,
const double h,
const VectorType & y )

Clear internal memory and start with clean objects. This function is called when the simulation starts and when the user returns true to a call to solver_should_restart().

By default solver_should_restart() returns false. If the user needs to implement, for example, local adaptivity in space, he or she may assign a different function to solver_should_restart() that performs all mesh changes, transfers the solution to the new mesh, and returns true.

Parameters
tThe new starting time
hThe new starting time step
yThe new initial solution

Definition at line 247 of file arkode.cc.

◆ get_arkode_memory()

template<typename VectorType >
void * SUNDIALS::ARKode< VectorType >::get_arkode_memory ( ) const

Provides user access to the internally used ARKODE memory.

This functionality is intended for users who wish to query additional information directly from the ARKODE integrator, refer to the ARKODE manual for the various ARKStepGet... functions. The ARKStepSet... functions should not be called since this might lead to conflicts with various settings that are performed by this ARKode object.

Note
If custom settings of ARKODE functionality (that are not achievable via the interface of this class) are required, the function custom_setup() should be used.
Returns
pointer to the ARKODE memory block that can be passed to SUNDIALS functions

Definition at line 753 of file arkode.cc.

◆ do_evolve_time()

template<typename VectorType = Vector<double>>
unsigned int SUNDIALS::ARKode< VectorType >::do_evolve_time ( VectorType & solution,
::DiscreteTime & time,
const bool do_reset )
private

Internal routine to call ARKode repeatedly.

Definition at line 146 of file arkode.cc.

◆ setup_system_solver()

template<typename VectorType >
void SUNDIALS::ARKode< VectorType >::setup_system_solver ( const VectorType & solution)
private

Set up the (non)linear solver and preconditioners in the ARKODE memory object based on the user-specified functions.

Parameters
solutionThe solution vector which is used as a template to create new vectors.

Definition at line 392 of file arkode.cc.

◆ setup_mass_solver()

template<typename VectorType >
void SUNDIALS::ARKode< VectorType >::setup_mass_solver ( const VectorType & solution)
private

Set up the solver and preconditioner for a non-identity mass matrix in the ARKODE memory object based on the user-specified functions.

Parameters
solutionThe solution vector which is used as a template to create new vectors.

Definition at line 602 of file arkode.cc.

◆ set_functions_to_trigger_an_assert()

template<typename VectorType >
void SUNDIALS::ARKode< VectorType >::set_functions_to_trigger_an_assert ( )
private

This function is executed at construction time to set the std::function above to trigger an assert if they are not implemented.

Definition at line 742 of file arkode.cc.

Member Data Documentation

◆ explicit_function

template<typename VectorType = Vector<double>>
std::function< void(const double t, const VectorType &y, VectorType &explicit_f)> SUNDIALS::ARKode< VectorType >::explicit_function

A function object that users may supply and that is intended to compute the explicit part of the IVP right hand side. Sets \(explicit_f = f_E(t, y)\).

At least one of explicit_function() or implicit_function() must be provided. According to which one is provided, explicit, implicit, or mixed RK methods are used.

Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 593 of file arkode.h.

◆ implicit_function

template<typename VectorType = Vector<double>>
std::function<void(const double t, const VectorType &y, VectorType &res)> SUNDIALS::ARKode< VectorType >::implicit_function

A function object that users may supply and that is intended to compute the implicit part of the IVP right hand side. Sets \(implicit_f = f_I(t, y)\).

At least one of explicit_function() or implicit_function() must be provided. According to which one is provided, explicit, implicit, or mixed RK methods are used.

Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 612 of file arkode.h.

◆ mass_times_vector

template<typename VectorType = Vector<double>>
std::function<void(const double t, const VectorType &v, VectorType &Mv)> SUNDIALS::ARKode< VectorType >::mass_times_vector

A function object that users may supply and that is intended to compute the product of the mass matrix with a given vector v. This function will be called by ARKode (possibly several times) after mass_times_setup() has been called at least once. ARKode tries to do its best to call mass_times_setup() the minimum amount of times.

