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

Data Types
type	distributedsparsematrix_t
	A datatype for a distributed blocked CSR matrix. More...

Functions/Subroutines
pure subroutine, public	constructemptydistributedsparsematrix (this, matrix_dim_)
	Construct an empty sparse, distributed, matrix. More...

pure subroutine, public	destructdistributedsparsematrix (this)
	Destruct a distributed sparse matrix. More...

pure subroutine, public	copydistributedsparsematrix (matA, matB)
	Copy a distributed sparse matrix in a safe way. More...

subroutine, public	constructfrommatrixmarket (this, file_name)
	Construct distributed sparse matrix from a matrix market file in parallel. More...

subroutine, public	constructfrombinary (this, file_name)
	Construct a distributed sparse matrix from a binary file in parallel. More...

subroutine, public	writetobinary (this, file_name)
	Save a distributed sparse matrix to a file. More...

subroutine, public	writetomatrixmarket (this, file_name)
	Write a distributed sparse matrix to a matrix market file. More...

subroutine, public	fillfromtripletlist (this, triplet_list)
	This routine fills in a matrix based on local triplet lists. Each process should pass in triplet lists with global coordinates. It doesn't matter where each triplet is stored, as long as global coordinates are given. More...

pure subroutine, public	filldistributedidentity (this)
	Fill in the values of a distributed matrix with the identity matrix. More...

pure subroutine, public	filldistributedpermutation (this, permutation_vector, permuterows)
	Fill in the values of a distributed matrix with a permutation. If you don't specify permuterows, will default to permuting rows. More...

pure subroutine, public	gettripletlist (this, triplet_list)
	Extracts a triplet list of the data that is stored on this process. Data is returned with absolute coordinates. More...

pure integer function, public	getactualdimension (this)
	Get the actual dimension of the matrix. More...

pure integer function, public	getlogicaldimension (this)
	Get the logical dimension of the matrix. More...

subroutine, public	incrementdistributedsparsematrix (matA, matB, alpha_in, threshold_in)
	Matrix B = alpha*Matrix A + Matrix B (AXPY) This will utilize the sparse vector increment routine. More...

real(ntreal) function, public	dotdistributedsparsematrix (matA, matB)
	product = dot(Matrix A,Matrix B) More...

real(ntreal) function, public	distributedgrandsum (matA)
	Sum up the elements in a matrix. More...

subroutine, public	distributedpairwisemultiply (matA, matB, matC)
	Elementwise multiplication. C_ij = A_ij * B_ij. More...

subroutine, public	distributedgemm (matA, matB, matC, alpha_in, beta_in, threshold_in, memory_pool_in)
	Multiply two matrices together, and add to the third. C := alphamatAmatB+ beta*matC. More...

pure subroutine, public	scaledistributedsparsematrix (this, constant)
	Will scale a distributed sparse matrix by a constant. More...

real(ntreal) function, public	distributedsparsenorm (this)
	Compute the norm of a distributed sparse matrix along the rows. More...

real(ntreal) function, public	trace (this)
	Compute the trace of the matrix. More...

subroutine, public	computesigma (this, sigma_value)
	Compute sigma for the inversion method. More...

subroutine, public	printdistributedsparsematrix (this, file_name_in)
	Print ouf a distributed sparse matrix. More...

subroutine, public	filtersparsematrix (this, threshold)
	A utility routine that filters a sparse matrix. All (absolute) values below the threshold are set to zero. More...

integer(c_long) function, public	getsize (this)
	Get the total number of non zero entries in the distributed sparse matrix. More...

subroutine, public	getloadbalance (this, min_size, max_size)
	Get a measure of how load balanced this matrix is. For each process, the number of non-zero entries is calculated. Then, this function returns the max and min of those values. More...

Detailed Description

A Module For Performing Distributed Sparse Matrix Operations.

