cp_cfm_elpa Module Reference

Wrapper for ELPA (complex matrices, i.e. cp_cfm_type) More...

Functions/Subroutines
subroutine, public	set_elpa_c_kernel (requested_kernel)
	Sets the active ELPA kernel for complex matrices.
subroutine, public	check_elpa_c_kernel_correctness (para_env)
	One-time runtime correctness check for ELPA complex BLOCK2 kernels. A small deterministic eigenproblem is solved on a single process with the same production routine used for the actual diagonalizations, and the eigenpair residual as well as the eigenvector orthogonality are verified. This guards against mis-compiled kernels (GCC 15.2 regression, marekandreas/elpa#77). If the check fails, a module flag is set such that cp_cfm_heevd falls back to ScaLAPACK.
logical function, public	is_elpa_c_broken ()
	Returns .TRUE. if the runtime check detected broken BLOCK2 kernels.
subroutine, public	cp_cfm_diag_elpa (matrix, eigenvectors, eigenvalues)
	Driver routine to diagonalize a CFM matrix with the ELPA library.

Variables
integer, dimension(1), parameter, public	elpa_c_kernel_ids = [-1]
character(len=14), dimension(1), parameter, public	elpa_c_kernel_names = ["AUTO"]
character(len=44), dimension(1), parameter, public	elpa_c_kernel_descriptions = ["Automatically selected kernel"]

Detailed Description

Wrapper for ELPA (complex matrices, i.e. cp_cfm_type)

Function/Subroutine Documentation

set_elpa_c_kernel()

subroutine, public cp_cfm_elpa::set_elpa_c_kernel

(

integer, intent(in)

requested_kernel

)

Sets the active ELPA kernel for complex matrices.

Parameters

requested_kernel

one of the elpa_c_kernel_ids

Definition at line 190 of file cp_cfm_elpa.F.

Here is the caller graph for this function:

check_elpa_c_kernel_correctness()

subroutine, public cp_cfm_elpa::check_elpa_c_kernel_correctness

(

type(mp_para_env_type), intent(in)

para_env

)

One-time runtime correctness check for ELPA complex BLOCK2 kernels. A small deterministic eigenproblem is solved on a single process with the same production routine used for the actual diagonalizations, and the eigenpair residual as well as the eigenvector orthogonality are verified. This guards against mis-compiled kernels (GCC 15.2 regression, marekandreas/elpa#77). If the check fails, a module flag is set such that cp_cfm_heevd falls back to ScaLAPACK.

Parameters

para_env

communicator of the run, used to agree on the check result

Definition at line 263 of file cp_cfm_elpa.F.

Here is the call graph for this function:

Here is the caller graph for this function:

is_elpa_c_broken()

logical function, public cp_cfm_elpa::is_elpa_c_broken

Returns .TRUE. if the runtime check detected broken BLOCK2 kernels.

Returns

...

Definition at line 378 of file cp_cfm_elpa.F.

Here is the caller graph for this function:

cp_cfm_diag_elpa()

subroutine, public cp_cfm_elpa::cp_cfm_diag_elpa	(	type(cp_cfm_type), intent(in)	matrix,
		type(cp_cfm_type), intent(in)	eigenvectors,
		real(kind=dp), dimension(:), intent(out)	eigenvalues
	)

Driver routine to diagonalize a CFM matrix with the ELPA library.

Parameters

matrix	the matrix that is diagonalized
eigenvectors	eigenvectors of the input matrix
eigenvalues	eigenvalues of the input matrix

Definition at line 390 of file cp_cfm_elpa.F.

Here is the call graph for this function:

Here is the caller graph for this function:

Variable Documentation

elpa_c_kernel_ids

integer, dimension(1), parameter, public cp_cfm_elpa::elpa_c_kernel_ids = [-1]

Definition at line 141 of file cp_cfm_elpa.F.

elpa_c_kernel_names

character(len=14), dimension(1), parameter, public cp_cfm_elpa::elpa_c_kernel_names = ["AUTO"]

Definition at line 142 of file cp_cfm_elpa.F.

elpa_c_kernel_descriptions

character(len=44), dimension(1), parameter, public cp_cfm_elpa::elpa_c_kernel_descriptions = ["Automatically selected kernel"]

Definition at line 143 of file cp_cfm_elpa.F.