aggregator Class Reference

#include <aggregator.h>

Public Member Functions
	aggregator (void)
void	assemble_aggr_unknowns (gtopology *gt)
void	assemble_lnagr ()
void	assemble_lnnagr ()
void	assemble_lnuagr_sp (long agrid, long *i)
void	assemble_lnuagr_sp (long *i)
void	assemble_lnuagr_tp (long *i)
void	assemble_smoothed_prol (long agrid, long cid, long i, double a)
void	assemble_smoothed_prol (gtopology gt, gmatrix mtx)
void	assemble_smoothed_prol2 (long agrid, long cid, long i, compvect cvi, compvect cvo, gmatrix mtx)
void	assemble_tentative_prol (long agrid, long cid, long i, double a)
void	boss (gmatrix gm, double u, double *v)
void	clean_memory ()
void	coarse_matrix (gmatrix *gm)
void	define_degree (long i)
void	define_degree ()
void	define_graph (long nnod, long nneigh, long *listneigh)
void	define_matsize (long i, long j)
void	define_na (long i)
void	define_na (long i)
	defines number of aggregates
void	gener_rbm_1 (gtopology gt, long agid, double v)
void	gener_rbm_2 (gtopology gt, long agid, long cnid, double v)
long	give_deg ()
long	give_deg ()
void	local_matrices (long bsize, gmatrix *gm)
void	matrix_assemble (long i, long j, double *v)
void	metis_aggr (gtopology *gt)
void	mulA (long st, double x, gmatrix mtx)
void	mulS (long st, double x, gmatrix mtx)
	lokalni metody - casem asi private
void	prepare_boss (gtopology gt, gmatrix gm, FILE *out)
void	prepare_lnagr ()
void	prepare_tlnagr (void)
void	print (FILE *out)
void	read (gtopology gt, XFILE in, long mespr)
	~aggregator (void)
Public Attributes
long **	adjnodnod
long *	ci
densemat *	cm
	coarse matrix
long	cms
	size of the coarse matrix
long	degree_k
long	degree_r
long	exinex
long	impl
comprow *	lmcr
densemat *	lmdm
skyline *	lmsky
long **	lnagr
long *	lnnagr
long *	lnntagr
	list of numbers of nodes in tentative aggregates (without overlap)
long **	lntagr
	list of node numbers in tentative aggregates (without overlap)
long *	lnuaggr
long **	luaggr
long	maxnu
	maximum number of unknowns in aggregate
long	n
	number of unknowns in the whole problem
long	na
	number of aggregates
long *	nadjnodnod
long	nn
	number of nodes in the finite element mesh
long	nrbm
	number of rigid body modes of one aggregate
long	nz
double *	p
long *	ri
ISolver *	sdirect
double	specrad
slesolv *	ssle
	data about solver of system of linear equations
long	trbm
	type of factorization of system of linear algebraic equations
double *	val

Detailed Description

class aggregator deals with preconditioning of solvers of systems of linear algebraic equations the preconditioner is based on the BOSS method (Black-box Overlapping Schwarz with Smoothed Coarse Space) the BOSS method was developed ba M. Brezina in his Ph.D. thesis

JK

Definition at line 20 of file aggregator.h.

Constructor & Destructor Documentation

aggregator ( void )

Definition at line 5 of file aggregator.cpp.

References cms, degree_k, degree_r, exinex, impl, lmcr, lmdm, lmsky, lnagr, lnnagr, lnntagr, lntagr, lnuaggr, luaggr, n, na, nn, nrbm, sdirect, specrad, ssle, and trbm.

~aggregator ( void )

Definition at line 57 of file aggregator.cpp.

References lmcr, lmdm, lmsky, lnuaggr, luaggr, and sdirect.

Member Function Documentation

void assemble_aggr_unknowns ( gtopology * gt )

function assembles list of unknowns belonging to the aggregates with overlap

function assembles array luaggr which contains unknow numbers in usual (not C) notation it means, it starts from 1 and not from 0

Parameters:

gt

- pointer to general topology

JK, 19.2.2007

Definition at line 745 of file aggregator.cpp.

References gtopology::give_dof(), gtopology::give_ndofn(), lnagr, lnnagr, lnuaggr, luaggr, maxnu, and na.

Referenced by prepare_boss().

void assemble_lnagr ( )

Referenced by smaggreg::localization_matrices().

void assemble_lnnagr ( )

Referenced by smaggreg::localization_matrices().

void assemble_lnuagr_sp	(	long	agrid,
		long *	i
	)

void assemble_lnuagr_sp ( long * i )

Referenced by smaggreg::smoothed_prolong().

void assemble_lnuagr_tp ( long * i )

void assemble_smoothed_prol	(	long	agrid,
		long	cid,
		long *	i,
		double *	a
	)

void assemble_smoothed_prol	(	gtopology *	gt,
		gmatrix *	mtx
	)

vytvori zhlazeny prolongator input: agrid . cislo pozadovaneho agregatu cid ... cislo pozadovaneho sloupce v agregatu ? i ... pole indexu ? a ... pole hodnot (v pripade rigid body motions v mechanice) mtx ... matice problemu PM

Definition at line 515 of file aggregator.cpp.

References copyv(), degree_k, gener_rbm_1(), gener_rbm_2(), mulS(), n, na, nrbm, p, print_err(), and trbm.

Referenced by prepare_boss().

void assemble_smoothed_prol2	(	long	agrid,
		long	cid,
		long *	i,
		compvect *	cvi,
		compvect *	cvo,
		gmatrix *	mtx
	)

Definition at line 479 of file aggregator.cpp.

References gmatrix::cr, comprow::crxcv_cv(), and compvect::nz.

void assemble_tentative_prol	(	long	agrid,
		long	cid,
		long *	i,
		double *	a
	)

void boss	(	gmatrix *	gm,
		double *	u,
		double *	v
	)

function computes Black-box Overlapping Schwarz with Smoothed coarse space see Brezina, page 75 and 76

Parameters:

	gm	- matrix of the system
	u	- input vector
	v	- output vector

JK, 20.2.2007

Definition at line 1275 of file aggregator.cpp.

References slesolv::aerr, slesolv::ani, comprow::cg(), cms, copyv(), slesolv::err, exinex, fillv(), globloc(), gmatrix::gmxv(), skyline::ldl_sky(), lmcr, lmsky, lnuaggr, locglob(), luaggr, maxnu, n, na, slesolv::ni, ssle, subv(), and slesolv::zero.

Referenced by precond::preconditioning().

void clean_memory ( )

Definition at line 946 of file aggregator.cpp.

References lnagr, lnnagr, lnntagr, lntagr, and na.

Referenced by prepare_boss().

void coarse_matrix ( gmatrix * gm )

function computes coarse matrix of the problem

at this time, first and unefficient implementation is used the matrix is stored in the dense matrix storage scheme

Parameters:

gm

- general matrix

JK, 27.8.2008

Definition at line 708 of file aggregator.cpp.

References densemat::add_entry(), densemat::alloc(), cm, cms, gmatrix::gmxv(), n, p, and ss().

Referenced by prepare_boss().

void define_degree ( long i )

void define_degree ( )

function sets up degree of recursion real degree of polynomial is equal to (3^dg-1)/2

function defines depth of recursion for smoothed prolongator dg = 0 .... use polynomial of degree 0

PM

Definition at line 471 of file aggregator.cpp.

References degree_k, and degree_r.

Referenced by prepare_boss().

void define_graph	(	long	nnod,
		long *	nneigh,
		long **	listneigh
	)

vytvori graf matice nn - pocet uzlu nneigh - pocty sousednich uzlu listneigh - seznamy sousednich uzlu

function defines graph of the matrix

Parameters:

	nnod	- number of nodes
	nneigh	- numbers of adjacent nodes nneigh[i]=j - there are j adjacent nodes to the i-th node
	listneigh	- list of node numbers of adjacent nodes listneigh[i][j]=k - the j-th adjacent node to the i-th node has number k

JK, 4.2.2007

Definition at line 203 of file aggregator.cpp.

References adjnodnod, nadjnodnod, and nn.

void define_matsize	(	long	i,
		long	j
	)

void define_na ( long i )

defines number of aggregates

function defines required number of aggregates

Parameters:

i

- required number of aggregates

JK, 4.2.2007

Definition at line 187 of file aggregator.cpp.

References na.

void gener_rbm_1	(	gtopology *	gt,
		long	agid,
		double *	v
	)

function computes kernels for heat transfer, 1D mechancs, etc.

Parameters:

	gt	- pointer to the general topology
	agid	- aggregate id
	v	- array containing kernel

JK, 27.8.2008

Definition at line 638 of file aggregator.cpp.

References gtopology::give_dof(), lnagr, and lnnagr.

Referenced by assemble_smoothed_prol().

void gener_rbm_2	(	gtopology *	gt,
		long	agid,
		long	cnid,
		double *	v
	)

function computes kernels for plane stress

Parameters:

	gt	- pointer to the general topology
	agid	- aggregate id
	cnid	- code number id
	v	- array containing kernel

JK, 27.8.2008

Definition at line 660 of file aggregator.cpp.

References gtopology::give_dof(), gtopology::gnodes, lnagr, lnnagr, gnode::x, and gnode::y.

Referenced by assemble_smoothed_prol().

long give_deg ( )

void local_matrices	(	long	bsize,
		gmatrix *	gm
	)

function assembles local matrices of aggregates

Parameters:

	bsize	- size of blocks for sparse direct solvers only
	gm	- matrix of the system

JK, 20.2.2007

Definition at line 814 of file aggregator.cpp.

References comprow::adr, skyline::assemble_from_cr(), ci, gmatrix::cr, exinex, lmcr, lmsky, lnuaggr, luaggr, n, comprow::n, na, skyline::negm, and comprow::select_submatrix().

Referenced by prepare_boss().

void matrix_assemble	(	long *	i,
		long *	j,
		double *	v
	)

void metis_aggr ( gtopology * gt )

function copies arrays lnntagr and lntagr from the object stop (class seqtop) allocated in class gtopology it is an alternative to the function void prepare_tlnagr(void);

function copies arrays lnntagr and lntagr from the object stop (class seqtop) allocated in class gtopology

it is an alternative to the function void prepare_tlnagr(void);

Parameters:

gt

- global topology

JK, 31.5.2008

Definition at line 367 of file aggregator.cpp.

References lnntagr, lntagr, seqtop::ltg, na, seqtop::nnsd, and gtopology::stop.

Referenced by prepare_boss().

void mulA	(	long	st,
		double *	x,
		gmatrix *	mtx
	)

st .... stupen S, t.j. dolni index x .... vektor mtx .... matice

Definition at line 597 of file aggregator.cpp.

References copyv(), gmatrix::gmxv(), n, and specrad.

Referenced by mulS().

void mulS	(	long	st,
		double *	x,
		gmatrix *	mtx
	)

lokalni metody - casem asi private

provedena aplikaci s(st) na pozadovany vekor st .... stupen S, t.j. dolni index x .... vektor mtx .... matice

Definition at line 569 of file aggregator.cpp.

References copyv(), mulA(), n, and specrad.

Referenced by assemble_smoothed_prol().

void prepare_boss	(	gtopology *	gt,
		gmatrix *	gm,
		FILE *	out
	)

function prepares all necessary variables, arrays, etc. for BOSS algorithm

Parameters:

	gt	- general topology
	gm	- matrix of the system stored in the class gmatrix
	out	- output stream

JK, 25.2.2007

Definition at line 976 of file aggregator.cpp.

References gtopology::adjacnodes(), gtopology::adjnodnod, adjnodnod, assemble_aggr_unknowns(), assemble_smoothed_prol(), clean_memory(), cm, cms, coarse_matrix(), define_degree(), gmatrix::estim_spect_radius(), exinex, gtopology::gnodes, impl, skyline::ldl_sky(), densemat::ll(), lmsky, lnagr, lnnagr, local_matrices(), metis_aggr(), gmatrix::n, n, na, gtopology::nadjnodnod, nadjnodnod, gnode::ndofn, gtopology::nn, nn, nrbm, gtopology::ns, prepare_lnagr(), prepare_tlnagr(), print_err(), specrad, time, and trbm.

Referenced by precond::initiation().

void prepare_lnagr ( )

assembles list of numbers of nodes in aggregates with overlap pocet uzlu v agregatech s prekryvem assembles node numbers in aggregates with overlap seznamy uzlu v jednotlivych agregatech s prekryvem lnnagr[i] = k - v i-tem agregatu je k uzlu lnagr[i][j] = k - j-ty uzel v i-tem agregatu ma cislo k

pripravi seznamy smoothed prolongatoru

pred zavolanim teto funkce se musi zavolat funkce define_degree(long dg);

PM

Definition at line 397 of file aggregator.cpp.

References adjnodnod, degree_r, lnagr, lnnagr, lnntagr, lntagr, memset(), na, nadjnodnod, and nn.

Referenced by prepare_boss().

void prepare_tlnagr ( void )

assembles list of numbers of nodes in aggregates without overlap pocet uzlu v agregatech bez prekryvu assembles node numbers in aggregates without overlap seznamy uzlu v jednotlivych agregatech bez prekryvu lnntagr[i] = k - v i-tem agregatu je k uzlu lntagr[i][j] = k - j-ty uzel v i-tem agregatu ma cislo k

functions creates disjoint covering of node graph set both list of sizes and list of lists

PM

Definition at line 220 of file aggregator.cpp.

References adjnodnod, lnntagr, lntagr, memset(), na, nadjnodnod, and nn.

Referenced by prepare_boss().

void print ( FILE * out )

function prints data about aggregation

Parameters:

out

- output stream

31.5.2008, JK

Definition at line 155 of file aggregator.cpp.