Teko Version of the Day
Teko::NS::LSCSIMPLECStrategy Class Reference

A strategy that takes a single inverse factory and uses that for all inverses. If no mass matrix is passed in the diagonal of the 1,1 block is used. More...

#include <Teko_LSCSIMPLECStrategy.hpp>

Inheritance diagram for Teko::NS::LSCSIMPLECStrategy:
Inheritance graph
[legend]

List of all members.

Public Member Functions

Constructors
 LSCSIMPLECStrategy ()
virtual void buildState (BlockedLinearOp &A, BlockPreconditionerState &state) const
 Functions inherited from LSCStrategy.
virtual LinearOp getInvBQBt (const BlockedLinearOp &A, BlockPreconditionerState &state) const
virtual LinearOp getInvBHBt (const BlockedLinearOp &A, BlockPreconditionerState &state) const
virtual LinearOp getInvF (const BlockedLinearOp &A, BlockPreconditionerState &state) const
virtual LinearOp getOuterStabilization (const BlockedLinearOp &A, BlockPreconditionerState &state) const
virtual LinearOp getInnerStabilization (const BlockedLinearOp &A, BlockPreconditionerState &state) const
virtual LinearOp getInvMass (const BlockedLinearOp &A, BlockPreconditionerState &state) const
virtual LinearOp getHScaling (const BlockedLinearOp &A, BlockPreconditionerState &state) const
virtual bool useFullLDU () const
virtual void initializeFromParameterList (const Teuchos::ParameterList &pl, const InverseLibrary &invLib)
 Initialize from a parameter list.
virtual void initializeState (const BlockedLinearOp &A, LSCPrecondState *state) const
 Initialize the state object using this blocked linear operator.
void computeInverses (const BlockedLinearOp &A, LSCPrecondState *state) const
virtual void setUseFullLDU (bool val)
 Set to true to use the Full LDU decomposition, false otherwise.
virtual void setSymmetric (bool isSymmetric)

Detailed Description

A strategy that takes a single inverse factory and uses that for all inverses. If no mass matrix is passed in the diagonal of the 1,1 block is used.

A strategy that takes a single inverse factory and uses that for all inverses. Optionally the mass matrix can be passed in, if it is the diagonal is extracted and that is used to form the inverse approximation.

Definition at line 66 of file Teko_LSCSIMPLECStrategy.hpp.


Member Function Documentation

void Teko::NS::LSCSIMPLECStrategy::buildState ( BlockedLinearOp &  A,
BlockPreconditionerState state 
) const [virtual]

Functions inherited from LSCStrategy.

This informs the strategy object to build the state associated with this operator.

Parameters:
[in]AThe linear operator to be preconditioned by LSC.
[in]stateState object for storying reusable information about the operator A.

Implements Teko::NS::LSCStrategy.

Definition at line 91 of file Teko_LSCSIMPLECStrategy.cpp.

LinearOp Teko::NS::LSCSIMPLECStrategy::getInvBQBt ( const BlockedLinearOp &  A,
BlockPreconditionerState state 
) const [virtual]

Get the inverse of $B Q_u^{-1} B^T$.

Parameters:
[in]AThe linear operator to be preconditioned by LSC.
[in]stateState object for storying reusable information about the operator A.
Returns:
An (approximate) inverse of $B Q_u^{-1} B^T$.

Implements Teko::NS::LSCStrategy.

Definition at line 127 of file Teko_LSCSIMPLECStrategy.cpp.

LinearOp Teko::NS::LSCSIMPLECStrategy::getInvBHBt ( const BlockedLinearOp &  A,
BlockPreconditionerState state 
) const [virtual]

Get the inverse of $B H B^T - \gamma C$.

Parameters:
[in]AThe linear operator to be preconditioned by LSC.
[in]stateState object for storying reusable information about the operator A.
Returns:
An (approximate) inverse of $B H B^T - \gamma C$.

Implements Teko::NS::LSCStrategy.

Definition at line 132 of file Teko_LSCSIMPLECStrategy.cpp.

LinearOp Teko::NS::LSCSIMPLECStrategy::getInvF ( const BlockedLinearOp &  A,
BlockPreconditionerState state 
) const [virtual]

Get the inverse of the $F$ block.

Parameters:
[in]AThe linear operator to be preconditioned by LSC.
[in]stateState object for storying reusable information about the operator A.
Returns:
An (approximate) inverse of $F$.

Implements Teko::NS::LSCStrategy.

Definition at line 137 of file Teko_LSCSIMPLECStrategy.cpp.

virtual LinearOp Teko::NS::LSCSIMPLECStrategy::getOuterStabilization ( const BlockedLinearOp &  A,
BlockPreconditionerState state 
) const [inline, virtual]

Get the inverse for stabilizing the whole schur complement approximation. Returns $\alpha D^{-1}$.

Parameters:
[in]AThe linear operator to be preconditioned by LSC.
[in]stateState object for storying reusable information about the operator A.
Returns:
The operator to stabilize the whole Schur complement.

Implements Teko::NS::LSCStrategy.

Definition at line 126 of file Teko_LSCSIMPLECStrategy.hpp.

LinearOp Teko::NS::LSCSIMPLECStrategy::getInnerStabilization ( const BlockedLinearOp &  A,
BlockPreconditionerState state 
) const [virtual]

Get the inverse to stablized stabilizing the Schur complement approximation using a placement on the ``inside''. That is what is the value for $C_I$. This quantity may be null.

Parameters:
[in]AThe linear operator to be preconditioned by LSC.
[in]stateState object for storying reusable information about the operator A.
Returns:
The operator to stabilize the whole Schur complement.

Implements Teko::NS::LSCStrategy.

Definition at line 142 of file Teko_LSCSIMPLECStrategy.cpp.

LinearOp Teko::NS::LSCSIMPLECStrategy::getInvMass ( const BlockedLinearOp &  A,
BlockPreconditionerState state 
) const [virtual]

Get the inverse mass matrix.

Parameters:
[in]AThe linear operator to be preconditioned by LSC.
[in]stateState object for storying reusable information about the operator A.
Returns:
The inverse of the mass matrix $Q_u$.

Implements Teko::NS::LSCStrategy.

Definition at line 153 of file Teko_LSCSIMPLECStrategy.cpp.

LinearOp Teko::NS::LSCSIMPLECStrategy::getHScaling ( const BlockedLinearOp &  A,
BlockPreconditionerState state 
) const [virtual]

Get the $H$ scaling matrix.

Parameters:
[in]AThe linear operator to be preconditioned by LSC.
[in]stateState object for storying reusable information about the operator A.
Returns:
The $H$ scaling matrix.

Implements Teko::NS::LSCStrategy.

Definition at line 162 of file Teko_LSCSIMPLECStrategy.cpp.

virtual bool Teko::NS::LSCSIMPLECStrategy::useFullLDU ( ) const [inline, virtual]

Should the approximation of the inverse use a full LDU decomposition, or is a upper triangular approximation sufficient.

Returns:
True if the full LDU decomposition should be used, otherwise only an upper triangular version is used.

Implements Teko::NS::LSCStrategy.

Definition at line 157 of file Teko_LSCSIMPLECStrategy.hpp.

void Teko::NS::LSCSIMPLECStrategy::initializeFromParameterList ( const Teuchos::ParameterList &  pl,
const InverseLibrary &  invLib 
) [virtual]

Initialize from a parameter list.

Reimplemented from Teko::NS::LSCStrategy.

Definition at line 239 of file Teko_LSCSIMPLECStrategy.cpp.

void Teko::NS::LSCSIMPLECStrategy::initializeState ( const BlockedLinearOp &  A,
LSCPrecondState state 
) const [virtual]

Initialize the state object using this blocked linear operator.

Definition at line 168 of file Teko_LSCSIMPLECStrategy.cpp.

void Teko::NS::LSCSIMPLECStrategy::computeInverses ( const BlockedLinearOp &  A,
LSCPrecondState state 
) const

Compute the inverses required for the LSC Schur complement

Note:
This method assumes that the BQBt and BHBt operators have been constructed.

Definition at line 199 of file Teko_LSCSIMPLECStrategy.cpp.

virtual void Teko::NS::LSCSIMPLECStrategy::setUseFullLDU ( bool  val) [inline, virtual]

Set to true to use the Full LDU decomposition, false otherwise.

Definition at line 174 of file Teko_LSCSIMPLECStrategy.hpp.

virtual void Teko::NS::LSCSIMPLECStrategy::setSymmetric ( bool  isSymmetric) [inline, virtual]

Tell strategy that this operator is supposed to be symmetric. Behavior of LSC is slightly different for non-symmetric case.

Parameters:
[in]isSymmetricIs this operator symmetric?

Implements Teko::NS::LSCStrategy.

Definition at line 176 of file Teko_LSCSIMPLECStrategy.hpp.


The documentation for this class was generated from the following files:
 All Classes Files Functions Variables