A call to this function should store in Mv the result of \(M\) applied to v.

Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 632 of file arkode.h.

◆ mass_times_setup

template<typename VectorType = Vector<double>>
std::function<void(const double t)> SUNDIALS::ARKode< VectorType >::mass_times_setup

A function object that users may supply and that is intended to set up the mass matrix. This function is called by ARKode any time a mass matrix update is required. The user should compute the mass matrix (or update all the variables that allow the application of the mass matrix). This function is guaranteed to be called by ARKode at least once, before any call to mass_times_vector().

ARKode supports the case where the mass matrix may depend on time, but not the case where the mass matrix depends on the solution itself.

If the user does not provide a mass_times_vector() function, then the identity is used. If the mass_times_setup() function is not provided, then mass_times_vector() should do all the work by itself.

If the user uses a matrix-based computation of the mass matrix, then this is the right place where an assembly routine should be called to assemble the matrix. Subsequent calls (possibly more than one) to mass_times_vector() can assume that this function has been called at least once.

Note
No assumption is made by this interface on what the user should do in this function. ARKode only assumes that after a call to mass_times_setup() it is possible to call mass_times_vector().
Parameters
tThe current evaluation time
Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 668 of file arkode.h.

◆ jacobian_times_vector

template<typename VectorType = Vector<double>>
std::function<void(const VectorType &v, VectorType &Jv, const double t, const VectorType &y, const VectorType &fy)> SUNDIALS::ARKode< VectorType >::jacobian_times_vector

A function object that users may supply and that is intended to compute the product of the Jacobian matrix with a given vector v. The Jacobian here refers to \(J=\frac{\partial f_I}{\partial y}\), i.e., the Jacobian of the user-specified implicit_function.

A call to this function should store in Jv the result of \(J\) applied to v.

Arguments to the function are

Parameters
[in]vThe vector to be multiplied by the Jacobian
[out]JvThe vector to be filled with the product J*v
[in]tThe current time
[in]yThe current \(y\) vector for the current ARKode internal step
[in]fyThe current value of the implicit right-hand side at y, \(f_I (t_n, y)\).
Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 701 of file arkode.h.

◆ jacobian_times_setup

template<typename VectorType = Vector<double>>
std::function< void(const double t, const VectorType &y, const VectorType &fy)> SUNDIALS::ARKode< VectorType >::jacobian_times_setup

A function object that users may supply and that is intended to set up all data necessary for the application of jacobian_times_vector(). This function is called by ARKode any time a Jacobian update is required. The user should compute the Jacobian (or update all the variables that allow the application of Jacobian). This function is guaranteed to be called by ARKode at least once, before any call to jacobian_times_vector().

If the jacobian_times_setup() function is not provided, then jacobian_times_vector() should do all the work by itself.

If the user uses a matrix based computation of the Jacobian, then this is the right place where an assembly routine should be called to assemble the matrix. Subsequent calls (possibly more than one) to jacobian_times_vector() can assume that this function has been called at least once.

Note
No assumption is made by this interface on what the user should do in this function. ARKode only assumes that after a call to jacobian_times_setup() it is possible to call jacobian_times_vector().
Parameters
tThe current time
yThe current ARKode internal solution vector \(y\)
fyThe implicit right-hand side function evaluated at the current time \(t\) and state \(y\), i.e., \(f_I(y,t)\)
Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 739 of file arkode.h.

