Multilevel BDDC based linear system solver. More...

#include <bddcml_wrapper.hh>

Collaboration diagram for la::BddcmlWrapper:

[legend]

Public Types
enum	matrixTypeEnum { GENERAL, SPD, SYMMETRICGENERAL, SPD_VIA_SYMMETRICGENERAL }

typedef enum matrixTypeEnum	MatrixType

Public Member Functions
	BddcmlWrapper (const unsigned numDofs, const unsigned numDofsSub, const MatrixType matrixType=GENERAL, const MPI_Comm comm=MPI_COMM_WORLD, int numSubLoc=1)

	~BddcmlWrapper ()
	Destructor frees the BDDCML structures. More...

void	loadRawMesh (const int nDim, const int numNodes, const std::vector< int > &inet, const std::vector< int > &nnet, const std::vector< int > &nndf, const std::vector< int > &isegn, const std::vector< int > &isngn, const std::vector< int > &isvgvn, const std::vector< double > &xyz, const std::vector< double > &element_data, const int meshDim=0)
	Load raw data about mesh. More...

void	prepareMatAssembly (unsigned numElements, unsigned elMatSize)
	Prepare assembly of matrix - reserve space in underlying coordinate matrix object. More...

void	insertToMatrix (const SubMat_ &eStiff, const VecUint_ &rowIndices, const VecUint_ &colIndices)
	insert submatrix to system matrix - into coordinate matrix object More...

void	finishMatAssembly ()
	Finalize assembly of matrix. More...

void	insertToRhs (const SubVec_ &eForce, const VecUint_ &dofIndices)
	insert subvector to system rhs vector More...

void	writeMatrix (std::ostream &out)
	Outputs. More...

void	finishAssembly ()
	Compatibility call which does absolutely nothing. More...

void	applyConstraints (ConstraintVec_ &constraints, const double factor, const double scalar)

void	fixDOF (const unsigned index, const double scalar=1.)

void	insertToDiagonalWeights (const int &index, const double &value)

void	finishDiagAssembly ()
	Finalize assembly of diagonal. More...

void	solveSystem (double tol=1.e-7, int numLevels=2, std::vector< int > *numSubAtLevels=NULL, int verboseLevel=0, int maxIt=1000, int ndecrMax=30, bool use_adaptive=false)
	Solve the system. More...

double	normRhs ()
	Get norm of right-hand side. More...

double	normSol ()
	Get norm of solution. More...

int	giveNumIterations ()
	Give number of iteration. More...

int	giveConvergedReason ()
	Give reason of convergence. More...

double	giveCondNumber ()
	Give estimate of condition number. More...

void	clearMatrix ()
	Fill matrix with zeros. More...

void	clearRhs ()
	Fill RHS with zeros. More...

void	clearBC ()
	Blank arrays for boundary conditions. More...

void	clearSol ()
	Fill solution with zeros. More...

void	destroyMatrix ()
	Destroy matrix. More...

void	destroyRhs ()
	Destroy RHS. More...

void	destroyBC ()
	Destroy boundary conditions. More...

void	destroySol ()
	Destroy solution. More...

template<typename VEC1 , typename VEC2 >
void	giveSolution (const VEC1 &dofIndices, VEC2 &result) const
	Give access to the solution vector. More...

Private Types
typedef ublas::matrix< double >	SubMat_

typedef ublas::vector< double >	SubVec_

typedef std::vector< unsigned >	VecUint_

typedef std::vector< std::pair< unsigned, double > >	ConstraintVec_

typedef std::map< unsigned, unsigned >	Global2LocalMap_

Private Attributes
const int	numDofs_
	number of total dofs More...

const int	numDofsSub_
	number of subdomain dofs More...

const MatrixType	matrixType_
	type of matrix More...

const MPI_Comm	comm_
	communicator More...

int	numSubLoc_
	number of subdomains on process More...

int	nProc_
	number of processors in communicator More...

int	rank_
	ID of processors in communicator. More...

int	nDim_
	space dimension More...

int	meshDim_
	mesh dimension, may differ from space dimension - e.g. 2 for shells in 3D More...

int	numElemSub_
	number of elements in subdomain More...

int	numElem_
	total number of elements More...

int	numNodesSub_
	number of nodes in subdomain More...

int	numNodes_
	total number of nodes More...

bool	meshLoaded_

std::vector< int >	inet_
	safety state variable to warn if solver is called without loading subdomain mesh More...

std::vector< int >	nnet_

std::vector< int >	nndf_

std::vector< double >	xyz_

std::vector< double >	element_data_
	array with data on elements, here used for permeability for each element More...

std::vector< int >	isngn_
	Indices of Subdomain Nodes in Global Numbering. More...

std::vector< int >	isvgvn_
	Indices of Subdomain Variables in Global Variable Numbering ( i.e. dof mapping ) More...

std::vector< int >	isegn_
	Indices of Subdomain Elements in Global Numbering. More...

Global2LocalMap_	global2LocalDofMap_
	type for storage of global to local map More...

std::vector< int >	ifix_
	indices of fixed boundary conditions - ( 0 for free variable, 1 for constrained dof ) More...

std::vector< double >	fixv_
	values of fixed boundary conditions at places of ifix_ - values outside fixed dofs are ignored More...

la::MatrixCoo< int, double >	coo_
	matrix in coordinate format (COO) More...

bool	isMatAssembled_
	true if matrix is assembled More...

la::MatrixCoo< int, double >	diagWeightsCoo_
	diagonal of subdomain matrix in sparse format and global numbers More...

bool	isDiagAssembled_
	true if diagonal is assembled More...

std::vector< double >	rhsVec_
	vector with RHS values restricted to subdomain, i.e. values are repeated at shared nodes More...

double	normRhs_
	norm of the global right-hand side More...

std::vector< double >	sol_
	distributed solution vector More...

double	normSol_
	norm of the global solution More...

int	numIter_
	required number of iterations More...

int	convergedReason_

double	condNumber_

Detailed Description

Multilevel BDDC based linear system solver.

This class implements solution of a system of linear equations by multilevel BDDC implementation in BDDCML solver.

The class uses assembly of the matrix into the coordinate matrix format implemented in matrix_coo.hpp file.

prepareMatAssembly - (optional) - reserves memory for sparse coordinate matrix insertToMatrix - called many times, inserts values into coordinate matrix insertToRhs - called many times, inserts values into RHS. If matrix is created and assembled, structure of the vector is deduced from it. Otherwise, it is created as empty. applyConstraints - currently only passes list of boundary conditions into the BDDCML solver, rather then modifying the matrix and rhs

as a result, no nonsymmetry is forced to SPD and symmetric indefinite problems by constraints, and these can be accelerated and be more memory efficient by specifying proper matrixType solveSystem - solves the system

Definition at line 80 of file bddcml_wrapper.hh.

Member Typedef Documentation

typedef std::vector< std::pair<unsigned,double> > la::BddcmlWrapper::ConstraintVec_

private

Definition at line 96 of file bddcml_wrapper.hh.

typedef std::map<unsigned,unsigned> la::BddcmlWrapper::Global2LocalMap_

private

Definition at line 271 of file bddcml_wrapper.hh.

typedef enum matrixTypeEnum la::BddcmlWrapper::MatrixType

Definition at line 90 of file bddcml_wrapper.hh.

typedef ublas::matrix< double > la::BddcmlWrapper::SubMat_

private

Definition at line 93 of file bddcml_wrapper.hh.

typedef ublas::vector< double > la::BddcmlWrapper::SubVec_

private

Definition at line 94 of file bddcml_wrapper.hh.

typedef std::vector< unsigned > la::BddcmlWrapper::VecUint_

private

Definition at line 95 of file bddcml_wrapper.hh.

Member Enumeration Documentation

enum la::BddcmlWrapper::matrixTypeEnum

Enumerator
GENERAL	general (default),
SPD	symmetric positive definite,
SYMMETRICGENERAL	general symmetric ( e.g. saddle point ),
SPD_VIA_SYMMETRICGENERAL	(advanced) interface problem is symm. pos. def., but subdomain problems are general symmetric ( e.g. saddle point )

Definition at line 83 of file bddcml_wrapper.hh.

