Peano

Searching...

No Matches

Introduce additive CC Coupling. More...

Inheritance diagram for solvers.api.actionsets.DGCGCoupling.AdditiveDGCGCoupling:

[legend]

Collaboration diagram for solvers.api.actionsets.DGCGCoupling.AdditiveDGCGCoupling:

[legend]

Public Member Functions
	__init__ (self, dg_solver, cg_solver, prolongation_matrix, prolongation_matrix_scaling, restriction_matrix, restriction_matrix_scaling, injection_matrix, injection_matrix_scaling, use_fas, smoothing_steps_per_cycle=1)
	Construct action set.

	get_attributes (self)
	Add a new static attribute to the class.

	get_static_initialisations (self, full_qualified_classname)
	Initialise the new counter.

	get_body_of_prepareTraversal (self)

Public Member Functions inherited from solvers.api.actionsets.DGCGCoupling.AbstractDGCGCoupling
	get_body_of_operation (self, operation_name)
	Return actual C++ code snippets to be inserted into C++ code.

	get_action_set_name (self)
	Configure name of generated C++ action set.

	user_should_modify_template (self)
	The action set that Peano will generate that corresponds to this class should not be modified by users and can safely be overwritten every time we run the Python toolkit.

	get_includes (self)
	We need the solver repository in this action set, as we directly access the solver object.

Public Member Functions inherited from peano4.solversteps.ActionSet.ActionSet
	get_constructor_body (self)
	Define a tailored constructor body.

	get_destructor_body (self)

	get_body_of_getGridControlEvents (self)

	get_body_of_unprepareTraversal (self)

Data Fields
	d

Data Fields inherited from solvers.api.actionsets.DGCGCoupling.AbstractDGCGCoupling
	d

Data Fields inherited from peano4.solversteps.ActionSet.ActionSet
	descend_invocation_order

	parallel

Additional Inherited Members
Static Public Attributes inherited from solvers.api.actionsets.DGCGCoupling.AbstractDGCGCoupling
str	templateTouchCellFirstTime

Static Public Attributes inherited from peano4.solversteps.ActionSet.ActionSet
str	OPERATION_BEGIN_TRAVERSAL = "beginTraversal"

str	OPERATION_END_TRAVERSAL = "endTraversal"

str	OPERATION_CREATE_PERSISTENT_VERTEX = "createPersistentVertex"

str	OPERATION_DESTROY_PERSISTENT_VERTEX = "destroyPersistentVertex"

str	OPERATION_CREATE_HANGING_VERTEX = "createHangingVertex"

str	OPERATION_DESTROY_HANGING_VERTEX = "destroyHangingVertex"

str	OPERATION_CREATE_PERSISTENT_FACE = "createPersistentFace"

str	OPERATION_DESTROY_PERSISTENT_FACE = "destroyPersistentFace"

str	OPERATION_CREATE_HANGING_FACE = "createHangingFace"

str	OPERATION_DESTROY_HANGING_FACE = "destroyHangingFace"

str	OPERATION_CREATE_CELL = "createCell"

str	OPERATION_DESTROY_CELL = "destroyCell"

str	OPERATION_TOUCH_VERTEX_FIRST_TIME = "touchVertexFirstTime"

str	OPERATION_TOUCH_VERTEX_LAST_TIME = "touchVertexLastTime"

str	OPERATION_TOUCH_FACE_FIRST_TIME = "touchFaceFirstTime"

str	OPERATION_TOUCH_FACE_LAST_TIME = "touchFaceLastTime"

str	OPERATION_TOUCH_CELL_FIRST_TIME = "touchCellFirstTime"

str	OPERATION_TOUCH_CELL_LAST_TIME = "touchCellLastTime"

Protected Attributes inherited from solvers.api.actionsets.DGCGCoupling.AbstractDGCGCoupling
	_use_fas

Detailed Description

Introduce additive CC Coupling.

The control logic here is simplistic and prescripted. We follow the following steps for the additive solver with one iteration:

Pre step
- Interpolate correction, i.e. sum all the hierarchies up. This happens in touchCellFirstTime().
- Compute and restrict residual and inject solution (if FAS). This happens in touchCellLastTime().
- Suspend both solvers, but let DG solver project its solution onto the faces.
Compute step
- All solvers compute.
- DG solver can project solution onto faces.
- Return to step 1.

Nothing stops us from repeating the step (2) multiple times, which gives us the opportunity to play around with various multigrid schemes. If we are in the last sweep of (2), the projection onto the faces of the DG solver is not necessary. We'll overwrite this solution a minute later anyway.

Realisation

This action set will map onto a class of its own. We give the class a class attribute which we increment by one in each step. We then make the actual logic depend upon this counter, as we inject the predicates into the superclass: That is, the superclass has some boolean expressions which allow us to switch its features on and off. We make those guys depends upon the new counter.

Definition at line 529 of file DGCGCoupling.py.

Constructor & Destructor Documentation

◆ init()

solvers.api.actionsets.DGCGCoupling.AdditiveDGCGCoupling.__init__	(	self,
		dg_solver,
		cg_solver,
		prolongation_matrix,
		prolongation_matrix_scaling,
		restriction_matrix,
		restriction_matrix_scaling,
		injection_matrix,
		injection_matrix_scaling,
		use_fas,
		smoothing_steps_per_cycle = 1 )

Construct action set.

Parameters

prolongation_matrix	Please consult the action set ComputeAndRestrictBiasedHierarchicalResidual.
prolongation_matrix_scaling	Please consult the action set ComputeAndRestrictBiasedHierarchicalResidual.
injection_matrix	[ \( \mathbb{R}^{2^dK \times (p+1}^dK } \) ] Sequence of matrices. Each takes the solution over the cell spanned by \( (p+1}^dK \) weights and maps them onto \( K \) quantities on the vertices. These quantities are typically just the value of the polynomial in the vertex, but you can pick other schemes, too. Most of the time, it will be a single injection (with one scaling, see below). To be in line with the other signatures, you can however specify the total injection through a sequence of matrices, each one with its own \( h^s \) scaling.
injection_matrix_scaling	[ Integer ] Specifies a sequence of values \( s \). They define with which \( h^s \) the corresponding matrices in injection_matrix are to be scaled. That is, if you pass in a [3,0], then the first matrix in injection_matrix will be scaled with \( h^3 \) and the second one with one. If you work with a simple function evaluation as injection, then you are fine with one matrix, and its scaline will be [ 0 ].
use_fas	Boolean If this flag is set, we realise a FAS following the HTMG idea by Griebel et al.