MLAPI Namespace Reference

MLAPI: Default namespace for all MLAPI objects and functions. More...

Classes
class	MLAPI::BaseLinearCombination
class	MLAPI::BaseObject
	Basic class for MLAPI objects. More...
class	MLAPI::BaseOperator
	Base class for all MLAPI objects. More...
class	MLAPI::CompObject
	Class to count flops. More...
class	MLAPI::DistributedMatrix
class	MLAPI::EpetraBaseOperator
	Basic class to wrap MLAPI::InverseOperator into Epetra_Operator. More...
struct	MLAPI::StackEntry
class	MLAPI::InverseOperator
	InverseOperator: basic class to define smoother and coarse solvers. More...
class	MLAPI::LinearCombinationAdd
class	MLAPI::LinearCombinationMixed
class	MLAPI::LinearCombinationScaled
class	MLAPI::MultiVectorScaled
class	MLAPI::MultiVectorCombination
class	MLAPI::BaseOperatorTimesMultiVector
class	MLAPI::Residual
class	MLAPI::MATLABStream
	Basic stream to save in a MATLAB-compatible file MLAPI objects. More...
class	MLAPI::MultiLevelAdaptiveSA
	Black-box multilevel adaptive smoothed aggregation preconditioner. More...
class	MLAPI::MultiLevelSA
	Black-box multilevel smoothed aggregation preconditioner. More...
class	MLAPI::DoubleVector
class	MLAPI::MultiVector
	Basic class for distributed double-precision vectors. More...
class	MLAPI::Operator
	Operator: basic class to define operators within MLAPI. More...
class	MLAPI::ML_Operator_Box
	Simple wrapper for ML_Operator struct. More...
class	MLAPI::Epetra_SerialMatrix
class	MLAPI::SerialMatrix
class	MLAPI::Space
	Specifies the number and distribution among processes of elements. More...
class	MLAPI::TimeObject
	Class to track time spent in an object. More...
Typedefs
typedef MLAPI::StackEntry	Entry
Functions
void	GetPtent (const Operator &A, Teuchos::ParameterList &List, const MultiVector &ThisNS, Operator &Ptent, MultiVector &NextNS)
	Builds the tentative prolongator using aggregation.
void	GetPtent (const Operator &A, Teuchos::ParameterList &List, Operator &Ptent)
	Builds the tentative prolongator with default null space.
void	GetPtent (const Epetra_RowMatrix &A, Teuchos::ParameterList &List, double *thisns, Teuchos::RCP< Epetra_CrsMatrix > &Ptent, Teuchos::RCP< Epetra_MultiVector > &NextNS, const int domainoffset=0)
	Builds the tentative prolongator using aggregation.
void	GetPtent (const Epetra_RowMatrix &A, Teuchos::ParameterList &List, double *thisns, Teuchos::RCP< Epetra_CrsMatrix > &Ptent, const int domainoffset=0)
	Builds the tentative prolongator using aggregation.
int	GetAggregates (Epetra_RowMatrix &A, Teuchos::ParameterList &List, double *thisns, Epetra_IntVector &aggrinfo)
	Call ML aggregation on A according to parameters supplied in List. Return aggregates in aggrinfo.
int	GetGlobalAggregates (Epetra_RowMatrix &A, Teuchos::ParameterList &List, double *thisns, Epetra_IntVector &aggrinfo)
	Call ML aggregation on A according to parameters supplied in List. Return aggregates in aggrinfo.
std::ostream &	operator<< (std::ostream &os, const BaseObject &obj)
void	SetDefaults (Teuchos::ParameterList &List)
	Sets default values in input List.
double	MaxEigAnorm (const Operator &Op, const bool DiagonalScaling=false)
	Computes the maximum eigenvalue of Op using the A-norm of the operator.
double	MaxEigCG (const Operator &Op, const bool DiagonalScaling=false)
	Computes the maximum eigenvalue of Op using the CG method.
double	MaxEigPowerMethod (const Operator &Op, const bool DiagonalScaling=false)
	Computes the maximum eigenvalue of Op using the power method.
double	MaxEigAnasazi (const Operator &Op, const bool DiagonalScaling=false)
	Computes the maximum eigenvalue of Op using Anasazi.
void	Eig (const Operator &Op, MultiVector &ER, MultiVector &EI)