Function/Subroutine Documentation

◆ computesigma()

subroutine, public distributedsparsematrixmodule::computesigma	(	type(distributedsparsematrix_t), intent(in)	this,
		real(ntreal), intent(out)	sigma_value
	)

Compute sigma for the inversion method.

Todo:: describe this better.

Parameters

[in]	this	the matrix to compute the sigma value of.
[out]	sigma_value	sigma.

◆ constructemptydistributedsparsematrix()

pure subroutine, public distributedsparsematrixmodule::constructemptydistributedsparsematrix	(	type(distributedsparsematrix_t), intent(inout)	this,
		integer, intent(in)	matrix_dim_
	)

Construct an empty sparse, distributed, matrix.

Parameters

[out]	this	the matrix to be constructed.
[in]	matrix_dim_	the dimension of the full matrix.

◆ constructfrombinary()

subroutine, public distributedsparsematrixmodule::constructfrombinary	(	type(distributedsparsematrix_t)	this,
		character(len=*), intent(in)	file_name
	)

Construct a distributed sparse matrix from a binary file in parallel.

Parameters

[out]	this	the file being constructed.
[in]	file_name	name of the file to read.

◆ constructfrommatrixmarket()

subroutine, public distributedsparsematrixmodule::constructfrommatrixmarket	(	type(distributedsparsematrix_t)	this,
		character(len=*), intent(in)	file_name
	)

Construct distributed sparse matrix from a matrix market file in parallel.

Parameters

[out]	this	the file being constructed.
[in]	file_name	name of the file to read.

◆ copydistributedsparsematrix()

pure subroutine, public distributedsparsematrixmodule::copydistributedsparsematrix	(	type(distributedsparsematrix_t), intent(in)	matA,
		type(distributedsparsematrix_t), intent(inout)	matB
	)

Copy a distributed sparse matrix in a safe way.

Parameters

[in]	matA	matrix to copy
[in,out]	matB	= matA

◆ destructdistributedsparsematrix()

pure subroutine, public distributedsparsematrixmodule::destructdistributedsparsematrix ( type(distributedsparsematrix_t), intent(inout) this )

Destruct a distributed sparse matrix.

Parameters

[in,out] this the matrix to destruct

◆ distributedgemm()

subroutine, public distributedsparsematrixmodule::distributedgemm	(	type(distributedsparsematrix_t), intent(in)	matA,
		type(distributedsparsematrix_t), intent(in)	matB,
		type(distributedsparsematrix_t), intent(inout)	matC,
		real(ntreal), intent(in), optional	alpha_in,
		real(ntreal), intent(in), optional	beta_in,
		real(ntreal), intent(in), optional	threshold_in,
		type(distributedmatrixmemorypool_t), intent(inout), optional	memory_pool_in
	)

Multiply two matrices together, and add to the third. C := alpha*matA*matB+ beta*matC.

Parameters

[in]	matA	Matrix A
[in]	matB	Matrix B
[out]	matC	= alphamatAmatB + beta*matC
[in]	alpha_in	scales the multiplication
[in]	beta_in	scales matrix we sum on to
[in]	threshold_in	for flushing values to zero. Default value is 0.0.
[in,out]	memory_pool_in	a memory pool that can be used for the calculation.

◆ distributedgrandsum()

real(ntreal) function, public distributedsparsematrixmodule::distributedgrandsum ( type(distributedsparsematrix_t), intent(in) matA )

Sum up the elements in a matrix.

Parameters

[in] matA Matrix A.

Returns: sum the sum of all elements.

◆ distributedpairwisemultiply()

subroutine, public distributedsparsematrixmodule::distributedpairwisemultiply	(	type(distributedsparsematrix_t), intent(in)	matA,
		type(distributedsparsematrix_t), intent(in)	matB,
		type(distributedsparsematrix_t), intent(inout)	matC
	)

Elementwise multiplication. C_ij = A_ij * B_ij.