◆ solve_linearized_system

template<typename VectorType = Vector<double>>
LinearSolveFunction<VectorType> SUNDIALS::ARKode< VectorType >::solve_linearized_system

A LinearSolveFunction object that users may supply and that is intended to solve the linearized system \(Ax=b\), where \(A = M-\gamma J\) is the Jacobian of the nonlinear residual. The application of the mass matrix \(M\) and Jacobian \(J\) are known through the functions mass_times_vector() and jacobian_times_vector() and \(\gamma\) is a factor provided by SUNDIALS. The matrix-vector product \(Ax\) is encoded in the supplied SundialsOperator. If a preconditioner was set through jacobian_preconditioner_solve(), it is encoded in the SundialsPreconditioner. If no preconditioner was supplied this way, the preconditioner is the identity matrix, i.e., no preconditioner. The user is free to use a custom preconditioner in this function object that is not supplied through SUNDIALS.

If you do not specify a solve_linearized_system() function, then a SUNDIALS packaged SPGMR solver with default settings is used.

For more details on the function type refer to LinearSolveFunction.

Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 767 of file arkode.h.

◆ solve_mass

template<typename VectorType = Vector<double>>
LinearSolveFunction<VectorType> SUNDIALS::ARKode< VectorType >::solve_mass

A LinearSolveFunction object that users may supply and that is intended to solve the mass system \(Mx=b\). The matrix-vector product \(Mx\) is encoded in the supplied SundialsOperator. If a preconditioner was set through mass_preconditioner_solve(), it is encoded in the SundialsPreconditioner. If no preconditioner was supplied this way, the preconditioner is the identity matrix, i.e., no preconditioner. The user is free to use a custom preconditioner in this function object that is not supplied through SUNDIALS.

The user must specify this function if a non-identity mass matrix is used and applied in mass_times_vector().

For more details on the function type refer to LinearSolveFunction.

Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 791 of file arkode.h.

◆ jacobian_preconditioner_solve

template<typename VectorType = Vector<double>>
std::function<void(const double t, const VectorType &y, const VectorType &fy, const VectorType &r, VectorType &z, const double gamma, const double tol, const int lr)> SUNDIALS::ARKode< VectorType >::jacobian_preconditioner_solve

A function object that users may supply to either pass a preconditioner to a SUNDIALS built-in solver or to apply a custom preconditioner within the user's own linear solve specified in solve_linearized_system().

This function should compute the solution to the preconditioner equation \(Pz=r\) and store it in z. In this equation \(P\) should approximate the Jacobian \(M-\gamma J\) of the nonlinear system.

Parameters
[in]tThe current time
[in]yThe current \(y\) vector for the current ARKode internal step
[in]fyThe current value of the implicit right-hand side at y, \(f_I (t_n, y)\).
[in]rThe right-hand side of the preconditioner equation
[out]zThe solution of applying the preconditioner, i.e., solving \(Pz=r\)
[in]gammaThe value \(\gamma\) in the preconditioner equation
[in]tolThe tolerance up to which the system should be solved
[in]lrAn input flag indicating whether the preconditioner solve is to use the left preconditioner (lr = 1) or the right preconditioner (lr = 2). Only relevant if used with a SUNDIALS packaged solver. If used with a custom solve_mass() function this will be set to zero.
Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 832 of file arkode.h.

◆ jacobian_preconditioner_setup

template<typename VectorType = Vector<double>>
std::function<void(const double t, const VectorType &y, const VectorType &fy, const int jok, int &jcur, const double gamma)> SUNDIALS::ARKode< VectorType >::jacobian_preconditioner_setup

A function object that users may supply to set up a preconditioner specified in jacobian_preconditioner_solve().

This function should prepare the solution of the preconditioner equation \(Pz=r\). In this equation \(P\) should approximate the Jacobian \(M-\gamma J\) of the nonlinear system.

If the jacobian_preconditioner_setup() function is not provided, then jacobian_preconditioner_solve() should do all the work by itself.