	Computes eigenvalues and eigenvectors using LAPACK (w/ one process only).
void	Eigs (const Operator &A, int NumEigenvalues, MultiVector &ER, MultiVector &EI)
void	StackPop ()
void	StackPrint ()
MultiVectorCombination	operator+ (const MultiVector &x, const MultiVector &y)
	Creates a new MultiVector, defined as x + y.
LinearCombinationAdd	operator+ (const BaseLinearCombination &left, const BaseLinearCombination &right)
LinearCombinationMixed	operator+ (const BaseLinearCombination &left, const MultiVector &right)
LinearCombinationMixed	operator+ (const MultiVector &left, const BaseLinearCombination &right)
MultiVectorCombination	operator+ (const MultiVectorScaled &left, const MultiVectorScaled &right)
	Creates a new MultiVector, defined as alpha * x + beta * y.
Residual	operator+ (const MultiVectorScaled &left, const BaseOperatorTimesMultiVector &right)
	Creates a new MultiVector, defined as alpha * x + A * y.
Residual	operator+ (const MultiVector &left, const BaseOperatorTimesMultiVector &right)
	Creates a new MultiVector, defined as x + A * y.
MultiVectorCombination	operator- (const MultiVector &x, const MultiVector &y)
	Creates a new MultiVector, defined as x - y.
LinearCombinationAdd	operator- (const BaseLinearCombination &left, const BaseLinearCombination &right)
LinearCombinationMixed	operator- (const BaseLinearCombination &left, const MultiVector &right)
LinearCombinationMixed	operator- (const MultiVector &left, const BaseLinearCombination &right)
Residual	operator- (const MultiVector &left, const BaseOperatorTimesMultiVector &right)
	Creates a new MultiVector, defined as x - A * y.
MultiVector	operator+ (const MultiVector &x, const double alpha)
	Creates a new MultiVector, defined as x + alpha.
MultiVector	operator+ (const double alpha, const MultiVector &x)
	Creates a new MultiVector, defined as alpha + x.
MultiVector	operator- (const MultiVector &x, const double alpha)
	Creates a new MultiVector, defined as x - alpha.
MultiVector	operator- (const double alpha, const MultiVector &x)
	Creates a new MultiVector, defined as alpha - y.
Operator	operator+ (const Operator &A, const Operator &B)
	Creates a new Operator, defined as A + B.
Operator	operator- (const Operator &A, const Operator &B)
	Creates a new Operator, defined as A - B.
Operator	operator * (const Operator &A, const Operator &B)
	Creates a new Operator, defined as A * B.
Operator	operator * (const Operator &A, const double alpha)
	Creates a new Operator, defined as A * alpha.
Operator	operator * (const double alpha, const Operator &A)
	Creates a new Operator, defined as alpha * A.
Operator	operator/ (const Operator &A, const double alpha)
	Creates a new Operator, defined as A / alpha.
MultiVector	operator * (const MultiVector &x, const double alpha)
	Creates a new MultiVector, defined as x * alpha.
MultiVector	operator * (const double alpha, const MultiVector &x)
MultiVector	operator/ (const MultiVector &x, const double alpha)
	Creates a new MultiVector y, such that y = x / alpha.
BaseOperatorTimesMultiVector	operator * (const BaseOperator &A, const MultiVector &x)
	Creates a new MultiVector y, such that y = A * x.
BaseOperatorTimesMultiVector	operator * (const BaseOperator &A, const BaseLinearCombination &x)
	Creates a new MultiVector y, such that y = A * x (x is a BaseLinearCombination).
double	operator * (const MultiVector &x, const MultiVector &y)
	Computes the dot product between the first vector in x and y.
double	operator * (const MultiVector &x, const BaseLinearCombination &y)
double	operator * (const BaseLinearCombination &x, const MultiVector &y)
double	operator * (const BaseLinearCombination &x, const BaseLinearCombination &y)
