(git:34ef472)
Data Types | Functions/Subroutines
cp_fm_types Module Reference

represent a full matrix distributed on many processors More...

Functions/Subroutines

subroutine, public cp_fm_create (matrix, matrix_struct, name, use_sp)
 creates a new full matrix with the given structure More...
 
subroutine, public cp_fm_init_random (matrix, ncol, start_col)
 fills a matrix with random numbers More...
 
subroutine, public cp_fm_set_all (matrix, alpha, beta)
 set all elements of a matrix to the same value, and optionally the diagonal to a different one More...
 
subroutine, public cp_fm_get_diag (matrix, diag)
 returns the diagonal elements of a fm More...
 
subroutine, public cp_fm_get_element (matrix, irow_global, icol_global, alpha, local)
 returns an element of a fm this value is valid on every cpu using this call is expensive More...
 
subroutine, public cp_fm_set_element (matrix, irow_global, icol_global, alpha)
 sets an element of a matrix More...
 
subroutine, public cp_fm_set_submatrix (fm, new_values, start_row, start_col, n_rows, n_cols, alpha, beta, transpose)
 sets a submatrix of a full matrix fm(start_row:start_row+n_rows,start_col:start_col+n_cols) = alpha*op(new_values)(1:n_rows,1:n_cols)+ beta More...
 
subroutine, public cp_fm_get_submatrix (fm, target_m, start_row, start_col, n_rows, n_cols, transpose)
 gets a submatrix of a full matrix op(target_m)(1:n_rows,1:n_cols) =fm(start_row:start_row+n_rows,start_col:start_col+n_cols) target_m is replicated on all cpus using this call is expensive More...
 
subroutine, public cp_fm_get_info (matrix, name, nrow_global, ncol_global, nrow_block, ncol_block, nrow_local, ncol_local, row_indices, col_indices, local_data, context, nrow_locals, ncol_locals, matrix_struct, para_env)
 returns all kind of information about the full matrix More...
 
subroutine, public cp_fm_write_info (matrix, io_unit)
 Write nicely formatted info about the FM to the given I/O unit (including the underlying FM struct) More...
 
subroutine, public cp_fm_maxabsval (matrix, a_max, ir_max, ic_max)
 find the maximum absolute value of the matrix element maxval(abs(matrix)) More...
 
subroutine, public cp_fm_maxabsrownorm (matrix, a_max)
 find the maximum over the rows of the sum of the absolute values of the elements of a given row = || A ||_infinity More...
 
subroutine, public cp_fm_vectorsnorm (matrix, norm_array)
 find the inorm of each column norm_{j}= sqrt( \sum_{i} A_{ij}*A_{ij} ) More...
 
subroutine, public cp_fm_vectorssum (matrix, sum_array, dir)
 summing up all the elements along the matrix's i-th index \( \mathrm{sum}_{j} = \sum_{i} A_{ij} \) or \( \mathrm{sum}_{i} = \sum_{j} A_{ij} \) More...
 
subroutine, public cp_fm_to_fm_triangular (msource, mtarget, uplo)
 copy just a triangular matrix More...
 
subroutine, public cp_fm_to_fm_submat (msource, mtarget, nrow, ncol, s_firstrow, s_firstcol, t_firstrow, t_firstcol)
 copy just a part ot the matrix More...
 
subroutine, public cp_fm_copy_general (source, destination, para_env)
 General copy of a fm matrix to another fm matrix. Uses non-blocking MPI rather than ScaLAPACK. More...
 
subroutine, public cp_fm_start_copy_general (source, destination, para_env, info)
 Initiates the copy operation: get distribution data, post MPI isend and irecvs. More...
 
subroutine, public cp_fm_finish_copy_general (destination, info)
 Completes the copy operation: wait for comms, unpack, clean up MPI state. More...
 
subroutine, public cp_fm_cleanup_copy_general (info)
 Completes the copy operation: wait for comms clean up MPI state. More...
 
subroutine, public cp_fm_to_fm_submat_general (source, destination, nrows, ncols, s_firstrow, s_firstcol, d_firstrow, d_firstcol, global_context)
 General copy of a submatrix of fm matrix to a submatrix of another fm matrix. The two matrices can have different contexts. More...
 
