Teko Version of the Day
Loading...
Searching...
No Matches
Functions
Teko_Utilities.hpp File Reference
#include "Teko_ConfigDefs.hpp"
#include "Tpetra_CrsMatrix.hpp"
#include "Teuchos_VerboseObject.hpp"
#include "Thyra_LinearOpBase.hpp"
#include "Thyra_PhysicallyBlockedLinearOpBase.hpp"
#include "Thyra_ProductVectorSpaceBase.hpp"
#include "Thyra_VectorSpaceBase.hpp"
#include "Thyra_ProductMultiVectorBase.hpp"
#include "Thyra_MultiVectorStdOps.hpp"
#include "Thyra_MultiVectorBase.hpp"
#include "Thyra_VectorBase.hpp"
#include "Thyra_VectorStdOps.hpp"
#include "Thyra_DefaultBlockedLinearOp.hpp"
#include "Thyra_DefaultMultipliedLinearOp.hpp"
#include "Thyra_DefaultScaledAdjointLinearOp.hpp"
#include "Thyra_DefaultAddedLinearOp.hpp"
#include "Thyra_DefaultIdentityLinearOp.hpp"
#include "Thyra_DefaultZeroLinearOp.hpp"

Go to the source code of this file.

Functions

RCP< Tpetra::CrsMatrix< ST, LO, GO, NT > > Teko::buildGraphLaplacian (int dim, ST *coords, const Tpetra::CrsMatrix< ST, LO, GO, NT > &stencil)
 Build a graph Laplacian stenciled on a Epetra_CrsMatrix.
RCP< Tpetra::CrsMatrix< ST, LO, GO, NT > > Teko::buildGraphLaplacian (ST *x, ST *y, ST *z, GO stride, const Tpetra::CrsMatrix< ST, LO, GO, NT > &stencil)
 Build a graph Laplacian stenciled on a Epetra_CrsMatrix.
const Teuchos::RCP< Teuchos::FancyOStream > Teko::getOutputStream ()
 Function used internally by Teko to find the output stream.

LinearOp utilities

Let the blocked operator know that you are going to set the sub blocks.

typedef Teuchos::RCP<Thyra::PhysicallyBlockedLinearOpBase<ST> > BlockedLinearOp; typedef Teuchos::RCP<const Thyra::LinearOpBase<ST> > LinearOp; typedef Teuchos::RCP<Thyra::LinearOpBase<ST> > InverseLinearOp; typedef Teuchos::RCP<Thyra::LinearOpBase<ST> > ModifiableLinearOp;

Build a square zero operator from a single vector space inline LinearOp zero(const VectorSpace &vs) { return Thyra::zero<ST>(vs, vs); }

inline LinearOp zero(const VectorSpace &range, const VectorSpace &domain) { return Thyra::zero<ST>(range, domain); }

Replace nonzeros with a scalar value, used to zero out an operator

Get the range space of a linear operator inline VectorSpace rangeSpace(const LinearOp &lo) { return lo->range(); }

Get the domain space of a linear operator inline VectorSpace domainSpace(const LinearOp &lo) { return lo->domain(); }

Converts a LinearOp to a BlockedLinearOp inline BlockedLinearOp toBlockedLinearOp(LinearOp &clo) { Teuchos::RCP<Thyra::LinearOpBase<double> > lo = Teuchos::rcp_const_cast<Thyra::LinearOpBase<double> >(clo); return Teuchos::rcp_dynamic_cast<Thyra::PhysicallyBlockedLinearOpBase<double> >(lo); }

Converts a LinearOp to a BlockedLinearOp inline const BlockedLinearOp toBlockedLinearOp(const LinearOp &clo) { Teuchos::RCP<Thyra::LinearOpBase<double> > lo = Teuchos::rcp_const_cast<Thyra::LinearOpBase<double> >(clo); return Teuchos::rcp_dynamic_cast<Thyra::PhysicallyBlockedLinearOpBase<double> >(lo); }

Convert to a LinearOp from a BlockedLinearOp inline LinearOp toLinearOp(BlockedLinearOp &blo) { return blo; }

Convert to a LinearOp from a BlockedLinearOp inline const LinearOp toLinearOp(const BlockedLinearOp &blo) { return blo; }

Convert to a LinearOp from a BlockedLinearOp inline LinearOp toLinearOp(ModifiableLinearOp &blo) { return blo; }

Convert to a LinearOp from a BlockedLinearOp inline const LinearOp toLinearOp(const ModifiableLinearOp &blo) { return blo; }

Get the row count in a block linear operator inline int blockRowCount(const BlockedLinearOp &blo) { return blo->productRange()->numBlocks(); }

Get the column count in a block linear operator inline int blockColCount(const BlockedLinearOp &blo) { return blo->productDomain()->numBlocks(); }

Get the i,j block in a BlockedLinearOp object inline LinearOp getBlock(int i, int j, const BlockedLinearOp &blo) { return blo->getBlock(i, j); }

Set the i,j block in a BlockedLinearOp object inline void setBlock(int i, int j, BlockedLinearOp &blo, const LinearOp &lo) { return blo->setBlock(i, j, lo); }

Build a new blocked linear operator inline BlockedLinearOp createBlockedOp() { return rcp(new Thyra::DefaultBlockedLinearOp<double>()); }

/**

Let the blocked operator know that you are going to set the sub blocks. This is a simple wrapper around the member function of the same name in Thyra.

