IBAMR
An adaptive and distributed-memory parallel implementation of the immersed boundary (IB) method
|
Class PETScNewtonKrylovSolver provides a NewtonKrylovSolver interface for a PETSc inexact Newton-Krylov iterative nonlinear solver (SNES). More...
#include </home/runner/work/IBAMR/IBAMR/ibtk/include/ibtk/PETScNewtonKrylovSolver.h>
Public Member Functions | |
PETScNewtonKrylovSolver (std::string object_name, SAMRAI::tbox::Pointer< SAMRAI::tbox::Database > input_db, std::string default_options_prefix, MPI_Comm petsc_comm=PETSC_COMM_WORLD) | |
Constructor for a concrete NewtonKrylovSolver that employs the PETSc SNES solver framework. | |
PETScNewtonKrylovSolver (std::string object_name, SNES petsc_snes) | |
Constructor for a concrete NewtonKrylovSolver that acts as a "wrapper" for a supplied PETSc SNES object. More... | |
~PETScNewtonKrylovSolver () | |
Destructor. | |
void | setOptionsPrefix (const std::string &options_prefix) |
Set the options prefix used by this PETSc solver object. | |
Functions to access the underlying PETSc objects. | |
const SNES & | getPETScSNES () const |
Get the PETSc SNES object. | |
Public Member Functions inherited from IBTK::NewtonKrylovSolver | |
NewtonKrylovSolver () | |
Default constructor. | |
virtual | ~NewtonKrylovSolver ()=default |
Empty virtual destructor. | |
void | setHierarchyMathOps (SAMRAI::tbox::Pointer< HierarchyMathOps > hier_math_ops) override |
Set the HierarchyMathOps object used by the solver. | |
void | setHomogeneousBc (bool homogeneous_bc) override |
Set whether the solver should use homogeneous boundary conditions. | |
void | setSolutionTime (double solution_time) override |
Set the time at which the solution is to be evaluated. | |
void | setTimeInterval (double current_time, double new_time) override |
Set the current time interval. | |
virtual SAMRAI::tbox::Pointer< GeneralOperator > | getOperator () const |
Retrieve the nonlinear operator used by the solver. | |
virtual SAMRAI::tbox::Pointer< JacobianOperator > | getJacobian () const |
Retrieve the Jacobian operator used by the solver. | |
virtual SAMRAI::tbox::Pointer< KrylovLinearSolver > | getLinearSolver () const |
Retrieve the Krylov linear solver used in computing Newton step directions. | |
virtual void | setMaxEvaluations (int max_evaluations) |
Set the maximum number of function evaluations to use per solve. | |
virtual int | getMaxEvaluations () const |
Get the maximum number of function evaluations to use per solve. | |
virtual void | setSolutionTolerance (double solution_tol) |
Set the tolerance in terms of the norm of the change in the solution between steps. | |
virtual double | getSolutionTolerance () const |
Get the tolerance in terms of the norm of the change in the solution between steps. | |
virtual int | getNumLinearIterations () const |
Return the number of linear iterations from the most recent nonlinear solve. | |
Public Member Functions inherited from IBTK::GeneralSolver | |
GeneralSolver ()=default | |
Constructor. | |
virtual | ~GeneralSolver ()=default |
Empty virtual destructor. | |
const std::string & | getName () const |
Return the object name. | |
virtual bool | getIsInitialized () const |
Return whether the operator is initialized. | |
virtual bool | getHomogeneousBc () const |
Return whether the solver is using homogeneous boundary conditions. | |
virtual double | getSolutionTime () const |
Get the time at which the solution is being evaluated. | |
virtual std::pair< double, double > | getTimeInterval () const |
Get the current time interval. | |
virtual double | getDt () const |
Get the current time step size. | |
virtual SAMRAI::tbox::Pointer< HierarchyMathOps > | getHierarchyMathOps () const |
Get the HierarchyMathOps object used by the solver. | |
virtual void | setMaxIterations (int max_iterations) |
Set the maximum number of nonlinear iterations to use per solve. | |
virtual int | getMaxIterations () const |
Get the maximum number of nonlinear iterations to use per solve. | |
virtual void | setAbsoluteTolerance (double abs_residual_tol) |
Set the absolute residual tolerance for convergence. | |
virtual double | getAbsoluteTolerance () const |
Get the absolute residual tolerance for convergence. | |
virtual void | setRelativeTolerance (double rel_residual_tol) |
Set the relative residual tolerance for convergence. | |
virtual double | getRelativeTolerance () const |
Get the relative residual tolerance for convergence. | |
virtual int | getNumIterations () const |
Return the iteration count from the most recent solve. | |
virtual double | getResidualNorm () const |
Return the residual norm from the most recent iteration. | |
virtual void | setLoggingEnabled (bool enable_logging=true) |
Enable or disable logging. | |
virtual bool | getLoggingEnabled () const |
Determine whether logging is enabled or disabled. | |
virtual void | printClassData (std::ostream &stream) |
Print class data to stream. | |
Static Public Member Functions | |
static SAMRAI::tbox::Pointer< NewtonKrylovSolver > | allocate_solver (const std::string &object_name, SAMRAI::tbox::Pointer< SAMRAI::tbox::Database > input_db, const std::string &default_options_prefix) |
Static function to construct a PETScNewtonKrylovSolver. | |
Newton-Krylov solver functionality. | |
void | setOperator (SAMRAI::tbox::Pointer< GeneralOperator > op) override |
Set the nonlinear operator used by the solver. | |
SAMRAI::tbox::Pointer< SAMRAI::solv::SAMRAIVectorReal< NDIM, double > > | getSolutionVector () const override |