subroutine, public cp_fm_add_to_element (matrix, irow_global, icol_global, alpha)
 ... More...
 
subroutine, public cp_fm_write_unformatted (fm, unit)
 ... More...
 
subroutine, public cp_fm_write_formatted (fm, unit, header, value_format)
 Write out a full matrix in plain text. More...
 
subroutine, public cp_fm_read_unformatted (fm, unit)
 ... More...
 
integer function, public cp_fm_indxg2p (INDXGLOB, NB, IPROC, ISRCPROC, NPROCS)
 wrapper to scalapack function INDXG2P that computes the process coordinate which possesses the entry of a distributed matrix specified by a global index INDXGLOB. More...
 
integer function, public cp_fm_indxg2l (INDXGLOB, NB, IPROC, ISRCPROC, NPROCS)
 wrapper to scalapack function INDXG2L that computes the local index of a distributed matrix entry pointed to by the global index INDXGLOB. More...
 
integer function, public cp_fm_indxl2g (INDXLOC, NB, IPROC, ISRCPROC, NPROCS)
 wrapper to scalapack function INDXL2G that computes the global index of a distributed matrix entry pointed to by the local index INDXLOC of the process indicated by IPROC. More...
 
subroutine, public cp_fm_setup (mult_type)
 ... More...
 
integer function, public cp_fm_get_mm_type ()
 ... More...
 
integer function, public cp_fm_pilaenv (ictxt, prec)
 ... More...
 

Detailed Description

represent a full matrix distributed on many processors

History
3) separated structure object, removed globenv, renamed to full matrix many changes (fawzi 08.2002)
Author
Matthias Krack (22.05.2001)

Function/Subroutine Documentation

◆ cp_fm_create()

subroutine, public cp_fm_types::cp_fm_create ( type(cp_fm_type), intent(out)  matrix,
type(cp_fm_struct_type), intent(in), target  matrix_struct,
character(len=*), intent(in), optional  name,
logical, intent(in), optional  use_sp 
)

creates a new full matrix with the given structure

Parameters
matrixthe matrix to be created
matrix_structthe structure of matrix
name...
use_sp...
History
08.2002 created [fawzi]
Author
Fawzi Mohamed
Note
preferred allocation routine

Definition at line 166 of file cp_fm_types.F.

Here is the call graph for this function:

◆ cp_fm_init_random()

subroutine, public cp_fm_types::cp_fm_init_random ( type(cp_fm_type), intent(in)  matrix,
integer, intent(in), optional  ncol,
integer, intent(in), optional  start_col 
)

fills a matrix with random numbers

Parameters
matrix: to be initialized
ncol: numbers of cols to fill
start_col: starting at coll number
Author
Joost VandeVondele
Note
the value of a_ij is independent of the number of cpus

Definition at line 451 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_set_all()

subroutine, public cp_fm_types::cp_fm_set_all ( type(cp_fm_type), intent(in)  matrix,
real(kind=dp), intent(in)  alpha,
real(kind=dp), intent(in), optional  beta 
)

set all elements of a matrix to the same value, and optionally the diagonal to a different one

Parameters
matrixinput matrix
alphascalar used to set all elements of the matrix
betascalar used to set diagonal of the matrix
Note
can be used to zero a matrix can be used to create a unit matrix (I-matrix) alpha=0.0_dp beta=1.0_dp

Definition at line 534 of file cp_fm_types.F.

Here is the call graph for this function:

◆ cp_fm_get_diag()

subroutine, public cp_fm_types::cp_fm_get_diag ( type(cp_fm_type), intent(in)  matrix,
real(kind=dp), dimension(:), intent(out)  diag 
)

returns the diagonal elements of a fm

Parameters
matrix...
diag...