Parameters
[in,out]bloBlocked operator to have its fill stage activated
[in]rowCntNumber of block rows in this operator
[in]colCntNumber of block columns in this operator
enum  Teko::DiagonalType {
  Teko::Diagonal , Teko::Lumped , Teko::AbsRowSum , Teko::BlkDiag ,
  Teko::NotDiag
}
void Teko::beginBlockFill (BlockedLinearOp &blo)
 Let the blocked operator know that you are going to set the sub blocks.
BlockedLinearOp Teko::getUpperTriBlocks (const BlockedLinearOp &blo, bool callEndBlockFill)
 Get the strictly upper triangular portion of the matrix.
BlockedLinearOp Teko::getLowerTriBlocks (const BlockedLinearOp &blo, bool callEndBlockFill)
 Get the strictly lower triangular portion of the matrix.
BlockedLinearOp Teko::zeroBlockedOp (const BlockedLinearOp &blo)
 Build a zero operator mimicing the block structure of the passed in matrix.
ModifiableLinearOp Teko::getAbsRowSumMatrix (const LinearOp &op)
 Compute absolute row sum matrix.
ModifiableLinearOp Teko::getAbsRowSumInvMatrix (const LinearOp &op)
 Compute inverse of the absolute row sum matrix.
ModifiableLinearOp Teko::getLumpedMatrix (const LinearOp &op)
 Compute the lumped version of this matrix.
ModifiableLinearOp Teko::getInvLumpedMatrix (const LinearOp &op)
 Compute the inverse of the lumped version of this matrix.
void Teko::applyOp (const LinearOp &A, const MultiVector &x, MultiVector &y, double alpha, double beta)
 Apply a linear operator to a multivector (think of this as a matrix vector multiply).
void Teko::applyTransposeOp (const LinearOp &A, const MultiVector &x, MultiVector &y, double alpha, double beta)
 Apply a transposed linear operator to a multivector (think of this as a matrix vector multiply).
void Teko::applyOp (const LinearOp &A, const BlockedMultiVector &x, BlockedMultiVector &y, double alpha=1.0, double beta=0.0)
 Apply a linear operator to a blocked multivector (think of this as a matrix vector multiply).
void Teko::applyTransposeOp (const LinearOp &A, const BlockedMultiVector &x, BlockedMultiVector &y, double alpha=1.0, double beta=0.0)
 Apply a transposed linear operator to a blocked multivector (think of this as a matrix vector multiply).
void Teko::update (double alpha, const MultiVector &x, double beta, MultiVector &y)
 Update the y vector so that $y = \alpha x+\beta y$.
const ModifiableLinearOp Teko::getDiagonalOp (const LinearOp &op)
 Get the diaonal of a linear operator.
const MultiVector Teko::getDiagonal (const LinearOp &op)
 Get the diagonal of a linear operator.
const ModifiableLinearOp Teko::getInvDiagonalOp (const LinearOp &op)
 Get the diaonal of a linear operator.
const LinearOp Teko::explicitMultiply (const LinearOp &opl, const LinearOp &opm, const LinearOp &opr)
 Multiply three linear operators.
const ModifiableLinearOp Teko::explicitMultiply (const LinearOp &opl, const LinearOp &opm, const LinearOp &opr, const ModifiableLinearOp &destOp)
 Multiply three linear operators.
const LinearOp Teko::explicitMultiply (const LinearOp &opl, const LinearOp &opr)
 Multiply two linear operators.
const ModifiableLinearOp Teko::explicitMultiply (const LinearOp &opl, const LinearOp &opr, const ModifiableLinearOp &destOp)
 Multiply two linear operators.
const LinearOp Teko::explicitAdd (const LinearOp &opl_in, const LinearOp &opr_in)
 Add two linear operators.
const ModifiableLinearOp Teko::explicitAdd (const LinearOp &opl_in, const LinearOp &opr_in, const ModifiableLinearOp &destOp)
 Add two linear operators.
const ModifiableLinearOp Teko::explicitSum (const LinearOp &op, const ModifiableLinearOp &destOp)
 Sum an operator.
const LinearOp Teko::explicitTranspose (const LinearOp &op)
const LinearOp Teko::explicitScale (double scalar, const LinearOp &op)
double Teko::frobeniusNorm (const LinearOp &op_in)
const LinearOp Teko::buildDiagonal (const MultiVector &v, const std::string &lbl="ANYM")
 Take the first column of a multivector and build a diagonal linear operator.
const LinearOp Teko::buildInvDiagonal (const MultiVector &v, const std::string &lbl="ANYM")
 Using the first column of a multivector, take the elementwise build a inverse and build the inverse diagonal operator.
double Teko::computeSpectralRad (const Teuchos::RCP< const Thyra::LinearOpBase< double > > &A, double tol, bool isHermitian=false, int numBlocks=5, int restart=0, int verbosity=0)
 Compute the spectral radius of a matrix.
double Teko::computeSmallestMagEig (const Teuchos::RCP< const Thyra::LinearOpBase< double > > &A, double tol, bool isHermitian=false, int numBlocks=5, int restart=0, int verbosity=0)
 Compute the smallest eigenvalue of an operator.
ModifiableLinearOp Teko::getDiagonalOp (const Teko::LinearOp &A, const DiagonalType &dt)
ModifiableLinearOp Teko::getInvDiagonalOp (const Teko::LinearOp &A, const Teko::DiagonalType &dt)
const MultiVector Teko::getDiagonal (const LinearOp &op, const DiagonalType &dt)
 Get the diagonal of a sparse linear operator.
std::string Teko::getDiagonalName (const DiagonalType &dt)
DiagonalType Teko::getDiagonalType (std::string name)
double Teko::norm_1 (const MultiVector &v, std::size_t col)
double Teko::norm_2 (const MultiVector &v, std::size_t col)

