The WmVariableAdjusterGradient class is a gradient adjuster. More...

#include <gradient.h>

Inheritance diagram for wxm::dfem::variable_adjuster::gradient::gradient:

Detailed Description

The WmVariableAdjusterGradient class is a gradient adjuster.

Public Member Functions
	gradient ()=default

void	setup (const WxCryptSet &wxc) override
	Setup the spatial solver using the cryptset.

void	solve (real time, variables_type &input) override
	Solves the spatial system and puts result in _rhs.

Public Member Functions inherited from wxm::dfem::variable_adjuster::variable_adjuster_t
	variable_adjuster_t ()
	Constructor.

void	setup (const WxCryptSet &wxc) override
	Setup the spatial solver using the cryptset.

virtual void	solve (real time, variables_type &input)=0
	applies the variable adjuster to `input`

std::vector< wxm::array::patch_array_t * >	get_patch_arrays (const variables_type &in)

void	process () override

virtual const std::vector< size_t > &	get_input_indices () const

int	priority () const
	Getter function for priority.

virtual void	Barrier (const WxMsgBase &msg, const real time, variables_type &input)
	An opportunity for the variable adjuster to perform any MPI operations it needs to after `solve` is called.

Public Member Functions inherited from WmPatchProcess
	WmPatchProcess ()

virtual	~WmPatchProcess ()=default

void	step ()

virtual void	process ()=0

void	setPatch (const WmUnstructuredPatch *patch)

void	setParentTaskProcessor (WmPatchProcessor *taskProcessor)

virtual std::string	name (std::string prefix="patchedProcessor.") const

virtual void	setup (const WxCryptSet &wxc)

void	setParentSolver (const WmSolverBase *solver)

void	setDt (WxStepper::time_t dt)

void	setTime (WxStepper::time_t time)

WxStepper::time_t	getSuggestedDt () const

const WmUnstructuredPatch &	get_patch () const

const WmSolver &	get_solver () const

Protected Member Functions
void	alloc_fluxes ()

void	zero_fluxes ()

void	ex_kernel (const WxRange &face_scope_range, std::vector< wxm::array::patch_array_t * > &input)

void	ldg_scheme (const int faceIndex, const int local_face_node_index, const int global_node_index_in, const int global_node_index_out, solverVariables_t pSV, std::vector< wxm::array::patch_array_t > &input)

void	ip_scheme (const int faceIndex, const int local_face_node_index, const int global_node_index_in, const int global_node_index_out, solverVariables_t pSV, std::vector< wxm::array::patch_array_t > &input)

void	in_kernel (const WxRange &element_scope_range, std::vector< wxm::array::patch_array_t * > &input)

void	co_kernel (const WxRange &element_scope_range, std::vector< wxm::array::patch_array_t * > &input)

Protected Attributes
std::vector< size_t >	_gradient_input_indexes

std::vector< size_t >	_gradient_output_indexes

std::string	_gradient_method

real	_penalty_beta

real	_penalty_eta

std::string	_basisSetName

WmBasisArraySet	_basis_set

std::unique_ptr< WmUDGGeometry >	_dg_geometry

std::vector< std::vector< std::vector< std::vector< real > > > >	_numerical_fluxes

std::vector< std::vector< std::vector< real > > >	_internal_fluxes

wxm::dfem::tools::scope_t	_element_flux

wxm::dfem::tools::scope_t	_face_flux

WxRange	_chunk_size_element

WxRange	_chunk_size_face

bool	_standard_ldg_flux_alternation
	"Standard" refers to the sign when adding the beta penalty for LDG, the u*_h numerical flux value as given in Table 7.3 in Hesthaven & Warburton, Nodal Discontinuous Galerkin Methods, 2008.