Definition at line 569 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_get_element()

subroutine, public cp_fm_types::cp_fm_get_element ( type(cp_fm_type), intent(in)  matrix,
integer, intent(in)  irow_global,
integer, intent(in)  icol_global,
real(kind=dp), intent(out)  alpha,
logical, intent(out), optional  local 
)

returns an element of a fm this value is valid on every cpu using this call is expensive

Parameters
matrixthe matrix to read
irow_globalthe row
icol_globalthe col
alphathe value of matrix(irow_global, icol_global)
localtrue if the element is on this cpu, false otherwise
Note
  • modified semantics. now this function always returns the value previously the value was zero on cpus that didn't own the relevant part of the matrix (Joost VandeVondele, May 2003)
  • usage of the function should be avoided, as it is likely to rather slow using row_indices/col_indices/local_data + some smart scheme normally yields a real parallel code

Definition at line 642 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_set_element()

subroutine, public cp_fm_types::cp_fm_set_element ( type(cp_fm_type), intent(in)  matrix,
integer, intent(in)  irow_global,
integer, intent(in)  icol_global,
real(kind=dp), intent(in)  alpha 
)

sets an element of a matrix

Parameters
matrix...
irow_global...
icol_global...
alpha...
Note
we expect all cpus to have the same arguments in the call to this function (otherwise one should use local_data tricks)

Definition at line 699 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_set_submatrix()

subroutine, public cp_fm_types::cp_fm_set_submatrix ( type(cp_fm_type), intent(in)  fm,
real(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,
real(kind=dp), intent(in), optional  alpha,
real(kind=dp), intent(in), optional  beta,
logical, intent(in), optional  transpose 
)

sets a submatrix of a full matrix fm(start_row:start_row+n_rows,start_col:start_col+n_cols) = alpha*op(new_values)(1:n_rows,1:n_cols)+ beta

  • fm(start_row:start_row+n_rows,start_col:start_col+n_cols)
    Parameters
    fmthe full to change
    new_valuesa replicated full matrix with the new values
    start_rowthe starting row of b_matrix (defaults to 1)
    start_colthe starting col of b_matrix (defaults to 1)
    n_rowsthe number of row to change in b (defaults to size(op(new_values),1))
    n_colsthe number of columns to change in b (defaults to size(op(new_values),2))
    alpharescaling factor for the new values (defaults to 1.0)
    betarescaling factor for the old values (defaults to 0.0)
    transposeif new_values should be transposed: if true op(new_values)=new_values^T, else op(new_values)=new_values (defaults to false)
    History
    07.2002 created borrowing from Joost's blacs_replicated_copy [fawzi]
    Author
    Fawzi Mohamed
    Note
    optimized for full column updates and alpha=1.0, beta=0.0 the new_values need to be valid on all cpus

Definition at line 766 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_get_submatrix()

subroutine, public cp_fm_types::cp_fm_get_submatrix ( type(cp_fm_type), intent(in)  fm,
real(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 
)

gets a submatrix of a full matrix op(target_m)(1:n_rows,1:n_cols) =fm(start_row:start_row+n_rows,start_col:start_col+n_cols) target_m is replicated on all cpus using this call is expensive

Parameters
fmthe full you want to get the info from
target_ma replicated full matrix that will contain the result
start_rowthe starting row of b_matrix (defaults to 1)
start_colthe starting col of b_matrix (defaults to 1)
n_rowsthe number of row to change in b (defaults to size(op(new_values),1))
n_colsthe number of columns to change in b (defaults to size(op(new_values),2))
transposeif target_m should be transposed: if true op(target_m)=target_m^T, else op(target_m)=target_m (defaults to false)
History
07.2002 created borrowing from Joost's blacs_replicated_copy [fawzi]
Author
Fawzi Mohamed
Note
optimized for full column updates. Zeros out a little too much of target_m the target_m is replicated and valid on all cpus

Definition at line 899 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_get_info()

subroutine, public cp_fm_types::cp_fm_get_info ( type(cp_fm_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,
real(kind=dp), dimension(:, :), optional, pointer, contiguous  local_data,
type(cp_blacs_env_type), optional, pointer  context,
integer, dimension(:), optional, pointer  nrow_locals,
integer, dimension(:), optional, pointer  ncol_locals,
type(cp_fm_struct_type), optional, pointer  matrix_struct,
type(mp_para_env_type), optional, pointer  para_env 
)

returns all kind of information about the full matrix

Parameters
matrix...
name...
nrow_global...
ncol_global...
nrow_block...
ncol_block...
nrow_local...
ncol_local...
row_indices...
col_indices...
local_data...
context...
nrow_locals...
ncol_locals...
matrix_struct...
para_env...
Note
see also cp_fm_struct for explanation
  • nrow_local, ncol_local, row_indices, col_indices, local_data are hooks for efficient access to the local blacs block

Definition at line 1012 of file cp_fm_types.F.