Operator	Gallery (const string ProblemType, const Space &MySpace)
	Creates a matrix using the TRIUTILS gallery.
Operator	GetShiftedLaplacian1D (const int NX, const double Factor=0.99)
	Creates a 1D shifted Laplacian.
Operator	GetShiftedLaplacian2D (const int NX, const int NY, const double Factor=0.99, const bool RandomScale=false)
	Creates a 2D shifted Laplacian.
Operator	ReadMatrix (const char *FileName)
	Reads a matrix in MATLAB format.
Operator	GetRecirc2D (const int NX, const int NY, const double conv, const double diff)
	Creates a recirculation problem in 2D.
Teuchos::ParameterList	ReadParameterList (const char *FileName)
	Populates a list from specified file.
void	Krylov (const Operator &A, const MultiVector &LHS, const MultiVector &RHS, const BaseOperator &Prec, Teuchos::ParameterList &List)
MultiVector	Duplicate (const MultiVector &y)
	Creates a new vector, x, such that x = y.
MultiVector	Duplicate (const MultiVector &y, const int v)
	Creates a new vector, x, such that x = y(:,v).
MultiVector	Extract (const MultiVector &y, const int v)
	Extracts a component from a vector.
MultiVector	Redistribute (const MultiVector &y, const int NumEquations)
	Redistributes the entry of a vector as a multivector.
Operator	GetRAP (const Operator &R, const Operator &A, const Operator &P)
	Performs a triple matrix-matrix product, res = R * A *P.
Operator	GetTranspose (const Operator &A, const bool byrow=true)
	Returns a newly created transpose of A.
Operator	GetIdentity (const Space &DomainSpace, const Space &RangeSpace)
	Returns the identity matrix.
MultiVector	GetDiagonal (const Operator &A)
	Returns a vector containing the diagonal elements of A.
MultiVector	GetDiagonal (const Operator &A, const int offset)
	Returns a vector containing the diagonal elements of A.
Operator	GetDiagonal (const MultiVector &D)
	Returns a newly created operator, containing D on the diagonal.
Operator	GetJacobiIterationOperator (const Operator &Amat, double Damping)
	Returns an operator defined as (I - Damping A).
Operator	GetPtent1D (const MultiVector &D, const int offset=0)
	Returns a newly created operator, containing D on the diagonal.
int	ML_Operator_Add2 (ML_Operator *A, ML_Operator *B, ML_Operator *C, int matrix_type, double scalarA, double scalarB)
void	AnalyzeCheap (const Operator &A)
	Performs a cheap analysis of the properties of the input operator.
void	PrintSparsity (const Operator &A, int NumPDEEquations=1)
	Prints on file the sparsity structure of input operator.
Operator	GetScaledOperator (const Operator &A, const double alpha)
	Multiply A by a double value, alpha.
Operator	Duplicate (const Operator &A)
	Duplicates a given operator.
void	ReadSAMISMatrix (const char *filen, Operator &A, int &NumPDEEqns)
	Reads symmetric matrix from SAMIS binary format.
void	ReadSAMISKernel (const char *myKerFileName, MultiVector &A, const int limKer=-1)
	Reads null space vectors from SAMIS binary format.
ML_Comm *	GetML_Comm ()
	Returns a pointer to the ML_Comm object defined on MPI_COMM_WORLD.
Epetra_Comm &	GetEpetra_Comm ()
	Returns a reference to the Epetra_Comm object defined on MPI_COMM_WORLD.
void	Barrier ()
	Calls Mpi_Barrier() if MPI is enabled.
int	GetMyPID ()
	Returns the ID of the calling process.
int	GetNumProcs ()
	Returns the total number of processes in the computation.
int	GetPrintLevel ()
	Retutns the level of output (always 0 if MyPID() != 0).
void	SetPrintLevel (int Level)
	Sets the level of output (from 0 to 10, 0 being verbose).
void	Init ()
	Initialize the MLAPI workspace.
void	Finalize ()
	Destroys the MLAPI workspace.
string	GetString (const int &x)
string	GetString (const double &x)
int	GetMatrixType ()