Return the vector in which the approximate solution is stored. | |
SAMRAI::tbox::Pointer< SAMRAI::solv::SAMRAIVectorReal< NDIM, double > > | getFunctionVector () const override |
Return the vector in which the nonlinear function evaluation is stored. | |
void | setJacobian (SAMRAI::tbox::Pointer< JacobianOperator > J) override |
Set the Jacobian operator used by the solver. More... | |
bool | solveSystem (SAMRAI::solv::SAMRAIVectorReal< NDIM, double > &x, SAMRAI::solv::SAMRAIVectorReal< NDIM, double > &b) override |
Solve the system for . More... | |
void | initializeSolverState (const SAMRAI::solv::SAMRAIVectorReal< NDIM, double > &x, const SAMRAI::solv::SAMRAIVectorReal< NDIM, double > &b) override |
Compute hierarchy dependent data required for solving . More... | |
void | deallocateSolverState () override |
Remove all hierarchy dependent data allocated by initializeSolverState(). More... | |
Additional Inherited Members | |
Protected Member Functions inherited from IBTK::GeneralSolver | |
void | init (const std::string &object_name, bool homogeneous_bc) |
virtual void | initSpecialized (const std::string &object_name, bool homogeneous_bc) |
Protected Attributes inherited from IBTK::NewtonKrylovSolver | |
SAMRAI::tbox::Pointer< GeneralOperator > | d_F |
SAMRAI::tbox::Pointer< JacobianOperator > | d_J |
SAMRAI::tbox::Pointer< KrylovLinearSolver > | d_krylov_solver |
SAMRAI::tbox::Pointer< SAMRAI::solv::SAMRAIVectorReal< NDIM, double > > | d_x |
SAMRAI::tbox::Pointer< SAMRAI::solv::SAMRAIVectorReal< NDIM, double > > | d_b |
SAMRAI::tbox::Pointer< SAMRAI::solv::SAMRAIVectorReal< NDIM, double > > | d_r |
int | d_max_evaluations = 10000 |
double | d_solution_tol = 1.0e-8 |
int | d_current_linear_iterations = 0 |
Protected Attributes inherited from IBTK::GeneralSolver | |
std::string | d_object_name = "unitialized" |
bool | d_is_initialized = false |
bool | d_homogeneous_bc = false |
double | d_solution_time = std::numeric_limits<double>::quiet_NaN() |
double | d_current_time = std::numeric_limits<double>::quiet_NaN() |
double | d_new_time = std::numeric_limits<double>::quiet_NaN() |
double | d_rel_residual_tol = 0.0 |
double | d_abs_residual_tol = 0.0 |
int | d_max_iterations = 100 |
int | d_current_iterations = 0 |
double | d_current_residual_norm = std::numeric_limits<double>::quiet_NaN() |
SAMRAI::tbox::Pointer< HierarchyMathOps > | d_hier_math_ops |
bool | d_hier_math_ops_external = false |
bool | d_enable_logging = false |
Class PETScNewtonKrylovSolver provides a NewtonKrylovSolver interface for a PETSc inexact Newton-Krylov iterative nonlinear solver (SNES).
This solver class provides access to inexact Newton-Krylov methods, including line search and trust region Newton methods, using the PETSc SNES nonlinear solver interface. Note that solver configuration is typically done at runtime via command line options.
Sample parameters for initialization from database (and their default values):
options_prefix = "" // see setOptionsPrefix() rel_residual_tol = 1.0e-8 // see setRelativeTolerance() abs_residual_tol = 1.0e-50 // see setAbsoluteTolerance() solution_tol = 1.0e-8 // see setSolutionTolerance() max_iterations = 50 // see setMaxIterations() enable_logging = FALSE // see setLoggingEnabled()
PETSc is developed in the Mathematics and Computer Science (MCS) Division at Argonne National Laboratory (ANL). For more information about PETSc, see http://www.mcs.anl.gov/petsc.
IBTK::PETScNewtonKrylovSolver::PETScNewtonKrylovSolver | ( | std::string | object_name, |
SNES | petsc_snes | ||
) |
Constructor for a concrete NewtonKrylovSolver that acts as a "wrapper" for a supplied PETSc SNES object.
object_name | Name of the solver |
petsc_snes | PETSc SNES object |
|
overridevirtual |
Remove all hierarchy dependent data allocated by initializeSolverState().
When any operator objects have been registered with this class via setOperator() or setJacobian(), they are also deallocated by this member function.
Reimplemented from IBTK::GeneralSolver.
|
overridevirtual |
Compute hierarchy dependent data required for solving .
By default, the solveSystem() method computes some required hierarchy dependent data before solving and removes that data after the solve. For multiple solves that use the same hierarchy configuration, it is more efficient to:
Note that it is generally necessary to reinitialize the solver state when the hierarchy configuration changes.
When any operator objects have been registered with this class via setOperator() or setJacobian(), they are also initialized by this member function.
x | solution vector |
b | right-hand-side vector |
Conditions on Parameters:
Reimplemented from IBTK::GeneralSolver.
|
overridevirtual |
Set the Jacobian operator used by the solver.
Reimplemented from IBTK::NewtonKrylovSolver.
|
overridevirtual |
Solve the system for .
Before calling solveSystem(), the form of the solution x and right-hand-side b vectors must be set properly by the user on all patch interiors on the specified range of levels in the patch hierarchy. The user is responsible for all data management for the quantities associated with the solution and right-hand-side vectors. In particular, patch data in these vectors must be allocated prior to calling this method.
x | solution vector |
b | right-hand-side vector |
Conditions on Parameters:
true
if the solver converged to the specified tolerances, false
otherwise Implements IBTK::GeneralSolver.