Note
No assumption is made by this interface on what the user should do in this function. ARKode only assumes that after a call to jacobian_preconditioner_setup() it is possible to call jacobian_preconditioner_solve().
Parameters
[in]tThe current time
[in]yThe current \(y\) vector for the current ARKode internal step
[in]fyThe current value of the implicit right-hand side at y, \(f_I (t_n, y)\).
[in]jokAn input flag indicating whether the Jacobian-related data needs to be updated. The jok argument provides for the reuse of Jacobian data in the preconditioner solve function. When jok = SUNFALSE, the Jacobian-related data should be recomputed from scratch. When jok = SUNTRUE the Jacobian data, if saved from the previous call to this function, can be reused (with the current value of gamma). A call with jok = SUNTRUE can only occur after a call with jok = SUNFALSE.
[out]jcurOn output this should be set to SUNTRUE if Jacobian data was recomputed, or set to SUNFALSE if Jacobian data was not recomputed, but saved data was still reused.
[in]gammaThe value \(\gamma\) in \(M-\gamma J\). The preconditioner should approximate the inverse of this matrix.
Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 882 of file arkode.h.

◆ mass_preconditioner_solve

template<typename VectorType = Vector<double>>
std::function<void(const double t, const VectorType &r, VectorType &z, const double tol, const int lr)> SUNDIALS::ARKode< VectorType >::mass_preconditioner_solve

A function object that users may supply to either pass a preconditioner to a SUNDIALS built-in solver or to apply a custom preconditioner within the user's own linear solve specified in solve_mass().

This function should compute the solution to the preconditioner equation \(Pz=r\) and store it in z. In this equation \(P\) should approximate the mass matrix \(M\).

Parameters
[in]tThe current time
[in]rThe right-hand side of the preconditioner equation
[out]zThe solution of applying the preconditioner, i.e., solving \(Pz=r\)
[in]gammaThe value \(\gamma\) in the preconditioner equation
[in]tolThe tolerance up to which the system should be solved
[in]lrAn input flag indicating whether the preconditioner solve is to use the left preconditioner (lr = 1) or the right preconditioner (lr = 2). Only relevant if used with a SUNDIALS packaged solver. If used with a custom solve_mass() function this will be set to zero.
Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 916 of file arkode.h.

◆ mass_preconditioner_setup

template<typename VectorType = Vector<double>>
std::function<void(const double t)> SUNDIALS::ARKode< VectorType >::mass_preconditioner_setup

A function object that users may supply to set up a preconditioner specified in mass_preconditioner_solve().

This function should prepare the solution of the preconditioner equation \(Pz=r\). In this equation \(P\) should approximate the mass matrix \(M\).

If the mass_preconditioner_setup() function is not provided, then mass_preconditioner_solve() should do all the work by itself.

Note
No assumption is made by this interface on what the user should do in this function. ARKode only assumes that after a call to mass_preconditioner_setup() it is possible to call mass_preconditioner_solve().
Parameters
[in]tThe current time
Note
This variable represents a user provided callback. See there for a description of how to deal with errors and other requirements and conventions. In particular, ARKode can deal with "recoverable" errors in some circumstances, so callbacks can throw exceptions of type RecoverableUserCallbackError.

Definition at line 942 of file arkode.h.

◆ output_step

template<typename VectorType = Vector<double>>
std::function<void(const double t, const VectorType &sol, const unsigned int step_number)> SUNDIALS::ARKode< VectorType >::output_step

A function object that users may supply and that is intended to postprocess the solution. This function is called by ARKode at fixed time increments (every output_period time units), and it is passed a polynomial interpolation of the solution, computed using the current ARK order and the (internally stored) previously computed solution steps.

Note
It is well possible that internally ARKode computes a time step which is much larger than the output_period step, and therefore calls this function consecutively several times by simply performing all intermediate interpolations. There is no relationship between how many times this function is called and how many time steps have actually been computed.

Definition at line 961 of file arkode.h.

◆ solver_should_restart

template<typename VectorType = Vector<double>>
std::function<bool(const double t, VectorType &sol)> SUNDIALS::ARKode< VectorType >::solver_should_restart

A function object that users may supply and that is intended to evaluate whether the solver should be restarted (for example because the number of degrees of freedom has changed).