Here is the call graph for this function:

◆ cp_fm_write_info()

subroutine, public cp_fm_types::cp_fm_write_info ( type(cp_fm_type), intent(in)  matrix,
integer, intent(in)  io_unit 
)

Write nicely formatted info about the FM to the given I/O unit (including the underlying FM struct)

Parameters
matrixa cp_fm_type instance
io_unitthe I/O unit to use for writing

Definition at line 1047 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_maxabsval()

subroutine, public cp_fm_types::cp_fm_maxabsval ( type(cp_fm_type), intent(in)  matrix,
real(kind=dp), intent(out)  a_max,
integer, intent(out), optional  ir_max,
integer, intent(out), optional  ic_max 
)

find the maximum absolute value of the matrix element maxval(abs(matrix))

Parameters
matrix...
a_max...
ir_max...
ic_max...

Definition at line 1063 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_maxabsrownorm()

subroutine, public cp_fm_types::cp_fm_maxabsrownorm ( type(cp_fm_type), intent(in)  matrix,
real(kind=dp), intent(out)  a_max 
)

find the maximum over the rows of the sum of the absolute values of the elements of a given row = || A ||_infinity

Parameters
matrix...
a_max...
Note
for a real symmetric matrix it holds that || A ||_2 = |lambda_max| < || A ||_infinity Hence this can be used to estimate an upper bound for the eigenvalues of a matrix http://mathworld.wolfram.com/MatrixNorm.html (but the bound is not so tight in the general case)

Definition at line 1165 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_vectorsnorm()

subroutine, public cp_fm_types::cp_fm_vectorsnorm ( type(cp_fm_type), intent(in)  matrix,
real(kind=dp), dimension(:), intent(out)  norm_array 
)

find the inorm of each column norm_{j}= sqrt( \sum_{i} A_{ij}*A_{ij} )

Parameters
matrix...
norm_array...

Definition at line 1206 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_vectorssum()

subroutine, public cp_fm_types::cp_fm_vectorssum ( type(cp_fm_type), intent(in)  matrix,
real(kind=dp), dimension(:), intent(out)  sum_array,
character(len=1), intent(in), optional  dir 
)

summing up all the elements along the matrix's i-th index \( \mathrm{sum}_{j} = \sum_{i} A_{ij} \) or \( \mathrm{sum}_{i} = \sum_{j} A_{ij} \)

Parameters
matrixan input matrix A
sum_arraysums of elements in each column/row
dir...
Note
forked from cp_fm_vectorsnorm() to be used with the maximum overlap method added row variation

Definition at line 1251 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_to_fm_triangular()

subroutine, public cp_fm_types::cp_fm_to_fm_triangular ( type(cp_fm_type), intent(in)  msource,
type(cp_fm_type), intent(in)  mtarget,
character(len=*), intent(in)  uplo 
)

copy just a triangular matrix

Parameters
msource...
mtarget...
uplo...

Definition at line 1427 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_to_fm_submat()

subroutine, public cp_fm_types::cp_fm_to_fm_submat ( type(cp_fm_type), intent(in)  msource,
type(cp_fm_type), intent(in)  mtarget,
integer, intent(in)  nrow,
integer, intent(in)  ncol,
integer, intent(in)  s_firstrow,
integer, intent(in)  s_firstcol,
integer, intent(in)  t_firstrow,
integer, intent(in)  t_firstcol 
)

copy just a part ot the matrix

Parameters
msource...
mtarget...
nrow...
ncol...
s_firstrow...
s_firstcol...
t_firstrow...
t_firstcol...

Definition at line 1472 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_copy_general()

subroutine, public cp_fm_types::cp_fm_copy_general ( type(cp_fm_type), intent(in)  source,
type(cp_fm_type), intent(in)  destination,
type(mp_para_env_type), intent(in)  para_env 
)

General copy of a fm matrix to another fm matrix. Uses non-blocking MPI rather than ScaLAPACK.

