dftfe::nonlinearSolverProblem Class Reference

Abstract class for solver functions. More...

#include <nonlinearSolverProblem.h>

Inheritance diagram for dftfe::nonlinearSolverProblem:

Public Member Functions
	nonlinearSolverProblem ()
	Constructor.
virtual	~nonlinearSolverProblem ()=0
	Destructor.
virtual unsigned int	getNumberUnknowns () const =0
	Obtain number of unknowns.
virtual void	value (std::vector< double > &functionValue)=0
	Compute function value (aka energy).
virtual void	gradient (std::vector< double > &gradient)=0
	Compute function gradient (aka forces).
virtual void	precondition (std::vector< double > &s, const std::vector< double > &gradient)=0
	Apply preconditioner to function gradient.
virtual void	update (const std::vector< double > &solution, const bool computeForces=true, const bool useSingleAtomSolutionsInitialGuess=false)=0
	Update solution.
virtual void	solution (std::vector< double > &solution)=0
	Obtain current solution.
virtual std::vector< unsigned int >	getUnknownCountFlag () const =0
	For each unknown indicate whether that unknown should be accumulated. This functionality is needed in the case of parallel execution when domain decomposition is employed. Unknowns residing on processor boundary should only be accumulated once when dot products of vertex fields are computed (e.g. residual).
virtual bool	isConverged () const =0
	check for convergence.
virtual void	save ()=0
	create checkpoint for the current state of the problem i.e problem domain and solution.
virtual const MPI_Comm &	getMPICommunicator ()=0
	get MPI communicator.

Detailed Description

Abstract class for solver functions.

Author

Sambit Das

Constructor & Destructor Documentation

nonlinearSolverProblem()

dftfe::nonlinearSolverProblem::nonlinearSolverProblem

(

)

Constructor.

~nonlinearSolverProblem()

virtual dftfe::nonlinearSolverProblem::~nonlinearSolverProblem ( )

pure virtual

Destructor.

Member Function Documentation

getMPICommunicator()

virtual const MPI_Comm & dftfe::nonlinearSolverProblem::getMPICommunicator ( )

pure virtual

get MPI communicator.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

getNumberUnknowns()

virtual unsigned int dftfe::nonlinearSolverProblem::getNumberUnknowns ( ) const

pure virtual

Obtain number of unknowns.

Returns

Number of unknowns.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

getUnknownCountFlag()

virtual std::vector< unsigned int > dftfe::nonlinearSolverProblem::getUnknownCountFlag ( ) const

pure virtual

For each unknown indicate whether that unknown should be accumulated. This functionality is needed in the case of parallel execution when domain decomposition is employed. Unknowns residing on processor boundary should only be accumulated once when dot products of vertex fields are computed (e.g. residual).

Returns

A vector of int values for each unknown. Value of 1 indicates that the unknown should be counted and 0 otherwise.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

gradient()

virtual void dftfe::nonlinearSolverProblem::gradient ( std::vector< double > & gradient )

pure virtual

Compute function gradient (aka forces).

Parameters

gradient

STL vector for gradient values.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

isConverged()

virtual bool dftfe::nonlinearSolverProblem::isConverged ( ) const

pure virtual

check for convergence.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

precondition()

virtual void dftfe::nonlinearSolverProblem::precondition ( std::vector< double > & s, const std::vector< double > & gradient )

pure virtual

Apply preconditioner to function gradient.

Parameters

s	STL vector for s=-M^{-1} gradient.
gradient	STL vector for gradient values.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

save()

virtual void dftfe::nonlinearSolverProblem::save ( )

pure virtual

create checkpoint for the current state of the problem i.e problem domain and solution.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

solution()

virtual void dftfe::nonlinearSolverProblem::solution ( std::vector< double > & solution )

pure virtual

Obtain current solution.

Parameters

solution

Space for current solution.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

update()

virtual void dftfe::nonlinearSolverProblem::update ( const std::vector< double > & solution, const bool computeForces = true, const bool useSingleAtomSolutionsInitialGuess = false )

pure virtual

Update solution.

Parameters

solution

Updated solution.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

value()

virtual void dftfe::nonlinearSolverProblem::value ( std::vector< double > & functionValue )

pure virtual

Compute function value (aka energy).

Returns

Function value.

Implemented in dftfe::geoOptCell, dftfe::geoOptIon, and dftfe::nudgedElasticBandClass.

---

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

• workspace/include/nonlinearSolverProblem.h