This function is supposed to perform all operations that are necessary in sol to make sure that the resulting vectors are consistent, and of the correct final size.

For example, one may decide that a local refinement is necessary at time t. This function should then return true, and change the dimension of sol to reflect the new dimension. Since ARKode does not know about the new dimension, an internal reset is necessary.

The default implementation simply returns false, i.e., no restart is performed during the evolution.

Definition at line 980 of file arkode.h.

◆ get_local_tolerances

template<typename VectorType = Vector<double>>
std::function<VectorType &()> SUNDIALS::ARKode< VectorType >::get_local_tolerances

A function object that users may supply and that is intended to return a vector whose components are the weights used by ARKode to compute the vector norm. The implementation of this function is optional, and it is used only if implemented.

Definition at line 988 of file arkode.h.

◆ custom_setup

template<typename VectorType = Vector<double>>
std::function<void(void *arkode_mem)> SUNDIALS::ARKode< VectorType >::custom_setup

A function object that users may supply and which is intended to perform custom settings on the supplied arkode_mem object. Refer to the SUNDIALS documentation for valid options.

For instance, the following code attaches two files for diagnostic and error output of the internal ARKODE implementation:

ode.custom_setup = [&](void *arkode_mem) {
ARKStepSetErrFile(arkode_mem, errfile);
ARKStepSetDiagnostics(arkode_mem, diagnostics_file);
};
void * arkode_mem
Definition arkode.h:1068
Note
This function will be called at the end of all other set up right before the actual time evolution is started or continued with solve_ode(). This function is also called when the solver is restarted, see solver_should_restart(). Consult the SUNDIALS manual to see which options are still available at this point.
Parameters
arkode_mempointer to the ARKODE memory block which can be used for custom calls to ARKStepSet... methods.

Definition at line 1014 of file arkode.h.

◆ data

template<typename VectorType = Vector<double>>
AdditionalData SUNDIALS::ARKode< VectorType >::data
private

ARKode configuration data.

Definition at line 1063 of file arkode.h.

◆ arkode_mem

template<typename VectorType = Vector<double>>
void* SUNDIALS::ARKode< VectorType >::arkode_mem
private

ARKode memory object.

Definition at line 1068 of file arkode.h.

◆ arkode_ctx

template<typename VectorType = Vector<double>>
SUNContext SUNDIALS::ARKode< VectorType >::arkode_ctx
private

A context object associated with the ARKode solver.

Definition at line 1074 of file arkode.h.

◆ mpi_communicator

template<typename VectorType = Vector<double>>
MPI_Comm SUNDIALS::ARKode< VectorType >::mpi_communicator
private

MPI communicator. Only used for SUNDIALS' logging routines - the actual solve routines will use the communicator provided by the vector class.

Definition at line 1081 of file arkode.h.

◆ last_end_time

template<typename VectorType = Vector<double>>
double SUNDIALS::ARKode< VectorType >::last_end_time
private

The final time in the last call to solve_ode().

Definition at line 1086 of file arkode.h.

◆ linear_solver

template<typename VectorType = Vector<double>>
std::unique_ptr<internal::LinearSolverWrapper<VectorType> > SUNDIALS::ARKode< VectorType >::linear_solver
private

Definition at line 1088 of file arkode.h.

◆ mass_solver

template<typename VectorType = Vector<double>>
std::unique_ptr<internal::LinearSolverWrapper<VectorType> > SUNDIALS::ARKode< VectorType >::mass_solver
private

Definition at line 1089 of file arkode.h.

◆ pending_exception

template<typename VectorType = Vector<double>>
std::exception_ptr SUNDIALS::ARKode< VectorType >::pending_exception
mutableprivate

A pointer to any exception that may have been thrown in user-defined call-backs and that we have to deal after the KINSOL function we call has returned.

Definition at line 1096 of file arkode.h.


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