This class performs the anderson mixing in a variable agnostic way This class takes can take different input variables as input in a std::vector format and computes the mixing coefficients These coefficients can then be used to compute the new variable at the start of the SCF. More...

#include <MixingScheme.h>

Collaboration diagram for dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >:

[legend]

Public Types
using	ValueType = linearAlgebra::blasLapack::scalar_type< ValueTypeWeights, ValueTypeMixingVariable >

Public Member Functions
	MixingScheme (const utils::mpi::MPIComm &mpiComm)

	~MixingScheme ()=default

size_type	lengthOfHistory ()

void	computeAndersonMixingCoeff (const std::vector< mixingVariable > mixingVariablesList, linearAlgebra::LinAlgOpContext< utils::MemorySpace::HOST > &linAlgOpContextHost)
	Computes the mixing coefficients. More...

void	computeAdaptiveAndersonMixingParameter ()
	Computes the adaptive mixing parameter. More...

void	popOldHistory (size_type mixingHistory)
	Deletes the old history if the length exceeds max length of history. More...

void	clearHistory ()
	Clears all the the history. More...

void	addMixingVariable (const mixingVariable mixingVariableList, const utils::MemoryStorage< ValueTypeWeights, utils::MemorySpace::HOST > &weightDotProducts, const bool performMPIReduce, const double mixingValue, const bool isAdaptiveMixing)
	This function is used to add the mixing variables and its corresponding JxW values For dependent variables which are not used in mixing, the weightDotProducts is set to a vector of size zero. Later the dependent variables can be mixed with the same mixing coefficients. More...

template<utils::MemorySpace memorySpace>
void	addVariableToInHist (const mixingVariable mixingVariableName, const ValueTypeMixingVariable *inputVariableToInHist, const size_type length)
	Adds to the input history. More...

template<utils::MemorySpace memorySpace>
void	addVariableToResidualHist (const mixingVariable mixingVariableName, const ValueTypeMixingVariable *inputVariableToResidualHist, const size_type length)
	Adds to the residual history. More...

template<utils::MemorySpace memorySpace>
void	mixVariable (const mixingVariable mixingVariableName, ValueTypeMixingVariable *outputVariable, const size_type lenVar)
	Computes the input for the next iteration based on the anderson coefficients. More...

Private Member Functions
void	computeMixingMatrices (const std::deque< std::vector< ValueTypeMixingVariable > > &inHist, const std::deque< std::vector< ValueTypeMixingVariable > > &outHist, const utils::MemoryStorage< ValueTypeWeights, utils::MemorySpace::HOST > &weightDotProducts, const bool isPerformMixing, const bool isMPIAllReduce, std::vector< ValueType > &A, std::vector< ValueType > &c)
	Computes the matrix A and c vector that will be needed for anderson mixing. This function computes the A and c values for each variable which will be then added up in computeAndersonMixingCoeff() More...

Private Attributes
std::vector< ValueType >	d_A

std::vector< ValueType >	d_c

ValueType	d_cFinal

std::map< mixingVariable, std::deque< std::vector< ValueTypeMixingVariable > > >	d_variableHistoryIn

std::map< mixingVariable, std::deque< std::vector< ValueTypeMixingVariable > > >	d_variableHistoryResidual

std::map< mixingVariable, utils::MemoryStorage< ValueTypeWeights, utils::MemorySpace::HOST > >	d_vectorDotProductWeights

std::map< mixingVariable, bool >	d_performMPIReduce

const MPI_Comm &	d_mpiComm

std::map< mixingVariable, double >	d_mixingParameter

std::map< mixingVariable, bool >	d_isAdaptiveMixing

bool	d_anyMixingParameterAdaptive

bool	d_adaptiveMixingParameterDecLastIteration

bool	d_adaptiveMixingParameterDecAllIterations

bool	d_adaptiveMixingParameterIncAllIterations

size_type	d_mixingHistory

std::map< mixingVariable, bool >	d_performMixing

utils::ConditionalOStream	d_rootCout

Detailed Description

template<typename ValueTypeMixingVariable, typename ValueTypeWeights>
class dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >

This class performs the anderson mixing in a variable agnostic way This class takes can take different input variables as input in a std::vector format and computes the mixing coefficients These coefficients can then be used to compute the new variable at the start of the SCF.

Author: Vishal Subramanian

Member Typedef Documentation

◆ ValueType

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

using dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::ValueType = linearAlgebra::blasLapack::scalar_type<ValueTypeWeights, ValueTypeMixingVariable>

Constructor & Destructor Documentation

◆ MixingScheme()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::MixingScheme ( const utils::mpi::MPIComm & mpiComm )

Here is the call graph for this function:

◆ ~MixingScheme()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::~MixingScheme ( )

default

Member Function Documentation

◆ addMixingVariable()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

void dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::addMixingVariable	(	const mixingVariable	mixingVariableList,
		const utils::MemoryStorage< ValueTypeWeights, utils::MemorySpace::HOST > &	weightDotProducts,
		const bool	performMPIReduce,
		const double	mixingValue,
		const bool	isAdaptiveMixing
	)

This function is used to add the mixing variables and its corresponding JxW values For dependent variables which are not used in mixing, the weightDotProducts is set to a vector of size zero. Later the dependent variables can be mixed with the same mixing coefficients.

Here is the call graph for this function:

◆ addVariableToInHist()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

template<utils::MemorySpace memorySpace>

void dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::addVariableToInHist	(	const mixingVariable	mixingVariableName,
		const ValueTypeMixingVariable *	inputVariableToInHist,
		const size_type	length
	)

Adds to the input history.

