A Module For Performing Distributed Sparse Matrix Operations. More...

#include <PSMatrix.h>

Public Member Functions
	Matrix_ps (int matrix_dimension)

	Matrix_ps (int matrix_dimension, const ProcessGrid &grid)

	Matrix_ps (std::string file_name, bool is_binary=false)

	Matrix_ps (std::string file_name, const ProcessGrid &grid, bool is_binary=false)

	Matrix_ps (const Matrix_ps &matB)

void	WriteToBinary (std::string file_name) const

void	WriteToMatrixMarket (std::string file_name) const

void	FillFromTripletList (const TripletList_r &triplet_list)

void	FillFromTripletList (const TripletList_c &triplet_list)

void	FillDistributedPermutation (const Permutation &lb, bool permuterows=true)

void	FillIdentity ()
	Fills this matrix as the identity matrix.

void	FillDense ()

int	GetActualDimension () const
	get the actual dimension of the matrix.

int	GetLogicalDimension () const
	the logical dimension is scaled so each process has an even slice.

long int	GetSize () const
	Get the total number of non-zero entries in the matrix.

void	GetTripletList (TripletList_r &triplet_list) const

void	GetTripletList (TripletList_c &triplet_list) const

void	GetMatrixBlock (TripletList_r &triplet_list, int start_row, int end_row, int start_column, int end_column)

void	GetMatrixBlock (TripletList_c &triplet_list, int start_row, int end_row, int start_column, int end_column)

void	GetMatrixSlice (Matrix_ps &submatrix, int start_row, int end_row, int start_column, int end_column)

bool	IsIdentity () const
	Determine if this is the identity matrix.

void	Transpose (const NTPoly::Matrix_ps &matA)
	Transpose a sparse matrix.

void	Conjugate ()
	Compute the complex conjugate of a matrix.

void	Resize (int new_size)

double	Dot (const Matrix_ps &matB)

std::complex< double >	Dot_c (const Matrix_ps &matB)

void	Increment (const Matrix_ps &matB, double alpha=1.0, double threshold=0.0)

void	PairwiseMultiply (const Matrix_ps &matA, const Matrix_ps &matB)
	Elementwise multiplication. C_ij = A_ij * B_ij. Also known as a Hadamard.

void	Gemm (const Matrix_ps &matA, const Matrix_ps &matB, PMatrixMemoryPool &memory_pool, double alpha=1.0, double beta=0.0, double threshold=0.0)

void	Scale (double constant)

double	Norm () const
	compute the norm of a matrix.

double	Trace () const
	compute the trace of a matrix.

	~Matrix_ps ()
	Destructor.

Friends
class	LoadBalancer

class	SolverBase

template<class T >
class	TripletList

class	PMatrixMemoryPool

class	MatrixMapper

class	MatrixConversion

Detailed Description

A Module For Performing Distributed Sparse Matrix Operations.

Constructor & Destructor Documentation

◆ Matrix_ps() [1/5]

NTPoly::Matrix_ps::Matrix_ps ( int matrix_dimension )

Construct an empty matrix.

Parameters

matrix_dimension size fo the matrix.

◆ Matrix_ps() [2/5]

NTPoly::Matrix_ps::Matrix_ps	(	int	matrix_dimension,
		const ProcessGrid &	grid
	)

Construct an empty matrix.

Parameters

matrix_dimension	size fo the matrix.
grid	the process grid this matrix is distributed on.

◆ Matrix_ps() [3/5]

NTPoly::Matrix_ps::Matrix_ps	(	std::string	file_name,
		bool	is_binary = `false`
	)

Construct a matrix from file.

Parameters

file_name	name of the file to build from.
is_binary	true if the file is a binary file.

◆ Matrix_ps() [4/5]

NTPoly::Matrix_ps::Matrix_ps	(	std::string	file_name,
		const ProcessGrid &	grid,
		bool	is_binary = `false`
	)

Construct a matrix from file.

Parameters

file_name	name of the file to build from.
is_binary	true if the file is a binary file.
grid	the process grid this matrix is distributed on.

◆ Matrix_ps() [5/5]

NTPoly::Matrix_ps::Matrix_ps ( const Matrix_ps & matB )

Copy constructor.

Parameters

matB	matrix to copy from.

Member Function Documentation

◆ Dot()

double NTPoly::Matrix_ps::Dot ( const Matrix_ps & matB )

this = dot(this,matB)

Parameters

matB	the matrix to dot.

◆ Dot_c()

complex< double > NTPoly::Matrix_ps::Dot_c ( const Matrix_ps & matB )

this = dot(this,matB)

Parameters

matB	the matrix to dot.

◆ FillDense()

void NTPoly::Matrix_ps::FillDense ( )

This routine will fill a dense matrix so that every element has a given a value of 1. This is useful as a starting point for further filtering or mapping operations.

◆ FillDistributedPermutation()

void NTPoly::Matrix_ps::FillDistributedPermutation	(	const Permutation &	lb,
		bool	permuterows = `true`
	)

Fill the matrix based on a permutation.

Parameters

lb	the permutation.
permuterows	true if this is a row permutation matrix.

◆ FillFromTripletList() [1/2]

void NTPoly::Matrix_ps::FillFromTripletList ( const TripletList_c & triplet_list )

Fill in the matrix based on the contents of triplet lists.

Parameters

triplet_list list of values. Need to be absolute coordinates.

◆ FillFromTripletList() [2/2]

void NTPoly::Matrix_ps::FillFromTripletList ( const TripletList_r & triplet_list )

Fill in the matrix based on the contents of triplet lists.