Protected Attributes inherited from wxm::dfem::variable_adjuster::variable_adjuster_t
std::vector< size_t >	_var_idcs

std::vector< std::unique_ptr< WmApplication > >	_apps

bool	_grab_all = false

int	_priority

Protected Attributes inherited from WmPatchProcess
std::string	_ppName

WmPatchProcessor *	_parentTaskProcessor

const WmUnstructuredPatch *	_patch

std::vector< std::string >	_onSubdomains

WxStepper::time_t	_time

WxStepper::time_t	_dt

WxStepper::time_t	_suggested_dt

WxLogStream	_debStrm = WxLogger::get("warpx-root.console")->getDebugStream()

const WmSolverBase *	_parentSolver = nullptr

Additional Inherited Members
Public Types inherited from wxm::dfem::variable_adjuster::variable_adjuster_t
typedef wxm::temporal_solver::variables_type	variables_type

Constructor & Destructor Documentation

◆ gradient()

wxm::dfem::variable_adjuster::gradient::gradient::gradient ( )

default

Member Function Documentation

◆ alloc_fluxes()

void wxm::dfem::variable_adjuster::gradient::gradient::alloc_fluxes ( )

protected

◆ co_kernel()

void wxm::dfem::variable_adjuster::gradient::gradient::co_kernel	(	const WxRange &	element_scope_range,
		std::vector< wxm::array::patch_array_t * > &	input
	)

protected

◆ ex_kernel()

void wxm::dfem::variable_adjuster::gradient::gradient::ex_kernel	(	const WxRange &	face_scope_range,
		std::vector< wxm::array::patch_array_t * > &	input
	)

protected

◆ in_kernel()

void wxm::dfem::variable_adjuster::gradient::gradient::in_kernel	(	const WxRange &	element_scope_range,
		std::vector< wxm::array::patch_array_t * > &	input
	)

protected

◆ ip_scheme()

void wxm::dfem::variable_adjuster::gradient::gradient::ip_scheme	(	const int	faceIndex,
		const int	local_face_node_index,
		const int	global_node_index_in,
		const int	global_node_index_out,
		solverVariables_t *	pSV,
		std::vector< wxm::array::patch_array_t * > &	input
	)

protected

◆ ldg_scheme()

void wxm::dfem::variable_adjuster::gradient::gradient::ldg_scheme	(	const int	faceIndex,
		const int	local_face_node_index,
		const int	global_node_index_in,
		const int	global_node_index_out,
		solverVariables_t *	pSV,
		std::vector< wxm::array::patch_array_t * > &	input
	)

protected

◆ setup()

void wxm::dfem::variable_adjuster::gradient::gradient::setup ( const WxCryptSet & wxc )

overridevirtual

Setup the spatial solver using the cryptset.

Parameters

wxc	The cryptset

Reimplemented from WmPatchProcess.

◆ solve()

void wxm::dfem::variable_adjuster::gradient::gradient::solve	(	real	time,
		variables_type &	input
	)

overridevirtual

Solves the spatial system and puts result in _rhs.

Implements wxm::dfem::variable_adjuster::variable_adjuster_t.

◆ zero_fluxes()

void wxm::dfem::variable_adjuster::gradient::gradient::zero_fluxes ( )

protected

Member Data Documentation

◆ _basis_set

WmBasisArraySet wxm::dfem::variable_adjuster::gradient::gradient::_basis_set

protected

◆ _basisSetName

std::string wxm::dfem::variable_adjuster::gradient::gradient::_basisSetName

protected

◆ _chunk_size_element

WxRange wxm::dfem::variable_adjuster::gradient::gradient::_chunk_size_element

protected

◆ _chunk_size_face

WxRange wxm::dfem::variable_adjuster::gradient::gradient::_chunk_size_face

protected

◆ _dg_geometry

std::unique_ptr<WmUDGGeometry> wxm::dfem::variable_adjuster::gradient::gradient::_dg_geometry

protected

◆ _element_flux

wxm::dfem::tools::scope_t wxm::dfem::variable_adjuster::gradient::gradient::_element_flux