Here is the call graph for this function:

◆ addVariableToResidualHist()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

template<utils::MemorySpace memorySpace>

void dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::addVariableToResidualHist	(	const mixingVariable	mixingVariableName,
		const ValueTypeMixingVariable *	inputVariableToResidualHist,
		const size_type	length
	)

Adds to the residual history.

Here is the call graph for this function:

◆ clearHistory()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

void dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::clearHistory

Clears all the the history.

◆ computeAdaptiveAndersonMixingParameter()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

void dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::computeAdaptiveAndersonMixingParameter

Computes the adaptive mixing parameter.

Here is the call graph for this function:

◆ computeAndersonMixingCoeff()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

void dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::computeAndersonMixingCoeff	(	const std::vector< mixingVariable >	mixingVariablesList,
		linearAlgebra::LinAlgOpContext< utils::MemorySpace::HOST > &	linAlgOpContextHost
	)

Computes the mixing coefficients.

◆ computeMixingMatrices()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

void dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::computeMixingMatrices	(	const std::deque< std::vector< ValueTypeMixingVariable > > &	inHist,
		const std::deque< std::vector< ValueTypeMixingVariable > > &	outHist,
		const utils::MemoryStorage< ValueTypeWeights, utils::MemorySpace::HOST > &	weightDotProducts,
		const bool	isPerformMixing,
		const bool	isMPIAllReduce,
		std::vector< ValueType > &	A,
		std::vector< ValueType > &	c
	)

private

Computes the matrix A and c vector that will be needed for anderson mixing. This function computes the A and c values for each variable which will be then added up in computeAndersonMixingCoeff()

◆ lengthOfHistory()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

size_type dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::lengthOfHistory

◆ mixVariable()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

template<utils::MemorySpace memorySpace>

void dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::mixVariable	(	const mixingVariable	mixingVariableName,
		ValueTypeMixingVariable *	outputVariable,
		const size_type	lenVar
	)

Computes the input for the next iteration based on the anderson coefficients.

Here is the call graph for this function:

◆ popOldHistory()

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

void dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::popOldHistory ( size_type mixingHistory )

Deletes the old history if the length exceeds max length of history.

Member Data Documentation

◆ d_A

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

std::vector<ValueType> dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_A

private

◆ d_adaptiveMixingParameterDecAllIterations

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

bool dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_adaptiveMixingParameterDecAllIterations

private

◆ d_adaptiveMixingParameterDecLastIteration

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

bool dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_adaptiveMixingParameterDecLastIteration

private

◆ d_adaptiveMixingParameterIncAllIterations

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

bool dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_adaptiveMixingParameterIncAllIterations

private

◆ d_anyMixingParameterAdaptive

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

bool dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_anyMixingParameterAdaptive

private

◆ d_c

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

std::vector<ValueType> dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_c

private

◆ d_cFinal

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

ValueType dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_cFinal

private

◆ d_isAdaptiveMixing

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

std::map<mixingVariable, bool> dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_isAdaptiveMixing

private

◆ d_mixingHistory

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

size_type dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_mixingHistory

private

◆ d_mixingParameter

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

std::map<mixingVariable, double> dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_mixingParameter

private

◆ d_mpiComm

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

const MPI_Comm& dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_mpiComm

private

◆ d_performMixing

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

std::map<mixingVariable, bool> dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_performMixing

private

◆ d_performMPIReduce

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

std::map<mixingVariable, bool> dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_performMPIReduce

private

◆ d_rootCout

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

utils::ConditionalOStream dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_rootCout

private

◆ d_variableHistoryIn

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

std::map<mixingVariable, std::deque<std::vector<ValueTypeMixingVariable> > > dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_variableHistoryIn

private

◆ d_variableHistoryResidual

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

std::map<mixingVariable, std::deque<std::vector<ValueTypeMixingVariable> > > dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_variableHistoryResidual

private

◆ d_vectorDotProductWeights

template<typename ValueTypeMixingVariable , typename ValueTypeWeights >

std::map<mixingVariable, utils::MemoryStorage<ValueTypeWeights, utils::MemorySpace::HOST> > dftefe::ksdft::MixingScheme< ValueTypeMixingVariable, ValueTypeWeights >::d_vectorDotProductWeights

private

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

src/ksdft/MixingScheme.h
src/ksdft/MixingScheme.t.cpp

Public Types

Public Member Functions

Private Member Functions

Private Attributes

Detailed Description

Member Typedef Documentation

◆ ValueType

Constructor & Destructor Documentation

◆ MixingScheme()

◆ ~MixingScheme()

Member Function Documentation

◆ addMixingVariable()

◆ addVariableToInHist()

◆ addVariableToResidualHist()

◆ clearHistory()

◆ computeAdaptiveAndersonMixingParameter()

◆ computeAndersonMixingCoeff()

◆ computeMixingMatrices()

◆ lengthOfHistory()

◆ mixVariable()

◆ popOldHistory()

Member Data Documentation

◆ d_A

◆ d_adaptiveMixingParameterDecAllIterations

◆ d_adaptiveMixingParameterDecLastIteration

◆ d_adaptiveMixingParameterIncAllIterations

◆ d_anyMixingParameterAdaptive

◆ d_c

◆ d_cFinal

◆ d_isAdaptiveMixing

◆ d_mixingHistory

◆ d_mixingParameter

◆ d_mpiComm

◆ d_performMixing

◆ d_performMPIReduce

◆ d_rootCout

◆ d_variableHistoryIn

◆ d_variableHistoryResidual

◆ d_vectorDotProductWeights