Parameters

[in]	matA	Matrix A.
[in]	matB	Matrix B.
[in,out]	matC	= MatA mult MatB.

◆ distributedsparsenorm()

real(ntreal) function, public distributedsparsematrixmodule::distributedsparsenorm ( type(distributedsparsematrix_t), intent(in) this )

Compute the norm of a distributed sparse matrix along the rows.

Parameters

[in] this the matrix to compute the norm of.

Returns: the norm value of the full distributed sparse matrix.

◆ dotdistributedsparsematrix()

real(ntreal) function, public distributedsparsematrixmodule::dotdistributedsparsematrix	(	type(distributedsparsematrix_t), intent(in)	matA,
		type(distributedsparsematrix_t), intent(in)	matB
	)

product = dot(Matrix A,Matrix B)

Parameters

[in]	matA	Matrix A.
[in,out]	matB	Matrix B.

Returns: product the dot product.

◆ filldistributedidentity()

pure subroutine, public distributedsparsematrixmodule::filldistributedidentity ( type(distributedsparsematrix_t), intent(inout) this )

Fill in the values of a distributed matrix with the identity matrix.

Parameters

[in,out] this the matrix being filled.

◆ filldistributedpermutation()

pure subroutine, public distributedsparsematrixmodule::filldistributedpermutation	(	type(distributedsparsematrix_t), intent(inout)	this,
		integer, dimension(:), intent(in)	permutation_vector,
		logical, intent(in), optional	permuterows
	)

Fill in the values of a distributed matrix with a permutation. If you don't specify permuterows, will default to permuting rows.

Parameters

[in,out]	this	the matrix being filled.
[in]	permutation_vector	describes for each row/column, where it goes.
[in]	permuterows	if true permute rows, false permute columns.

◆ fillfromtripletlist()

subroutine, public distributedsparsematrixmodule::fillfromtripletlist	(	type(distributedsparsematrix_t)	this,
		type(tripletlist_t)	triplet_list
	)

This routine fills in a matrix based on local triplet lists. Each process should pass in triplet lists with global coordinates. It doesn't matter where each triplet is stored, as long as global coordinates are given.

Parameters

[in,out]	this	the matrix to fill.
[in]	triplet_list	the triplet list of values.

◆ filtersparsematrix()

subroutine, public distributedsparsematrixmodule::filtersparsematrix	(	type(distributedsparsematrix_t), intent(inout)	this,
		real(ntreal), intent(in)	threshold
	)

A utility routine that filters a sparse matrix. All (absolute) values below the threshold are set to zero.

Parameters

[in,out]	this	matrix to filter
[in]	threshold	(absolute) values below this are filtered

◆ getactualdimension()

pure integer function, public distributedsparsematrixmodule::getactualdimension ( type(distributedsparsematrix_t), intent(in) this )

Get the actual dimension of the matrix.

Parameters

[in] this the matrix.

Returns: dimension of the matrix;

◆ getloadbalance()

subroutine, public distributedsparsematrixmodule::getloadbalance	(	type(distributedsparsematrix_t), intent(in)	this,
		integer, intent(out)	min_size,
		integer, intent(out)	max_size
	)

Get a measure of how load balanced this matrix is. For each process, the number of non-zero entries is calculated. Then, this function returns the max and min of those values.

Parameters

[in]	this	The matrix to compute the measure on.
[out]	min_size	the minimum entries contained on a single process.
[out]	max_size	the maximum entries contained on a single process.

◆ getlogicaldimension()

pure integer function, public distributedsparsematrixmodule::getlogicaldimension ( type(distributedsparsematrix_t), intent(in) this )

Get the logical dimension of the matrix.

Parameters

[in] this the matrix.

Returns: dimension of the matrix;

◆ getsize()

integer(c_long) function, public distributedsparsematrixmodule::getsize ( type(distributedsparsematrix_t), intent(in) this )

Get the total number of non zero entries in the distributed sparse matrix.

Parameters

[in] this the distributed sparse matrix to calculate the non-zero entries of.

Returns: the number of non-zero entries in the matrix.

◆ gettripletlist()

pure subroutine, public distributedsparsematrixmodule::gettripletlist	(	type(distributedsparsematrix_t), intent(in)	this,
		type(tripletlist_t), intent(inout)	triplet_list
	)

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

Parameters

[in]	this	the distributed sparse matrix.
[in,out]	triplet_list	the list to fill.

◆ incrementdistributedsparsematrix()

subroutine, public distributedsparsematrixmodule::incrementdistributedsparsematrix	(	type(distributedsparsematrix_t), intent(in)	matA,
		type(distributedsparsematrix_t), intent(inout)	matB,
		real(ntreal), intent(in), optional	alpha_in,
		real(ntreal), intent(in), optional	threshold_in
	)

Matrix B = alpha*Matrix A + Matrix B (AXPY) This will utilize the sparse vector increment routine.

Parameters

[in]	matA	Matrix A.
[in,out]	matB	Matrix B.
[in]	alpha_in	multiplier. Default value is 1.0
[in]	threshold_in	for flushing values to zero. Default value is 0.0

◆ printdistributedsparsematrix()

subroutine, public distributedsparsematrixmodule::printdistributedsparsematrix	(	type(distributedsparsematrix_t)	this,
		character(len=*), intent(in), optional	file_name_in
	)

Print ouf a distributed sparse matrix.

Parameters

[in]	this	the matrix to print.
[in]	file_name_in	optionally, you can pass a file to print to.

◆ scaledistributedsparsematrix()

pure subroutine, public distributedsparsematrixmodule::scaledistributedsparsematrix	(	type(distributedsparsematrix_t), intent(inout)	this,
		real(ntreal), intent(in)	constant
	)

Will scale a distributed sparse matrix by a constant.

Parameters

[in,out]	this	Matrix to scale.
[in]	constant	scale factor.

◆ trace()

real(ntreal) function, public distributedsparsematrixmodule::trace ( type(distributedsparsematrix_t), intent(in) this )

Compute the trace of the matrix.

Parameters

[in] this the matrix to compute the norm of.

Returns: the trace value of the full distributed sparse matrix.

◆ writetobinary()

subroutine, public distributedsparsematrixmodule::writetobinary	(	type(distributedsparsematrix_t)	this,
		character(len=*), intent(in)	file_name
	)

Save a distributed sparse matrix to a file.

Parameters

[in]	this	the Matrix to write.
[in]	file_name	name of the file to write to.

◆ writetomatrixmarket()

subroutine, public distributedsparsematrixmodule::writetomatrixmarket	(	type(distributedsparsematrix_t)	this,
		character(len=*), intent(in)	file_name
	)

Write a distributed sparse matrix to a matrix market file.

Parameters

[in]	this	the Matrix to write.
[in]	file_name	name of the file to write to.

Data Types

Functions/Subroutines

Detailed Description

Function/Subroutine Documentation

◆ computesigma()

◆ constructemptydistributedsparsematrix()

◆ constructfrombinary()

◆ constructfrommatrixmarket()

◆ copydistributedsparsematrix()

◆ destructdistributedsparsematrix()

◆ distributedgemm()

◆ distributedgrandsum()

◆ distributedpairwisemultiply()

◆ distributedsparsenorm()

◆ dotdistributedsparsematrix()

◆ filldistributedidentity()

◆ filldistributedpermutation()

◆ fillfromtripletlist()

◆ filtersparsematrix()

◆ getactualdimension()

◆ getloadbalance()

◆ getlogicaldimension()

◆ getsize()

◆ gettripletlist()

◆ incrementdistributedsparsematrix()

◆ printdistributedsparsematrix()

◆ scaledistributedsparsematrix()

◆ trace()

◆ writetobinary()

◆ writetomatrixmarket()