Parameters

triplet_list list of values. Need to be absolute coordinates.

◆ Gemm()

void NTPoly::Matrix_ps::Gemm	(	const Matrix_ps &	matA,
		const Matrix_ps &	matB,
		PMatrixMemoryPool &	memory_pool,
		double	alpha = `1.0`,
		double	beta = `0.0`,
		double	threshold = `0.0`
	)

this := alpha*matA*matB+ beta*this (GEMM)

Parameters

matA	first mat.
matB	second mat.
memory_pool	memory pool for intermediates.
alpha	scaling factor.
beta	scaling factor.
threshold	for flushing small values.

◆ GetMatrixBlock() [1/2]

void NTPoly::Matrix_ps::GetMatrixBlock	(	TripletList_c &	triplet_list,
		int	start_row,
		int	end_row,
		int	start_column,
		int	end_column
	)

Extract an arbitrary block of a matrix into a triplet list. Block is defined by the row/column start/end values. This is slower than GetTripletList, because communication is required. Data is returned with absolute coordinates.

Parameters

triplet_list	the list to fill.
start_row	the starting row for data to store on this process.
end_row	the ending row for data to store on this process.
start_column	the starting col for data to store on this process
end_column	the ending col for data to store on this process

◆ GetMatrixBlock() [2/2]

void NTPoly::Matrix_ps::GetMatrixBlock	(	TripletList_r &	triplet_list,
		int	start_row,
		int	end_row,
		int	start_column,
		int	end_column
	)

Extract an arbitrary block of a matrix into a triplet list. Block is defined by the row/column start/end values. This is slower than GetTripletList, because communication is required. Data is returned with absolute coordinates.

Parameters

triplet_list	the list to fill.
start_row	the starting row for data to store on this process.
end_row	the ending row for data to store on this process.
start_column	the starting col for data to store on this process
end_column	the ending col for data to store on this process

◆ GetMatrixSlice()

void NTPoly::Matrix_ps::GetMatrixSlice	(	Matrix_ps &	submatrix,
		int	start_row,
		int	end_row,
		int	start_column,
		int	end_column
	)

Copy an arbitrary slice from a matrix into a new smaller matrix. NTPoly only works with square matrices, so if the number of rows and columns is different the matrix is resized to the maximum size.

Parameters

submatrix	the slice to fill.
start_row	the starting row to include in this matrix.
end_row	the ending row to include in this matrix.
start_column	the starting column to include in this matrix.
end_column	the last column to include in this matrix.

◆ GetTripletList() [1/2]

void NTPoly::Matrix_ps::GetTripletList ( TripletList_c & triplet_list ) const

Extracts a triplet list of the data that is stored on this process. Data is returned with absolute coordinates.

Parameters

triplet_list the list to fill.

◆ GetTripletList() [2/2]

void NTPoly::Matrix_ps::GetTripletList ( TripletList_r & triplet_list ) const

Extracts a triplet list of the data that is stored on this process. Data is returned with absolute coordinates.

Parameters

triplet_list the list to fill.

◆ Increment()

void NTPoly::Matrix_ps::Increment	(	const Matrix_ps &	matB,
		double	alpha = `1.0`,
		double	threshold = `0.0`
	)

this = alpha*MatB + this (AXPY)

Parameters

matB	the matrix to add.
alpha	scaling factor.
threshold	for flushing small values.

◆ PairwiseMultiply()

void NTPoly::Matrix_ps::PairwiseMultiply	(	const Matrix_ps &	matA,
		const Matrix_ps &	matB
	)

Elementwise multiplication. C_ij = A_ij * B_ij. Also known as a Hadamard.

Parameters

matA	first mat.
matB	second mat.

◆ Resize()

void NTPoly::Matrix_ps::Resize ( int new_size )

Change the size of a matrix. If the new size is smaller, then values outside that range are deleted. IF the new size is bigger, zero padding is applied. Warning: this requires a full data redistribution.

◆ Scale()

void NTPoly::Matrix_ps::Scale ( double constant )

scale the matrix by a constatn. constant the value to scale by.

◆ WriteToBinary()

void NTPoly::Matrix_ps::WriteToBinary ( std::string file_name ) const

Write the matrix to a custom binary format.

Parameters

file_name file to write to.

◆ WriteToMatrixMarket()

void NTPoly::Matrix_ps::WriteToMatrixMarket ( std::string file_name ) const

Write the matrix to the matrix market format.

Parameters

file_name file to write to.

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

/Users/wddawson/Desktop/NTPoly/Source/CPlusPlus/PSMatrix.h
/Users/wddawson/Desktop/NTPoly/Source/CPlusPlus/PSMatrix.cc

Public Member Functions

Friends

Detailed Description

Constructor & Destructor Documentation

◆ Matrix_ps() [1/5]

◆ Matrix_ps() [2/5]

◆ Matrix_ps() [3/5]

◆ Matrix_ps() [4/5]

◆ Matrix_ps() [5/5]

Member Function Documentation

◆ Dot()

◆ Dot_c()

◆ FillDense()

◆ FillDistributedPermutation()

◆ FillFromTripletList() [1/2]

◆ FillFromTripletList() [2/2]

◆ Gemm()

◆ GetMatrixBlock() [1/2]

◆ GetMatrixBlock() [2/2]

◆ GetMatrixSlice()

◆ GetTripletList() [1/2]

◆ GetTripletList() [2/2]

◆ Increment()

◆ PairwiseMultiply()

◆ Resize()

◆ Scale()

◆ WriteToBinary()

◆ WriteToMatrixMarket()