protected

◆ _face_flux

wxm::dfem::tools::scope_t wxm::dfem::variable_adjuster::gradient::gradient::_face_flux

protected

◆ _gradient_input_indexes

std::vector<size_t> wxm::dfem::variable_adjuster::gradient::gradient::_gradient_input_indexes

protected

◆ _gradient_method

std::string wxm::dfem::variable_adjuster::gradient::gradient::_gradient_method

protected

◆ _gradient_output_indexes

std::vector<size_t> wxm::dfem::variable_adjuster::gradient::gradient::_gradient_output_indexes

protected

◆ _internal_fluxes

std::vector<std::vector<std::vector<real> > > wxm::dfem::variable_adjuster::gradient::gradient::_internal_fluxes

protected

◆ _numerical_fluxes

std::vector<std::vector<std::vector<std::vector<real> > > > wxm::dfem::variable_adjuster::gradient::gradient::_numerical_fluxes

protected

◆ _penalty_beta

real wxm::dfem::variable_adjuster::gradient::gradient::_penalty_beta

protected

◆ _penalty_eta

real wxm::dfem::variable_adjuster::gradient::gradient::_penalty_eta

protected

◆ _standard_ldg_flux_alternation

bool wxm::dfem::variable_adjuster::gradient::gradient::_standard_ldg_flux_alternation

protected

"Standard" refers to the sign when adding the beta penalty for LDG, the u*_h numerical flux value as given in Table 7.3 in Hesthaven & Warburton, Nodal Discontinuous Galerkin Methods, 2008.

For gradients when using LDG, you should use the "alternating principle", where the sign for adding the beta penalty alternates when calculating the numerical flux for the gradient and the time-advance equation. If you have a higher order derivative, you might want to keeep alternating, so if you use this gradient solver successively for multiple gradients, you might need to switch each one with _standard_ldg_flux_alternation = True / False / True, etc., for example. Note that when calculating the numerical flux for the time-advance equation (using the q*_h value for the numerical flux in Table 7.3), you have to make the consistent choice of sign on beta in accordance to the alternating principle. Typically, beta should be either -0.5, +0.5, or 0. If you choose beta = 0, you are reverting to using "central flux" where you are no longer alternating flux directions when calculating numerical fluxes for the gradient and solving the time-advance equation, which seems to work for parabolic equations, though if you were to use it to solve elliptic equations, it might be unstable; See: Arnold, Siam J. Numer. Anal., Vol. 39, No. 5, pp. 1749-1779, 2002 (Unified Analysis of Discontinuous Galerkin Methods for Elliptic Problems). Also see Xu & Shu, Commun. Comput. Phys., Vol. 7, No. 1, pp. 1-46, 2010 (Local Discontinuous Methods for High-Order Time-Dependent Partial Differential Equations) as well as Hesthaven Ch. 7 for more on LDG, the alternating principle, and examples of solved equations. Datta Dissertation 2021, Ch. 3, also has a description LDG and interior penalty as implemented here.

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

src/dfem/variable_adjusters/gradient/gradient.h

Detailed Description

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Constructor & Destructor Documentation

◆ gradient()

Member Function Documentation

◆ alloc_fluxes()

◆ co_kernel()

◆ ex_kernel()

◆ in_kernel()

◆ ip_scheme()

◆ ldg_scheme()

◆ setup()

◆ solve()

◆ zero_fluxes()

Member Data Documentation

◆ _basis_set

◆ _basisSetName

◆ _chunk_size_element

◆ _chunk_size_face

◆ _dg_geometry

◆ _element_flux

◆ _face_flux

◆ _gradient_input_indexes

◆ _gradient_method

◆ _gradient_output_indexes

◆ _internal_fluxes

◆ _numerical_fluxes

◆ _penalty_beta

◆ _penalty_eta

◆ _standard_ldg_flux_alternation