MultiVector utilities

MultiVector Teko::toMultiVector (BlockedMultiVector &bmv)
 Convert to a MultiVector from a BlockedMultiVector.
const MultiVector Teko::toMultiVector (const BlockedMultiVector &bmv)
 Convert to a MultiVector from a BlockedMultiVector.
const BlockedMultiVector Teko::toBlockedMultiVector (const MultiVector &bmv)
 Convert to a BlockedMultiVector from a MultiVector.
int Teko::blockCount (const BlockedMultiVector &bmv)
 Get the column count in a block linear operator.
MultiVector Teko::getBlock (int i, const BlockedMultiVector &bmv)
 Get the ith block from a BlockedMultiVector object.
MultiVector Teko::deepcopy (const MultiVector &v)
 Perform a deep copy of the vector.
MultiVector Teko::copyAndInit (const MultiVector &v, double scalar)
 Perform a deep copy of the vector.
BlockedMultiVector Teko::deepcopy (const BlockedMultiVector &v)
 Perform a deep copy of the blocked vector.
MultiVector Teko::datacopy (const MultiVector &src, MultiVector &dst)
 Copy the contents of a multivector to a destination vector.
BlockedMultiVector Teko::datacopy (const BlockedMultiVector &src, BlockedMultiVector &dst)
 Copy the contents of a blocked multivector to a destination vector.
BlockedMultiVector Teko::buildBlockedMultiVector (const std::vector< MultiVector > &mvv)
 build a BlockedMultiVector from a vector of MultiVectors

Detailed Description

This file contains a number of useful functions and classes used in Teko. They are distinct from the core functionality of the preconditioner factory, however, the functions are critical to construction of the preconditioners themselves.

Definition in file Teko_Utilities.hpp.

Enumeration Type Documentation

◆ DiagonalType

enum Teko::DiagonalType
Enumerator
Diagonal 

Specifies that just the diagonal is used.

Lumped 

Specifies that row sum is used to form a diagonal.

AbsRowSum 

Specifies that the $i^{th}$ diagonal entry is $\sum_j |A_{ij}|$.

BlkDiag 

Specifies that a block diagonal approximation is to be used.

NotDiag 

For user convenience, if Teko recieves this value, exceptions will be thrown.

Definition at line 809 of file Teko_Utilities.hpp.

Function Documentation

◆ buildGraphLaplacian() [1/2]

Teuchos::RCP< Tpetra::CrsMatrix< ST, LO, GO, NT > > Teko::buildGraphLaplacian ( int dim,
ST * coords,
const Tpetra::CrsMatrix< ST, LO, GO, NT > & stencil )

Build a graph Laplacian stenciled on a Epetra_CrsMatrix.

This function builds a graph Laplacian given a (locally complete) vector of coordinates and a stencil Epetra_CrsMatrix (could this be a graph of Epetra_RowMatrix instead?). The resulting matrix will have the negative of the inverse distance on off diagonals. And the sum of the positive inverse distance of the off diagonals on the diagonal. If there are no off diagonal entries in the stencil, the diagonal is set to 0.

Parameters
[in]dimNumber of physical dimensions (2D or 3D?).
[in]coordsA vector containing the coordinates, with the i-th coordinate beginning at coords[i*dim].
[in]stencilThe stencil matrix used to describe the connectivity of the graph Laplacian matrix.
Returns
The graph Laplacian matrix to be filled according to the stencil matrix.

Definition at line 201 of file Teko_Utilities.cpp.

◆ buildGraphLaplacian() [2/2]

Teuchos::RCP< Tpetra::CrsMatrix< ST, LO, GO, NT > > Teko::buildGraphLaplacian ( ST * x,
ST * y,
ST * z,
GO stride,
const Tpetra::CrsMatrix< ST, LO, GO, NT > & stencil )

Build a graph Laplacian stenciled on a Epetra_CrsMatrix.

This function builds a graph Laplacian given a (locally complete) vector of coordinates and a stencil Epetra_CrsMatrix (could this be a graph of Epetra_RowMatrix instead?). The resulting matrix will have the negative of the inverse distance on off diagonals. And the sum of the positive inverse distance of the off diagonals on the diagonal. If there are no off diagonal entries in the stencil, the diagonal is set to 0.

Parameters
[in]xA vector containing the x-coordinates, with the i-th coordinate beginning at coords[i*stride].
[in]yA vector containing the y-coordinates, with the i-th coordinate beginning at coords[i*stride].
[in]zA vector containing the z-coordinates, with the i-th coordinate beginning at coords[i*stride].
[in]strideStride between entries in the (x,y,z) coordinate array
[in]stencilThe stencil matrix used to describe the connectivity of the graph Laplacian matrix.
Returns
The graph Laplacian matrix to be filled according to the stencil matrix.

Definition at line 344 of file Teko_Utilities.cpp.

◆ getOutputStream()

const Teuchos::RCP< Teuchos::FancyOStream > Teko::getOutputStream ( )

Function used internally by Teko to find the output stream.

Function used internally by Teko to find the output stream.

Returns
An output stream to use for printing

Definition at line 94 of file Teko_Utilities.cpp.

◆ toMultiVector() [1/2]

MultiVector Teko::toMultiVector ( BlockedMultiVector & bmv)
inline

Convert to a MultiVector from a BlockedMultiVector.

Definition at line 192 of file Teko_Utilities.hpp.

◆ toMultiVector() [2/2]

const MultiVector Teko::toMultiVector ( const BlockedMultiVector & bmv)
inline

Convert to a MultiVector from a BlockedMultiVector.

Definition at line 195 of file Teko_Utilities.hpp.

◆ toBlockedMultiVector()

const BlockedMultiVector Teko::toBlockedMultiVector ( const MultiVector & bmv)
inline

Convert to a BlockedMultiVector from a MultiVector.

Definition at line 198 of file Teko_Utilities.hpp.

◆ blockCount()

int Teko::blockCount ( const BlockedMultiVector & bmv)
inline

Get the column count in a block linear operator.

Definition at line 203 of file Teko_Utilities.hpp.

◆ getBlock()

MultiVector Teko::getBlock ( int i,
const BlockedMultiVector & bmv )
inline

Get the ith block from a BlockedMultiVector object.

Definition at line 206 of file Teko_Utilities.hpp.

◆ deepcopy() [1/2]

MultiVector Teko::deepcopy ( const MultiVector & v)
inline

Perform a deep copy of the vector.

Definition at line 211 of file Teko_Utilities.hpp.

◆ copyAndInit()

MultiVector Teko::copyAndInit ( const MultiVector & v,
double scalar )
inline

Perform a deep copy of the vector.

Definition at line 214 of file Teko_Utilities.hpp.

◆ deepcopy() [2/2]

BlockedMultiVector Teko::deepcopy ( const BlockedMultiVector & v)
inline

Perform a deep copy of the blocked vector.

Definition at line 221 of file Teko_Utilities.hpp.

◆ datacopy() [1/2]

MultiVector Teko::datacopy ( const MultiVector & src,
MultiVector & dst )
inline

Copy the contents of a multivector to a destination vector.

Copy the contents of a multivector to a new vector. If the destination vector is null, a deep copy of the source multivector is made to a newly allocated vector. Also, if the destination and the source do not match, a new destination object is allocated and returned to the user.

Parameters
[in]srcSource multivector to be copied.
[in]dstDestination multivector. If null a new multivector will be allocated.
Returns
A copy of the source multivector. If dst is not null a pointer to this object is returned. Otherwise a new multivector is returned.

Definition at line 238 of file Teko_Utilities.hpp.

◆ datacopy() [2/2]

BlockedMultiVector Teko::datacopy ( const BlockedMultiVector & src,
BlockedMultiVector & dst )
inline

Copy the contents of a blocked multivector to a destination vector.

Copy the contents of a blocked multivector to a new vector. If the destination vector is null, a deep copy of the source multivector is made to a newly allocated vector. Also, if the destination and the source do not match, a new destination object is allocated and returned to the user.

Parameters
[in]srcSource multivector to be copied.
[in]dstDestination multivector. If null a new multivector will be allocated.
Returns
A copy of the source multivector. If dst is not null a pointer to this object is returned. Otherwise a new multivector is returned.

Definition at line 264 of file Teko_Utilities.hpp.

◆ buildBlockedMultiVector()

BlockedMultiVector Teko::buildBlockedMultiVector ( const std::vector< MultiVector > & mvv)

build a BlockedMultiVector from a vector of MultiVectors

Definition at line 2444 of file Teko_Utilities.cpp.

◆ beginBlockFill()

void Teko::beginBlockFill ( BlockedLinearOp & blo)
inline

Let the blocked operator know that you are going to set the sub blocks.

Let the blocked operator know that you are going to set the sub blocks. This is a simple wrapper around the member function of the same name in Thyra.

Parameters
[in,out]bloBlocked operator to have its fill stage activated

Definition at line 393 of file Teko_Utilities.hpp.

◆ getUpperTriBlocks()

*Get the strictly upper triangular portion of the matrix BlockedLinearOp Teko::getUpperTriBlocks ( const BlockedLinearOp & blo,
bool callEndBlockFill )

Get the strictly upper triangular portion of the matrix.

Definition at line 461 of file Teko_Utilities.cpp.

◆ getLowerTriBlocks()

*Get the strictly lower triangular portion of the matrix BlockedLinearOp Teko::getLowerTriBlocks ( const BlockedLinearOp & blo,
bool callEndBlockFill )

Get the strictly lower triangular portion of the matrix.

Definition at line 497 of file Teko_Utilities.cpp.

◆ zeroBlockedOp()

BlockedLinearOp Teko::zeroBlockedOp ( const BlockedLinearOp & blo)

Build a zero operator mimicing the block structure of the passed in matrix.

Build a zero operator mimicing the block structure of the passed in matrix. Currently this function assumes that the operator is "block" square. Also, this function calls beginBlockFill but does not call endBlockFill. This is so that the user can fill the matrix as they wish once created.

Parameters
[in]bloBlocked operator with desired structure.
Returns
A zero operator with the same block structure as the argument blo.

\notes The caller is responsible for calling endBlockFill on the returned blocked operator.

Build a zero operator mimicing the block structure of the passed in matrix. Currently this function assumes that the operator is "block" square. Also, this function calls beginBlockFill but does not call endBlockFill. This is so that the user can fill the matrix as they wish once created.

Parameters
[in]bloBlocked operator with desired structure.
Returns
A zero operator with the same block structure as the argument blo.
Note
The caller is responsible for calling endBlockFill on the returned blocked operator.

Definition at line 551 of file Teko_Utilities.cpp.

◆ getAbsRowSumMatrix()

ModifiableLinearOp Teko::getAbsRowSumMatrix ( const LinearOp & op)

Compute absolute row sum matrix.

Compute the absolute row sum matrix. That is a diagonal operator composed of the absolute value of the row sum.

Returns
A diagonal operator.

Definition at line 672 of file Teko_Utilities.cpp.

◆ getAbsRowSumInvMatrix()

ModifiableLinearOp Teko::getAbsRowSumInvMatrix ( const LinearOp & op)

Compute inverse of the absolute row sum matrix.

Compute the inverse of the absolute row sum matrix. That is a diagonal operator composed of the inverse of the absolute value of the row sum.

Returns
A diagonal operator.

Definition at line 713 of file Teko_Utilities.cpp.

◆ getLumpedMatrix()

ModifiableLinearOp Teko::getLumpedMatrix ( const LinearOp & op)

Compute the lumped version of this matrix.

Compute the lumped version of this matrix. That is a diagonal operator composed of the row sum.

Returns
A diagonal operator.

Definition at line 807 of file Teko_Utilities.cpp.

◆ getInvLumpedMatrix()

ModifiableLinearOp Teko::getInvLumpedMatrix ( const LinearOp & op)

Compute the inverse of the lumped version of this matrix.

Compute the inverse of the lumped version of this matrix. That is a diagonal operator composed of the row sum.

Returns
A diagonal operator.

Definition at line 829 of file Teko_Utilities.cpp.

◆ applyOp() [1/2]

*name Mathematical functions void Teko::applyOp ( const LinearOp & A,
const MultiVector & x,
MultiVector & y,
double alpha,
double beta )

Apply a linear operator to a multivector (think of this as a matrix vector multiply).

Apply a linear operator to a multivector. This also permits arbitrary scaling and addition of the result. This function gives

$ y = \alpha A x + \beta y $

Parameters
[in]A
[in]x
[in,out]y
[in]

alpha

Parameters
[in]

beta

Apply a linear operator to a multivector. This also permits arbitrary scaling and addition of the result. This function gives

$ y = \alpha A x + \beta y $

It is required that the range space of A is compatible with y and the domain space of A is compatible with x.

Parameters
[in]A
[in]x
[in,out]y
[in]alpha
[in]beta

Definition at line 422 of file Teko_Utilities.cpp.

◆ applyTransposeOp() [1/2]

void Teko::applyTransposeOp ( const LinearOp & A,
const MultiVector & x,
MultiVector & y,
double alpha,
double beta )

Apply a transposed linear operator to a multivector (think of this as a matrix vector multiply).

Apply a transposed linear operator to a multivector. This also permits arbitrary scaling and addition of the result. This function gives

$ y = \alpha A^T x + \beta y $

Parameters
[in]A
[in]x
[in,out]y
[in]

alpha

Parameters
[in]

beta

Apply a transposed linear operator to a multivector. This also permits arbitrary scaling and addition of the result. This function gives

$ y = \alpha A^T x + \beta y $

It is required that the domain space of A is compatible with y and the range space of A is compatible with x.

Parameters
[in]A
[in]x
[in,out]y
[in]alpha
[in]beta

Definition at line 441 of file Teko_Utilities.cpp.

◆ applyOp() [2/2]

void Teko::applyOp ( const LinearOp & A,
const BlockedMultiVector & x,
BlockedMultiVector & y,
double alpha = 1.0,
double beta = 0.0 )
inline

Apply a linear operator to a blocked multivector (think of this as a matrix vector multiply).

Apply a linear operator to a blocked multivector. This also permits arbitrary scaling and addition of the result. This function gives

$ y = \alpha A x + \beta y $

It is required that the range space of A is compatible with y and the domain space of A is compatible with x.

Parameters
[in]A
[in]x
[in,out]y
[in]alpha
[in]beta

Definition at line 532 of file Teko_Utilities.hpp.

◆ applyTransposeOp() [2/2]

void Teko::applyTransposeOp ( const LinearOp & A,
const BlockedMultiVector & x,
BlockedMultiVector & y,
double alpha = 1.0,
double beta = 0.0 )
inline

Apply a transposed linear operator to a blocked multivector (think of this as a matrix vector multiply).

Apply a transposed linear operator to a blocked multivector. This also permits arbitrary scaling and addition of the result. This function gives

$ y = \alpha A^T x + \beta y $

It is required that the domain space of A is compatible with y and the range space of A is compatible with x.

Parameters
[in]A
[in]x
[in,out]y
[in]alpha
[in]beta

Definition at line 557 of file Teko_Utilities.hpp.

◆ update()

void Teko::update ( double alpha,
const MultiVector & x,
double beta,
MultiVector & y )

Update the y vector so that $y = \alpha x+\beta y$.

Compute the linear combination $y=\alpha x + \beta y$.

Parameters
[in]alpha
[in]x
[in]beta
[in,out]y

Definition at line 448 of file Teko_Utilities.cpp.

◆ getDiagonalOp() [1/2]

*name Epetra_Operator specific functions const ModifiableLinearOp Teko::getDiagonalOp ( const LinearOp & op)

Get the diaonal of a linear operator.

Get the diagonal of a linear operator. Currently it is assumed that the underlying operator is an Epetra_RowMatrix.

Parameters
[in]opThe operator whose diagonal is to be extracted.
Returns
An diagonal operator.

Definition at line 900 of file Teko_Utilities.cpp.

◆ getDiagonal() [1/2]

const MultiVector Teko::getDiagonal ( const LinearOp & op)

Get the diagonal of a linear operator.

Get the diagonal of a linear operator, putting it in the first column of a multivector.

Definition at line 932 of file Teko_Utilities.cpp.

◆ getInvDiagonalOp() [1/2]

const ModifiableLinearOp Teko::getInvDiagonalOp ( const LinearOp & op)

Get the diaonal of a linear operator.

Get the inverse of the diagonal of a linear operator. Currently it is assumed that the underlying operator is an Epetra_RowMatrix.

Parameters
[in]opThe operator whose diagonal is to be extracted and inverted
Returns
An diagonal operator.

Definition at line 986 of file Teko_Utilities.cpp.

◆ explicitMultiply() [1/4]

const LinearOp Teko::explicitMultiply ( const LinearOp & opl,
const LinearOp & opm,
const LinearOp & opr )

Multiply three linear operators.

Multiply three linear operators. This currently assumes that the underlying implementation uses Epetra_CrsMatrix. The exception is that opm is allowed to be an diagonal matrix.

Parameters
[in]oplLeft operator (assumed to be a Epetra_CrsMatrix)
[in]opmMiddle operator (assumed to be a diagonal matrix)
[in]oprRight operator (assumed to be a Epetra_CrsMatrix)
Returns
Matrix product with a Epetra_CrsMatrix implementation

Multiply three linear operators. This currently assumes that the underlying implementation uses Epetra_CrsMatrix. The exception is that opm is allowed to be an diagonal matrix.

Parameters
[in]oplLeft operator (assumed to be a Epetra_CrsMatrix)
[in]opmMiddle operator (assumed to be a Epetra_CrsMatrix or a diagonal matrix)
[in]oprRight operator (assumed to be a Epetra_CrsMatrix)
Returns
Matrix product with a Epetra_CrsMatrix implementation

Definition at line 1070 of file Teko_Utilities.cpp.

◆ explicitMultiply() [2/4]

const ModifiableLinearOp Teko::explicitMultiply ( const LinearOp & opl,
const LinearOp & opm,
const LinearOp & opr,
const ModifiableLinearOp & destOp )

Multiply three linear operators.

Multiply three linear operators. This currently assumes that the underlying implementation uses Epetra_CrsMatrix. The exception is that opm is allowed to be an diagonal matrix.

Parameters
[in]oplLeft operator (assumed to be a Epetra_CrsMatrix)
[in]opmMiddle operator (assumed to be a Epetra_CrsMatrix or a diagonal matrix)
[in]oprRight operator (assumed to be a Epetra_CrsMatrix)
[in,out]destOpThe operator to be used as the destination operator, if this is null this function creates a new operator
Returns
Matrix product with a Epetra_CrsMatrix implementation

Definition at line 1321 of file Teko_Utilities.cpp.

◆ explicitMultiply() [3/4]

const LinearOp Teko::explicitMultiply ( const LinearOp & opl,
const LinearOp & opr )

Multiply two linear operators.

Multiply two linear operators. This currently assumes that the underlying implementation uses Epetra_CrsMatrix.

Parameters
[in]oplLeft operator (assumed to be a Epetra_CrsMatrix)
[in]oprRight operator (assumed to be a Epetra_CrsMatrix)
Returns
Matrix product with a Epetra_CrsMatrix implementation

Definition at line 1452 of file Teko_Utilities.cpp.

◆ explicitMultiply() [4/4]

const ModifiableLinearOp Teko::explicitMultiply ( const LinearOp & opl,
const LinearOp & opr,
const ModifiableLinearOp & destOp )

Multiply two linear operators.

Multiply two linear operators. This currently assumes that the underlying implementation uses Epetra_CrsMatrix. The exception is that opm is allowed to be an diagonal matrix.

Parameters
[in]oplLeft operator (assumed to be a Epetra_CrsMatrix)
[in]oprRight operator (assumed to be a Epetra_CrsMatrix)
[in,out]destOpThe operator to be used as the destination operator, if this is null this function creates a new operator
Returns
Matrix product with a Epetra_CrsMatrix implementation

Definition at line 1687 of file Teko_Utilities.cpp.

◆ explicitAdd() [1/2]

const LinearOp Teko::explicitAdd ( const LinearOp & opl_in,
const LinearOp & opr_in )

Add two linear operators.

Add two linear operators. This currently assumes that the underlying implementation uses Epetra_CrsMatrix.

Parameters
[in]oplLeft operator (assumed to be a Epetra_CrsMatrix)
[in]oprRight operator (assumed to be a Epetra_CrsMatrix)
Returns
Matrix sum with a Epetra_CrsMatrix implementation

Definition at line 1927 of file Teko_Utilities.cpp.

◆ explicitAdd() [2/2]

const ModifiableLinearOp Teko::explicitAdd ( const LinearOp & opl_in,
const LinearOp & opr_in,
const ModifiableLinearOp & destOp )

Add two linear operators.

Add two linear operators. This currently assumes that the underlying implementation uses Epetra_CrsMatrix.

Parameters
[in]oplLeft operator (assumed to be a Epetra_CrsMatrix)
[in]oprRight operator (assumed to be a Epetra_CrsMatrix)
[in,out]destOpThe operator to be used as the destination operator, if this is null this function creates a new operator
Returns
Matrix sum with a Epetra_CrsMatrix implementation

Definition at line 2075 of file Teko_Utilities.cpp.

◆ explicitSum()

const ModifiableLinearOp Teko::explicitSum ( const LinearOp & op,
const ModifiableLinearOp & destOp )

Sum an operator.

Returns
Matrix sum with a Epetra_CrsMatrix implementation

Sum into the modifiable linear op.

Definition at line 2239 of file Teko_Utilities.cpp.

◆ explicitTranspose()

const LinearOp Teko::explicitTranspose ( const LinearOp & op)

Build an explicit transpose of a linear operator. (Concrete data underneath.

Definition at line 2264 of file Teko_Utilities.cpp.

◆ explicitScale()

const LinearOp Teko::explicitScale ( double scalar,
const LinearOp & op )

Explicitely scale a linear operator.

Definition at line 2307 of file Teko_Utilities.cpp.

◆ frobeniusNorm()

double Teko::frobeniusNorm ( const LinearOp & op)

Rturn the frobenius norm of a linear operator

Definition at line 2337 of file Teko_Utilities.cpp.

◆ buildDiagonal()

const LinearOp Teko::buildDiagonal ( const MultiVector & src,
const std::string & lbl )

Take the first column of a multivector and build a diagonal linear operator.

Definition at line 2428 of file Teko_Utilities.cpp.

◆ buildInvDiagonal()

const LinearOp Teko::buildInvDiagonal ( const MultiVector & src,
const std::string & lbl )

Using the first column of a multivector, take the elementwise build a inverse and build the inverse diagonal operator.

Definition at line 2435 of file Teko_Utilities.cpp.

◆ computeSpectralRad()

double Teko::computeSpectralRad ( const Teuchos::RCP< const Thyra::LinearOpBase< double > > & A,
double tol,
bool isHermitian = false,
int numBlocks = 5,
int restart = 0,
int verbosity = 0 )

Compute the spectral radius of a matrix.

Compute the spectral radius of matrix A. This utilizes the Trilinos-Anasazi BlockKrylovShcur method for computing eigenvalues. It attempts to compute the largest (in magnitude) eigenvalue to a given level of tolerance.

Parameters
[in]Amatrix whose spectral radius is needed
[in]tolThe most accuracy needed (the algorithm will run until it reaches this level of accuracy and then it will quit). If this level is not reached it will return something to indicate it has not converged.
[in]isHermitianIs the matrix Hermitian
[in]numBlocksThe size of the orthogonal basis built (like in GMRES) before restarting. Increase the memory usage by O(restart*n). At least restart=3 is required.
[in]restartHow many restarts are permitted
[in]verbositySee the Anasazi documentation
Returns
The spectral radius of the matrix. If the algorithm didn't converge the number is the negative of the ritz-values. If a NaN is returned there was a problem constructing the Anasazi problem

◆ computeSmallestMagEig()

double Teko::computeSmallestMagEig ( const Teuchos::RCP< const Thyra::LinearOpBase< double > > & A,
double tol,
bool isHermitian = false,
int numBlocks = 5,
int restart = 0,
int verbosity = 0 )

Compute the smallest eigenvalue of an operator.

Compute the smallest eigenvalue of matrix A. This utilizes the Trilinos-Anasazi BlockKrylovShcur method for computing eigenvalues. It attempts to compute the smallest (in magnitude) eigenvalue to a given level of tolerance.

Parameters
[in]Amatrix whose spectral radius is needed
[in]tolThe most accuracy needed (the algorithm will run until it reaches this level of accuracy and then it will quit). If this level is not reached it will return something to indicate it has not converged.
[in]isHermitianIs the matrix Hermitian
[in]numBlocksThe size of the orthogonal basis built (like in GMRES) before restarting. Increase the memory usage by O(restart*n). At least restart=3 is required.
[in]restartHow many restarts are permitted
[in]verbositySee the Anasazi documentation
Returns
The smallest magnitude eigenvalue of the matrix. If the algorithm didn't converge the number is the negative of the ritz-values. If a NaN is returned there was a problem constructing the Anasazi problem

◆ getDiagonalOp() [2/2]

ModifiableLinearOp Teko::getDiagonalOp ( const Teko::LinearOp & A,
const DiagonalType & dt )

Get a diagonal operator from a matrix. The mechanism for computing the diagonal is specified by a DiagonalType arugment.

Parameters
[in]AEpetra_CrsMatrix to extract the diagonal from.
[in]dtSpecifies the type of diagonal that is desired.
Returns
A diagonal operator.

Definition at line 2664 of file Teko_Utilities.cpp.

◆ getInvDiagonalOp() [2/2]

ModifiableLinearOp Teko::getInvDiagonalOp ( const Teko::LinearOp & A,
const Teko::DiagonalType & dt )

Get the inverse of a diagonal operator from a matrix. The mechanism for computing the diagonal is specified by a DiagonalType arugment.

Parameters
[in]AEpetra_CrsMatrix to extract the diagonal from.
[in]dtSpecifies the type of diagonal that is desired.
Returns
A inverse of a diagonal operator.

Definition at line 2684 of file Teko_Utilities.cpp.

◆ getDiagonal() [2/2]

const MultiVector Teko::getDiagonal ( const LinearOp & op,
const DiagonalType & dt )

Get the diagonal of a sparse linear operator.

Parameters
[in]OpSparse linear operator to get diagonal of
[in]dtType of diagonal operator required.

Definition at line 967 of file Teko_Utilities.cpp.

◆ getDiagonalName()

std::string Teko::getDiagonalName ( const DiagonalType & dt)

Get a string corresponding to the type of diagonal specified.

Parameters
[in]dtThe type of diagonal.
Returns
A string name representing this diagonal type.

Get a string corresponding to the type of digaonal specified.

Parameters
[in]dtThe type of diagonal.
Returns
A string name representing this diagonal type.

Definition at line 2702 of file Teko_Utilities.cpp.

◆ getDiagonalType()

DiagonalType Teko::getDiagonalType ( std::string name)

Get a type corresponding to the name of a diagonal specified.

Parameters
[in]nameString representing the diagonal type
Returns
The type representation of the string, if the string is not recognized this function returns a NotDiag

Definition at line 2722 of file Teko_Utilities.cpp.

◆ norm_1()

double Teko::norm_1 ( const MultiVector & v,
std::size_t col )

Get the one norm of the vector

Definition at line 2748 of file Teko_Utilities.cpp.

◆ norm_2()

double Teko::norm_2 ( const MultiVector & v,
std::size_t col )

Get the two norm of the vector

Definition at line 2755 of file Teko_Utilities.cpp.

Generated on for Teko by doxygen 1.15.0