Parameters
sourceinput fm matrix
destinationoutput fm matrix
para_envparallel environment corresponding to the BLACS env that covers all parts of the input and output matrices
History
31-Jan-2017 : Re-implemented using non-blocking MPI [IainB, MarkT]

Definition at line 1537 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_start_copy_general()

subroutine, public cp_fm_types::cp_fm_start_copy_general ( type(cp_fm_type), intent(in)  source,
type(cp_fm_type), intent(in)  destination,
type(mp_para_env_type), intent(in)  para_env,
type(copy_info_type), intent(out)  info 
)

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

Parameters
sourceinput fm matrix
destinationoutput fm matrix
para_envparallel environment corresponding to the BLACS env that covers all parts of the input and output matrices
infoall of the data that will be needed to complete the copy operation

Definition at line 1567 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_finish_copy_general()

subroutine, public cp_fm_types::cp_fm_finish_copy_general ( type(cp_fm_type), intent(in)  destination,
type(copy_info_type), intent(inout)  info 
)

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

Parameters
destinationoutput fm matrix
infoall of the data that will be needed to complete the copy operation

Definition at line 1881 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_cleanup_copy_general()

subroutine, public cp_fm_types::cp_fm_cleanup_copy_general ( type(copy_info_type), intent(inout)  info)

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

Parameters
infoall of the data that will be needed to complete the copy operation

Definition at line 1945 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_to_fm_submat_general()

subroutine, public cp_fm_types::cp_fm_to_fm_submat_general ( type(cp_fm_type), intent(in)  source,
type(cp_fm_type), intent(in)  destination,
integer, intent(in)  nrows,
integer, intent(in)  ncols,
integer, intent(in)  s_firstrow,
integer, intent(in)  s_firstcol,
integer, intent(in)  d_firstrow,
integer, intent(in)  d_firstcol,
class(cp_blacs_type), intent(in)  global_context 
)

General copy of a submatrix of fm matrix to a submatrix of another fm matrix. The two matrices can have different contexts.

Summary of distribution routines for dense matrices The following will copy A(iA:iA+M-1,jA:jA+N-1) to B(iB:iB+M-1,jB:jB+N-1):

call pdgemr2d(M,N,Aloc,iA,jA,descA,Bloc,iB,jB,descB,context)

A process that is not a part of the context of A should set descA(2) to -1, and similarly for B.

Parameters
sourceinput fm matrix
destinationoutput fm matrix
nrowsnumber of rows of sub matrix to be copied
ncolsnumber of cols of sub matrix to be copied
s_firstrowstarting global row index of sub matrix in source
s_firstcolstarting global col index of sub matrix in source
d_firstrowstarting global row index of sub matrix in destination
d_firstcolstarting global col index of sub matrix in destination
global_contextprocess grid that covers all parts of either A or B.

Definition at line 1994 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_add_to_element()

subroutine, public cp_fm_types::cp_fm_add_to_element ( type(cp_fm_type), intent(in)  matrix,
integer, intent(in)  irow_global,
integer, intent(in)  icol_global,
real(kind=dp), intent(in)  alpha 
)

...

Parameters
matrix...
irow_global...
icol_global...
alpha...

Definition at line 2092 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_write_unformatted()

subroutine, public cp_fm_types::cp_fm_write_unformatted ( type(cp_fm_type), intent(in)  fm,
integer, intent(in)  unit 
)

...

Parameters
fm...
unit...

Definition at line 2146 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_write_formatted()

subroutine, public cp_fm_types::cp_fm_write_formatted ( type(cp_fm_type), intent(in)  fm,
integer, intent(in)  unit,
character(len=*), intent(in), optional  header,
character(len=*), intent(in), optional  value_format 
)

Write out a full matrix in plain text.

Parameters
fmthe matrix to be outputted
unitthe unit number for I/O
headeroptional header
value_format...

Definition at line 2250 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_read_unformatted()

subroutine, public cp_fm_types::cp_fm_read_unformatted ( type(cp_fm_type), intent(inout)  fm,
integer, intent(in)  unit 
)

...