Constructor & Destructor Documentation

la::BddcmlWrapper::BddcmlWrapper	(	const unsigned	numDofs,
		const unsigned	numDofsSub,
		const MatrixType	matrixType = `GENERAL`,
		const MPI_Comm	comm = `MPI_COMM_WORLD`,
		int	numSubLoc = `1`
	)

Constructor with total number of dofs, subdomain number of dofs, (optional) type of matrix, (optional) communicator ( MPI_COMM_WORLD is the default ), (optional) number of subdomains ( if not given, use number of processes ),

Author: Jakub Sistek

Date: 7/2012 Constructor for the BDDCML solver object

Definition at line 9 of file bddcml_wrapper.cc.

la::BddcmlWrapper::~BddcmlWrapper ( )

Destructor frees the BDDCML structures.

Destructor frees the BDDC solver structures

Definition at line 46 of file bddcml_wrapper.cc.

Member Function Documentation

void la::BddcmlWrapper::applyConstraints	(	ConstraintVec_ &	constraints,
		const double	factor,
		const double	scalar
	)

Apply constraints Currently, scalar is not used in the function and it is kept only for compatibility with constraint functors. Proper scalar is enforced within the BDDC solver.

Apply boundary conditions Since BDDCML handles constraints, just fill proper BC arrays. Scalar is not used, it is kept for compatibility with constraint functors.

Definition at line 248 of file bddcml_wrapper.cc.

Here is the caller graph for this function:

void la::BddcmlWrapper::clearBC ( )

inline

Blank arrays for boundary conditions.

Definition at line 209 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::clearMatrix ( )

inline

Fill matrix with zeros.

Definition at line 196 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::clearRhs ( )

inline

Fill RHS with zeros.

Definition at line 203 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::clearSol ( )

inline

Fill solution with zeros.

Definition at line 215 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::destroyBC ( )

inline

Destroy boundary conditions.

Definition at line 227 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::destroyMatrix ( )

inline

Destroy matrix.

Definition at line 221 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::destroyRhs ( )

inline

Destroy RHS.

Definition at line 225 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::destroySol ( )

inline

Destroy solution.

Definition at line 232 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::finishAssembly ( )

inline

Compatibility call which does absolutely nothing.

Definition at line 144 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::finishDiagAssembly ( )

Finalize assembly of diagonal.

Routine for diagonal finalization.

Definition at line 302 of file bddcml_wrapper.cc.

Here is the caller graph for this function:

void la::BddcmlWrapper::finishMatAssembly ( )

Finalize assembly of matrix.

Routine for triplet finalization.

Definition at line 205 of file bddcml_wrapper.cc.

Here is the caller graph for this function:

void la::BddcmlWrapper::fixDOF	(	const unsigned	index,
		const double	scalar = `1.`
	)

Set any DOF to zero

Parameters

[in]	index	Global index of DOF to fix
[in]	scalar	Value to put on diagonal (default 1.)

Simply set the dof with index 'index' to zero

Definition at line 275 of file bddcml_wrapper.cc.

double la::BddcmlWrapper::giveCondNumber ( )

inline

Give estimate of condition number.

Definition at line 193 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::giveConvergedReason ( )

inline

Give reason of convergence.

Definition at line 190 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::giveNumIterations ( )

inline

Give number of iteration.

Definition at line 187 of file bddcml_wrapper.hh.

template<typename VEC1 , typename VEC2 >

void la::BddcmlWrapper::giveSolution	(	const VEC1 &	dofIndices,
		VEC2 &	result
	)		const

Give access to the solution vector.

Definition at line 303 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::insertToDiagonalWeights	(	const int &	index,
		const double &	value
	)

Insert a value to diagonal for creating weights in BDDC

Parameters

[in]	index	Global index of DOF
[in]	value	Value to put on diagonal

Insert elements on diagonal for weights.

Definition at line 287 of file bddcml_wrapper.cc.

void la::BddcmlWrapper::insertToMatrix	(	const SubMat_ &	subMat,
		const VecUint_ &	rowIndices,
		const VecUint_ &	colIndices
	)

insert submatrix to system matrix - into coordinate matrix object

Insert element stiffness matrix into global system matrix

Definition at line 182 of file bddcml_wrapper.cc.

void la::BddcmlWrapper::insertToRhs	(	const SubVec_ &	subVec,
		const VecUint_ &	dofIndices
	)

insert subvector to system rhs vector

Insert subvector to the system's RHS vector

Definition at line 221 of file bddcml_wrapper.cc.

void la::BddcmlWrapper::loadRawMesh	(	const int	nDim,
		const int	numNodes,
		const std::vector< int > &	inet,
		const std::vector< int > &	nnet,
		const std::vector< int > &	nndf,
		const std::vector< int > &	isegn,
		const std::vector< int > &	isngn,
		const std::vector< int > &	isvgvn,
		const std::vector< double > &	xyz,
		const std::vector< double > &	element_data,
		const int	meshDim = `0`
	)

Load raw data about mesh.

Routine for loading raw mesh data into the solver - for cases of strange meshes, where these are not Mesh or Grid objects, user can create own raw description to exploit the flexibility of mesh format underlaying BDDCML.

Definition at line 80 of file bddcml_wrapper.cc.

double la::BddcmlWrapper::normRhs ( )

inline

Get norm of right-hand side.

Definition at line 181 of file bddcml_wrapper.hh.

Here is the caller graph for this function:

double la::BddcmlWrapper::normSol ( )

inline

Get norm of solution.

Definition at line 184 of file bddcml_wrapper.hh.

void la::BddcmlWrapper::prepareMatAssembly	(	unsigned	numElements,
		unsigned	elMatSize
	)

Prepare assembly of matrix - reserve space in underlying coordinate matrix object.

Routine for preparing MATRIX assembly. It reserves memory in triplet.

Definition at line 172 of file bddcml_wrapper.cc.

void la::BddcmlWrapper::solveSystem	(	double	tol = `1.e-7`,
		int	numLevels = `2`,
		std::vector< int > *	numSubAtLevels = `NULL`,
		int	verboseLevel = `0`,
		int	maxIt = `1000`,
		int	ndecrMax = `30`,
		bool	use_adaptive = `false`
	)

Solve the system.

Solve the system by BDDCML

< number of subdomains on first level

< numbers of subdomains on levels ( should be monotonically decreasing, e.g. for 3 levels 1024, 32, 1 )

Parameters

tol	tolerance on relative residual \|\|res\|\|/\|\|rhs\|\|
numLevels	number of levels
numSubAtLevels	number of subdomains at levels
verboseLevel	level of verbosity of BDDCML library ( 0 - only fatal errors reported, 1 - mild output, 2 - detailed output )
maxIt	maximum number of iterations
ndecrMax	maximum number of iterations with non-decreasing residual ( used to stop diverging process )
use_adaptive	if adaptive constraints should be used by BDDCML

Definition at line 319 of file bddcml_wrapper.cc.

void la::BddcmlWrapper::writeMatrix ( std::ostream & out )

inline

Outputs.

Definition at line 141 of file bddcml_wrapper.hh.

Here is the caller graph for this function:

Member Data Documentation

const MPI_Comm la::BddcmlWrapper::comm_

private

communicator

Definition at line 242 of file bddcml_wrapper.hh.

double la::BddcmlWrapper::condNumber_

private

estimated condition number of the preconditioned operator only provided for SPD problems solved by PCG ( -1.0 if not meaningful )

Definition at line 294 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::convergedReason_

private

reason of convergence: 0 - converged relative residual, desirable state -1 - reached maximal number of iterations -2 - reached maximal number of iterations with non-decreasing residual

Definition at line 290 of file bddcml_wrapper.hh.

la::MatrixCoo<int,double> la::BddcmlWrapper::coo_

private

matrix in coordinate format (COO)

Definition at line 278 of file bddcml_wrapper.hh.

la::MatrixCoo<int,double> la::BddcmlWrapper::diagWeightsCoo_

private

diagonal of subdomain matrix in sparse format and global numbers

Definition at line 280 of file bddcml_wrapper.hh.

std::vector<double> la::BddcmlWrapper::element_data_

private

array with data on elements, here used for permeability for each element

Definition at line 265 of file bddcml_wrapper.hh.

std::vector<double> la::BddcmlWrapper::fixv_

private

values of fixed boundary conditions at places of ifix_ - values outside fixed dofs are ignored

Definition at line 276 of file bddcml_wrapper.hh.

Global2LocalMap_ la::BddcmlWrapper::global2LocalDofMap_

private

type for storage of global to local map

map from global dof indices to subdomain local indices

Definition at line 272 of file bddcml_wrapper.hh.

std::vector<int> la::BddcmlWrapper::ifix_

private

indices of fixed boundary conditions - ( 0 for free variable, 1 for constrained dof )

Definition at line 275 of file bddcml_wrapper.hh.

std::vector<int> la::BddcmlWrapper::inet_

private

safety state variable to warn if solver is called without loading subdomain mesh

indices of nodes on elements (linearized connectivity) - numbering w.r.t. local subdomain numbering

Definition at line 257 of file bddcml_wrapper.hh.

bool la::BddcmlWrapper::isDiagAssembled_

private

true if diagonal is assembled

Definition at line 281 of file bddcml_wrapper.hh.

std::vector<int> la::BddcmlWrapper::isegn_

private

Indices of Subdomain Elements in Global Numbering.

Definition at line 269 of file bddcml_wrapper.hh.

bool la::BddcmlWrapper::isMatAssembled_

private

true if matrix is assembled

Definition at line 279 of file bddcml_wrapper.hh.

std::vector<int> la::BddcmlWrapper::isngn_

private

Indices of Subdomain Nodes in Global Numbering.

Definition at line 267 of file bddcml_wrapper.hh.

std::vector<int> la::BddcmlWrapper::isvgvn_

private

Indices of Subdomain Variables in Global Variable Numbering ( i.e. dof mapping )

Definition at line 268 of file bddcml_wrapper.hh.

const MatrixType la::BddcmlWrapper::matrixType_

private

type of matrix

Definition at line 241 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::meshDim_

private

mesh dimension, may differ from space dimension - e.g. 2 for shells in 3D

Definition at line 249 of file bddcml_wrapper.hh.

bool la::BddcmlWrapper::meshLoaded_

private

Definition at line 256 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::nDim_

private

space dimension

Definition at line 248 of file bddcml_wrapper.hh.

std::vector<int> la::BddcmlWrapper::nndf_

private

number of nodal degrees of freedom ( entry for each node ), e.g. [ 3, 3, 3, ... ] for 3D solids

Definition at line 261 of file bddcml_wrapper.hh.

std::vector<int> la::BddcmlWrapper::nnet_

private

numbers of nodes on elements (determines chunks of nodes for each element in inet), e.g. [ 8, 8, 8, ... ] for linear hexahedra

Definition at line 258 of file bddcml_wrapper.hh.

double la::BddcmlWrapper::normRhs_

private

norm of the global right-hand side

Definition at line 284 of file bddcml_wrapper.hh.

double la::BddcmlWrapper::normSol_

private

norm of the global solution

Definition at line 287 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::nProc_

private

number of processors in communicator

Definition at line 245 of file bddcml_wrapper.hh.

const int la::BddcmlWrapper::numDofs_

private

number of total dofs

Definition at line 239 of file bddcml_wrapper.hh.

const int la::BddcmlWrapper::numDofsSub_

private

number of subdomain dofs

Definition at line 240 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::numElem_

private

total number of elements

Definition at line 252 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::numElemSub_

private

number of elements in subdomain

Definition at line 251 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::numIter_

private

required number of iterations

Definition at line 289 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::numNodes_

private

total number of nodes

Definition at line 254 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::numNodesSub_

private

number of nodes in subdomain

Definition at line 253 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::numSubLoc_

private

number of subdomains on process

Definition at line 243 of file bddcml_wrapper.hh.

int la::BddcmlWrapper::rank_

private

ID of processors in communicator.

Definition at line 246 of file bddcml_wrapper.hh.

std::vector<double> la::BddcmlWrapper::rhsVec_

private

vector with RHS values restricted to subdomain, i.e. values are repeated at shared nodes

Definition at line 283 of file bddcml_wrapper.hh.

std::vector<double> la::BddcmlWrapper::sol_

private

distributed solution vector

Definition at line 286 of file bddcml_wrapper.hh.

std::vector<double> la::BddcmlWrapper::xyz_

private

linearized array of coordinates - all x values, all y values, all z values ( if present ) i.e. [ x_0 x_1 x_2 ... x_(numNodes - 1) y_0 y_1 y_2 ... y_(numNodes - 1) z_0 z_1 z_2 ... z_(numNodes - 1) ]

Definition at line 263 of file bddcml_wrapper.hh.

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

/opt/flow123d/flow123d/src/la/bddcml_wrapper.hh
/opt/flow123d/flow123d/src/la/bddcml_wrapper.cc

Public Types

Public Member Functions

Private Types

Private Attributes

Detailed Description

Member Typedef Documentation

Member Enumeration Documentation

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation