Represents a complex full matrix distributed on many processors. More...

Data Types
type	copy_cfm_info_type
	Stores the state of a copy between cp_cfm_start_copy_general and cp_cfm_finish_copy_general. More...

type	cp_cfm_p_type
	Just to build arrays of pointers to matrices. More...

interface	cp_cfm_to_cfm

type	cp_cfm_type
	Represent a complex full matrix. More...

Functions/Subroutines
subroutine, public	cp_cfm_create (matrix, matrix_struct, name)
	Creates a new full matrix with the given structure.

subroutine, public	cp_cfm_release (matrix)
	Releases a full matrix.

subroutine, public	cp_cfm_set_all (matrix, alpha, beta)
	Set all elements of the full matrix to alpha. Besides, set all diagonal matrix elements to beta (if given explicitly).

subroutine, public	cp_cfm_get_element (matrix, irow_global, icol_global, alpha)
	Get the matrix element by its global index.

subroutine, public	cp_cfm_set_element (matrix, irow_global, icol_global, alpha)
	Set the matrix element (irow_global,icol_global) of the full matrix to alpha.

subroutine, public	cp_cfm_get_submatrix (fm, target_m, start_row, start_col, n_rows, n_cols, transpose)
	Extract a sub-matrix from the full matrix: op(target_m)(1:n_rows,1:n_cols) = fm(start_row:start_row+n_rows,start_col:start_col+n_cols). Sub-matrix 'target_m' is replicated on each CPU. Using this call is expensive.

subroutine, public	cp_cfm_set_submatrix (matrix, new_values, start_row, start_col, n_rows, n_cols, alpha, beta, transpose)
	Set a sub-matrix of the full matrix: matrix(start_row:start_row+n_rows,start_col:start_col+n_cols) = alphaop(new_values)(1:n_rows,1:n_cols) + betamatrix(start_row:start_row+n_rows,start_col:start_col+n_cols)

subroutine, public	cp_cfm_get_info (matrix, name, nrow_global, ncol_global, nrow_block, ncol_block, nrow_local, ncol_local, row_indices, col_indices, local_data, context, matrix_struct, para_env)
	Returns information about a full matrix.

subroutine, public	cp_cfm_to_fm (msource, mtargetr, mtargeti)
	Copy real and imaginary parts of a complex full matrix into separate real-value full matrices.

subroutine, public	cp_fm_to_cfm (msourcer, msourcei, mtarget)
	Construct a complex full matrix by taking its real and imaginary parts from two separate real-value full matrices.

subroutine, public	cp_cfm_start_copy_general (source, destination, para_env, info)
	Initiate the copy operation: get distribution data, post MPI isend and irecvs.

subroutine, public	cp_cfm_finish_copy_general (destination, info)
	Complete the copy operation: wait for comms, unpack, clean up MPI state.

subroutine, public	cp_cfm_cleanup_copy_general (info)
	Complete the copy operation: wait for comms clean up MPI state.

Detailed Description

Represents a complex full matrix distributed on many processors.

Author: Joost VandeVondele, based on Fawzi's cp_fm_* routines

Function/Subroutine Documentation

◆ cp_cfm_create()

subroutine, public cp_cfm_types::cp_cfm_create	(	type(cp_cfm_type), intent(out)	matrix,
		type(cp_fm_struct_type), intent(in), target	matrix_struct,
		character(len=*), intent(in), optional	name
	)

Creates a new full matrix with the given structure.

Parameters

matrix	matrix to be created
matrix_struct	structure of the matrix
name	name of the matrix

Note: preferred allocation routine

Definition at line 120 of file cp_cfm_types.F.

Here is the call graph for this function:

◆ cp_cfm_release()

subroutine, public cp_cfm_types::cp_cfm_release ( type(cp_cfm_type), intent(inout) matrix )

Releases a full matrix.

Parameters

matrix the matrix to release

Definition at line 154 of file cp_cfm_types.F.

Here is the call graph for this function:

◆ cp_cfm_set_all()

subroutine, public cp_cfm_types::cp_cfm_set_all	(	type(cp_cfm_type), intent(in)	matrix,
		complex(kind=dp), intent(in)	alpha,
		complex(kind=dp), intent(in), optional	beta
	)

Set all elements of the full matrix to alpha. Besides, set all diagonal matrix elements to beta (if given explicitly).

Parameters

matrix	matrix to initialise
alpha	value of off-diagonal matrix elements
beta	value of diagonal matrix elements (equal to alpha if absent)

Date: 12.06.2001

Author: Matthias Krack

Version: 1.0

Definition at line 174 of file cp_cfm_types.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ cp_cfm_get_element()

subroutine, public cp_cfm_types::cp_cfm_get_element	(	type(cp_cfm_type), intent(in)	matrix,
		integer, intent(in)	irow_global,
		integer, intent(in)	icol_global,
		complex(kind=dp), intent(out)	alpha
	)

Get the matrix element by its global index.

Parameters

matrix	full matrix
irow_global	global row index
icol_global	global column index
alpha	matrix element

History: , TCH, created always return the answer

Definition at line 227 of file cp_cfm_types.F.

Here is the caller graph for this function:

◆ cp_cfm_set_element()

subroutine, public cp_cfm_types::cp_cfm_set_element	(	type(cp_cfm_type), intent(in)	matrix,
		integer, intent(in)	irow_global,
		integer, intent(in)	icol_global,
		complex(kind=dp), intent(in)	alpha
	)

Set the matrix element (irow_global,icol_global) of the full matrix to alpha.

Parameters

matrix	full matrix
irow_global	global row index
icol_global	global column index
alpha	value of the matrix element

Date: 12.06.2001

Author: Matthias Krack

Version: 1.0

Definition at line 274 of file cp_cfm_types.F.

Here is the caller graph for this function:

◆ cp_cfm_get_submatrix()

subroutine, public cp_cfm_types::cp_cfm_get_submatrix	(	type(cp_cfm_type), intent(in)	fm,
		complex(kind=dp), dimension(:, :), intent(out)	target_m,
		integer, intent(in), optional	start_row,
		integer, intent(in), optional	start_col,
		integer, intent(in), optional	n_rows,
		integer, intent(in), optional	n_cols,
		logical, intent(in), optional	transpose
	)

Extract a sub-matrix from the full matrix: op(target_m)(1:n_rows,1:n_cols) = fm(start_row:start_row+n_rows,start_col:start_col+n_cols). Sub-matrix 'target_m' is replicated on each CPU. Using this call is expensive.

Parameters

fm	full matrix you want to get the elements from
target_m	2-D array to store the extracted sub-matrix
start_row	global row index of the matrix element target_m(1,1) (defaults to 1)
start_col	global column index of the matrix element target_m(1,1) (defaults to 1)
n_rows	number of rows to extract (defaults to size(op(target_m),1))
n_cols	number of columns to extract (defaults to size(op(target_m),2))
transpose	indicates that the extracted sub-matrix target_m should be transposed: op(target_m) = target_m^T if .TRUE., op(target_m) = target_m if .FALSE. (defaults to false)

History

04.2016 created borrowing from Fawzi's cp_fm_get_submatrix [Lianheng Tong]
01.2018 drop innermost conditional branching [Sergey Chulkov]

Author: Lianheng Tong

Note: Optimized for full column updates. The matrix target_m is replicated and valid on all CPUs.

Definition at line 328 of file cp_cfm_types.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ cp_cfm_set_submatrix()

subroutine, public cp_cfm_types::cp_cfm_set_submatrix	(	type(cp_cfm_type), intent(in)	matrix,
		complex(kind=dp), dimension(:, :), intent(in)	new_values,
		integer, intent(in), optional	start_row,
		integer, intent(in), optional	start_col,
		integer, intent(in), optional	n_rows,
		integer, intent(in), optional	n_cols,
		complex(kind=dp), intent(in), optional	alpha,
		complex(kind=dp), intent(in), optional	beta,
		logical, intent(in), optional	transpose
	)

Set a sub-matrix of the full matrix: matrix(start_row:start_row+n_rows,start_col:start_col+n_cols) = alpha*op(new_values)(1:n_rows,1:n_cols) + beta*matrix(start_row:start_row+n_rows,start_col:start_col+n_cols)

Parameters

matrix	full to update
new_values	replicated 2-D array that holds new elements of the updated sub-matrix
start_row	global row index of the matrix element new_values(1,1) (defaults to 1)
start_col	global column index of the matrix element new_values(1,1) (defaults to 1)
n_rows	number of rows to update (defaults to size(op(new_values),1))
n_cols	number of columns to update (defaults to size(op(new_values),2))
alpha	scale factor for the new values (defaults to (1.0,0.0))
beta	scale factor for the old values (defaults to (0.0,0.0))
transpose	indicates that the matrix new_values should be transposed: op(new_values) = new_values^T if .TRUE., op(new_values) = new_values if .FALSE. (defaults to false)

History

04.2016 created borrowing from Fawzi's cp_fm_set_submatrix [Lianheng Tong]
01.2018 drop innermost conditional branching [Sergey Chulkov]

Author: Lianheng Tong

Note: Optimized for alpha=(1.0,0.0), beta=(0.0,0.0) All matrix elements 'new_values' need to be valid on all CPUs

Definition at line 464 of file cp_cfm_types.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ cp_cfm_get_info()

subroutine, public cp_cfm_types::cp_cfm_get_info	(	type(cp_cfm_type), intent(in)	matrix,
		character(len=*), intent(out), optional	name,
		integer, intent(out), optional	nrow_global,
		integer, intent(out), optional	ncol_global,
		integer, intent(out), optional	nrow_block,
		integer, intent(out), optional	ncol_block,
		integer, intent(out), optional	nrow_local,
		integer, intent(out), optional	ncol_local,
		integer, dimension(:), optional, pointer	row_indices,
		integer, dimension(:), optional, pointer	col_indices,
		complex(kind=dp), dimension(:, :), optional, pointer, contiguous	local_data,
		type(cp_blacs_env_type), optional, pointer	context,
		type(cp_fm_struct_type), optional, pointer	matrix_struct,
		type(mp_para_env_type), optional, pointer	para_env
	)

Returns information about a full matrix.

Parameters

matrix	matrix
name	name of the matrix
nrow_global	total number of rows
ncol_global	total number of columns
nrow_block	number of rows per ScaLAPACK block
ncol_block	number of columns per ScaLAPACK block
nrow_local	number of locally stored rows
ncol_local	number of locally stored columns
row_indices	global indices of locally stored rows
col_indices	global indices of locally stored columns
local_data	locally stored matrix elements
context	BLACS context
matrix_struct	matrix structure
para_env	parallel environment

Date: 12.06.2001

Author: Matthias Krack

Version: 1.0

Definition at line 599 of file cp_cfm_types.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ cp_cfm_to_fm()

subroutine, public cp_cfm_types::cp_cfm_to_fm	(	type(cp_cfm_type), intent(in)	msource,
		type(cp_fm_type), intent(in), optional	mtargetr,
		type(cp_fm_type), intent(in), optional	mtargeti
	)

Copy real and imaginary parts of a complex full matrix into separate real-value full matrices.

Parameters

msource	complex matrix
mtargetr	(optional) real part of the source matrix
mtargeti	(optional) imaginary part of the source matrix

Note: Matrix structures are assumed to be equivalent.

Definition at line 760 of file cp_cfm_types.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ cp_fm_to_cfm()

subroutine, public cp_cfm_types::cp_fm_to_cfm	(	type(cp_fm_type), intent(in), optional	msourcer,
		type(cp_fm_type), intent(in), optional	msourcei,
		type(cp_cfm_type), intent(in)	mtarget
	)

Construct a complex full matrix by taking its real and imaginary parts from two separate real-value full matrices.

Parameters

msourcer	(optional) real part of the complex matrix (defaults to 0.0)
msourcei	(optional) imaginary part of the complex matrix (defaults to 0.0)
mtarget	resulting complex matrix

Note: Matrix structures are assumed to be equivalent.

Definition at line 812 of file cp_cfm_types.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ cp_cfm_start_copy_general()

subroutine, public cp_cfm_types::cp_cfm_start_copy_general	(	type(cp_cfm_type), intent(in)	source,
		type(cp_cfm_type), intent(in)	destination,
		type(mp_para_env_type), intent(in), pointer	para_env,
		type(copy_cfm_info_type), intent(out)	info
	)

Initiate the copy operation: get distribution data, post MPI isend and irecvs.

Parameters

source	input complex-valued fm matrix
destination	output complex-valued fm matrix
para_env	parallel environment corresponding to the BLACS env that covers all parts of the input and output matrices
info	all of the data that will be needed to complete the copy operation

Note: a slightly modified version of the subroutine cp_fm_start_copy_general() that uses allocatable arrays instead of pointers wherever possible.

Definition at line 874 of file cp_cfm_types.F.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ cp_cfm_finish_copy_general()

subroutine, public cp_cfm_types::cp_cfm_finish_copy_general	(	type(cp_cfm_type), intent(in)	destination,
		type(copy_cfm_info_type), intent(inout)	info
	)

Complete the copy operation: wait for comms, unpack, clean up MPI state.

Parameters

destination	output cfm matrix
info	all of the data that will be needed to complete the copy operation

Note: a slightly modified version of the subroutine cp_fm_finish_copy_general() that uses allocatable arrays instead of pointers wherever possible.

Definition at line 1186 of file cp_cfm_types.F.

Here is the caller graph for this function:

◆ cp_cfm_cleanup_copy_general()

subroutine, public cp_cfm_types::cp_cfm_cleanup_copy_general ( type(copy_cfm_info_type), intent(inout) info )

Complete the copy operation: wait for comms clean up MPI state.

Parameters

info	all of the data that will be needed to complete the copy operation

Note: a slightly modified version of the subroutine cp_fm_cleanup_copy_general() that uses allocatable arrays instead of pointers wherever possible.

Definition at line 1253 of file cp_cfm_types.F.

Here is the caller graph for this function:

Data Types

Functions/Subroutines

Detailed Description

Function/Subroutine Documentation

◆ cp_cfm_create()

◆ cp_cfm_release()

◆ cp_cfm_set_all()

◆ cp_cfm_get_element()

◆ cp_cfm_set_element()

◆ cp_cfm_get_submatrix()

◆ cp_cfm_set_submatrix()

◆ cp_cfm_get_info()

◆ cp_cfm_to_fm()

◆ cp_fm_to_cfm()

◆ cp_cfm_start_copy_general()

◆ cp_cfm_finish_copy_general()

◆ cp_cfm_cleanup_copy_general()