---

Detailed Description

MLAPI: Default namespace for all MLAPI objects and functions.

---

Function Documentation

int GetAggregates (  Epetra_RowMatrix &  A, Teuchos::ParameterList &  List, double *  thisns, Epetra_IntVector &  aggrinfo )

Call ML aggregation on A according to parameters supplied in List. Return aggregates in aggrinfo. On input, map of aggrinfo has to map row map of A. On output, aggrinfo[i] contains number of aggregate the row belongs to, where aggregates are numbered starting from 0. Return value is the processor-local number of aggregates build. If aggrinfo[i] >= return-value, then i is a processor local row of a row that ML has detected to be on a Dirichlet BC. Parameters: A  (in): Matrix to be aggregated on List  (in): ParameterList containing ML options thisns  (in): nullspace aggrinfo(out):  vector containing aggregation information Note: Map of aggrinfo has to match rowmap of A on input. Returns: returns processor-local number of aggregates Author: Michael Gee (gee@lnm.mw.tum.de)

int GetGlobalAggregates (  Epetra_RowMatrix &  A, Teuchos::ParameterList &  List, double *  thisns, Epetra_IntVector &  aggrinfo )

Call ML aggregation on A according to parameters supplied in List. Return aggregates in aggrinfo. On input, map of aggrinfo has to map row map of A. On output, aggrinfo[i] contains number of global aggregate the row belongs to, where aggregates are numbered starting from 0 globally. Return value is the processor-local number of aggregates build. If aggrinfo[i] < 0, then i is a processor local row that ML has detected to be on a Dirichlet BC. if aggrinfo[i] >= 0, then i is a processor local row and aggrinfo[i] is a global aggregate id. Parameters: A  (in): Matrix to be aggregated on List  (in): ParameterList containing ML options thisns  (in): nullspace aggrinfo(out):  vector containing aggregation information in global numbering Note: Map of aggrinfo has to match rowmap of A on input. Returns: returns processor-local number of aggregates Author: Michael Gee (gee@lnm.mw.tum.de)

void GetPtent (  const Epetra_RowMatrix &  A, Teuchos::ParameterList &  List, double *  thisns, Teuchos::RCP< Epetra_CrsMatrix > &  Ptent, const int  domainoffset = 0 )

Builds the tentative prolongator using aggregation. Build Ptent and NextNS as usual but as Epetra objects. Parameters: A  (in): Matrix to be aggregated on List  (in): ParameterList containing ML options thisns  (in): nullspace in format ML accepts Ptent(out):  Matrix containing tentative prolongator domainoffset  (in,optional): give an offset such that the domainmap of Ptent starts global numbering from domainoffset instead from zero. This is useful to create block matrices. Author: Michael Gee (gee@lnm.mw.tum.de)

void GetPtent (  const Epetra_RowMatrix &  A, Teuchos::ParameterList &  List, double *  thisns, Teuchos::RCP< Epetra_CrsMatrix > &  Ptent, Teuchos::RCP< Epetra_MultiVector > &  NextNS, const int  domainoffset = 0 )

Builds the tentative prolongator using aggregation. Build Ptent and NextNS as usual but as Epetra objects. Parameters: A  (in): Matrix to be aggregated on List  (in): ParameterList containing ML options thisns  (in): nullspace in format ML accepts Ptent(out):  Matrix containing tentative prolongator NextNS  (out): MultiVector containing next level nullspace. domainoffset  (in,optional): give an offset such that the domainmap of Ptent starts global numbering from domainoffset instead from zero. This is useful to create block matrices. Author: Michael Gee (gee@lnm.mw.tum.de)

---