Parameters
fm...
unit...

Definition at line 2378 of file cp_fm_types.F.

Here is the call graph for this function:
Here is the caller graph for this function:

◆ cp_fm_indxg2p()

integer function, public cp_fm_types::cp_fm_indxg2p ( integer, intent(in)  INDXGLOB,
integer, intent(in)  NB,
integer, intent(in)  IPROC,
integer, intent(in)  ISRCPROC,
integer, intent(in)  NPROCS 
)

wrapper to scalapack function INDXG2P that computes the process coordinate which possesses the entry of a distributed matrix specified by a global index INDXGLOB.

Arguments

INDXGLOB (global input) INTEGER The global index of the element.

NB (global input) INTEGER Block size, size of the blocks the distributed matrix is split into.

IPROC (local dummy) INTEGER Dummy argument in this case in order to unify the calling sequence of the tool-routines.

ISRCPROC (global input) INTEGER The coordinate of the process that possesses the first row/column of the distributed matrix.

NPROCS (global input) INTEGER The total number processes over which the matrix is distributed.

Parameters
INDXGLOB...
NB...
IPROC...
ISRCPROC...
NPROCS...
Returns
...
Author
Mauro Del Ben [MDB] - 12.2012

Definition at line 2465 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_indxg2l()

integer function, public cp_fm_types::cp_fm_indxg2l ( integer, intent(in)  INDXGLOB,
integer, intent(in)  NB,
integer, intent(in)  IPROC,
integer, intent(in)  ISRCPROC,
integer, intent(in)  NPROCS 
)

wrapper to scalapack function INDXG2L that computes the local index of a distributed matrix entry pointed to by the global index INDXGLOB.

Arguments

INDXGLOB (global input) INTEGER The global index of the distributed matrix entry.

NB (global input) INTEGER Block size, size of the blocks the distributed matrix is split into.

IPROC (local dummy) INTEGER Dummy argument in this case in order to unify the calling sequence of the tool-routines.

ISRCPROC (local dummy) INTEGER Dummy argument in this case in order to unify the calling sequence of the tool-routines.

NPROCS (global input) INTEGER The total number processes over which the distributed matrix is distributed.

Parameters
INDXGLOB...
NB...
IPROC...
ISRCPROC...
NPROCS...
Returns
...
Author
Mauro Del Ben [MDB] - 12.2012

Definition at line 2524 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_indxl2g()

integer function, public cp_fm_types::cp_fm_indxl2g ( integer, intent(in)  INDXLOC,
integer, intent(in)  NB,
integer, intent(in)  IPROC,
integer, intent(in)  ISRCPROC,
integer, intent(in)  NPROCS 
)

wrapper to scalapack function INDXL2G that computes the global index of a distributed matrix entry pointed to by the local index INDXLOC of the process indicated by IPROC.

Arguments

INDXLOC (global input) INTEGER The local index of the distributed matrix entry.

NB (global input) INTEGER Block size, size of the blocks the distributed matrix is split into.

IPROC (local input) INTEGER The coordinate of the process whose local array row or column is to be determined.

ISRCPROC (global input) INTEGER The coordinate of the process that possesses the first row/column of the distributed matrix.

NPROCS (global input) INTEGER The total number processes over which the distributed matrix is distributed.

Parameters
INDXLOC...
NB...
IPROC...
ISRCPROC...
NPROCS...
Returns
...
Author
Mauro Del Ben [MDB] - 12.2012

Definition at line 2584 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_setup()

subroutine, public cp_fm_types::cp_fm_setup ( integer, intent(in)  mult_type)

...

Parameters
mult_type...

Definition at line 2612 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_get_mm_type()

integer function, public cp_fm_types::cp_fm_get_mm_type

...

Returns
...

Definition at line 2622 of file cp_fm_types.F.

Here is the caller graph for this function:

◆ cp_fm_pilaenv()

integer function, public cp_fm_types::cp_fm_pilaenv ( integer  ictxt,
character(len=1)  prec 
)

...

Parameters
ictxt...
prec...
Returns
...

Definition at line 2634 of file cp_fm_types.F.

Here is the caller graph for this function: