aAMG_level

This class is used to manage the aAMG hierarchy. More…

#include "aAMG_level.h"

Public Functions

	Name
	aAMG_level(const DSMat & mat_A) Constructs the object.
	~aAMG_level() Deletes the object.
void	SetOffsets(const iReg mynrows, const iReg nRHS) Sets the offsets for RHS and SOL in the last level.
void	Prepare_MxV(const iReg nRHS) Prepares the object for the application to a vector (DDMat).
void	UndoPrepare_MxV() Undo the data structure necessary for MxV.

Public Attributes

	Name
int	level Hierarchy level.
struct aAMG_OperInfo	OperInfo Operator information.
struct aAMG_TimeInfo	TimeInfo Set-up wall time of the AMG components.
MPI_Comm	Comm Communicator for the current level.
bool	is_SYM Flag to indicate symmetry of the coarsest level operator.
bool	CoarsestFac Flag for the coarsest level with factorization.
bool	CoarsestDiag Flag for the coarsest level without factorization (diagonal case).
SMOOTHER *	ptr_SMOOTH Pointer to the smoother for the current level.
PROLONGATION *	ptr_PROLO Pointer to the prolongation for the current level.
DSMat *	ptr_A Pointer to the sparse matrix for the current level.
DDMat *	ptr_LUFacA Pointer to the LU factorization of A (only for the last level)
VEC_CPU< type_LAPACK_int >	vec_IPIV Permution LU in case of non-symmetric coarsest operator.
aAMG_level *	ptr_nextlevel Pointer to the next level.
VEC_CPU< type_MPI_int >	nRecv_loc Number of receiving coefficiets for RHS communication in the last level.
VEC_CPU< type_MPI_int >	offsets RHS and SOL offsets in the last level.
DDMat *	WR1 First work array (used in recursive ApplyToVec).
DDMat *	WR2 Second work array (used in recursive ApplyToVec).
DDMat *	WR3 Third work array (used in recursive ApplyToVec).
DDMat *	nextRHS RHS vector for teh next level (used in recursive ApplyToVec).
DDMat *	nextSOL SOL vector for the next level (used in recursive ApplyToVec).

Detailed Description

class aAMG_level;

This class is used to manage the aAMG hierarchy.

class aAMG_level.

Public Functions Documentation

function aAMG_level

aAMG_level(
    const DSMat & mat_A
)

Constructs the object.

function ~aAMG_level

~aAMG_level()

Deletes the object.

function SetOffsets

void SetOffsets(
    const iReg mynrows,
    const iReg nRHS
)

Sets the offsets for RHS and SOL in the last level.

Parameters:

• mynrows local size of the matrix.
• nRHS number of Right Hand Sides.

function Prepare_MxV

void Prepare_MxV(
    const iReg nRHS
)

Prepares the object for the application to a vector (DDMat).

Parameters:

• nRHS number of Right Hand Sides.

function UndoPrepare_MxV

void UndoPrepare_MxV()

Undo the data structure necessary for MxV.

Public Attributes Documentation

variable level

int level = 0;

Hierarchy level.

variable OperInfo

struct aAMG_OperInfo OperInfo;

Operator information.

variable TimeInfo

struct aAMG_TimeInfo TimeInfo;

Set-up wall time of the AMG components.

variable Comm

MPI_Comm Comm = MPI_COMM_WORLD;

Communicator for the current level.

variable is_SYM

bool is_SYM = false;

Flag to indicate symmetry of the coarsest level operator.

variable CoarsestFac

bool CoarsestFac = false;

Flag for the coarsest level with factorization.

variable CoarsestDiag

bool CoarsestDiag = false;

Flag for the coarsest level without factorization (diagonal case).

variable ptr_SMOOTH

SMOOTHER * ptr_SMOOTH = nullptr;

Pointer to the smoother for the current level.

variable ptr_PROLO

PROLONGATION * ptr_PROLO = nullptr;

Pointer to the prolongation for the current level.

variable ptr_A

DSMat * ptr_A = nullptr;

Pointer to the sparse matrix for the current level.

variable ptr_LUFacA

DDMat * ptr_LUFacA = nullptr;

Pointer to the LU factorization of A (only for the last level)

variable vec_IPIV

VEC_CPU< type_LAPACK_int > vec_IPIV;

Permution LU in case of non-symmetric coarsest operator.

variable ptr_nextlevel

aAMG_level * ptr_nextlevel = nullptr;

Pointer to the next level.

variable nRecv_loc

VEC_CPU< type_MPI_int > nRecv_loc;

Number of receiving coefficiets for RHS communication in the last level.

variable offsets

VEC_CPU< type_MPI_int > offsets;

RHS and SOL offsets in the last level.

variable WR1

DDMat * WR1 = nullptr;

First work array (used in recursive ApplyToVec).

variable WR2

DDMat * WR2 = nullptr;

Second work array (used in recursive ApplyToVec).

variable WR3

DDMat * WR3 = nullptr;

Third work array (used in recursive ApplyToVec).

variable nextRHS

DDMat * nextRHS = nullptr;

RHS vector for teh next level (used in recursive ApplyToVec).

variable nextSOL

DDMat * nextSOL = nullptr;

SOL vector for the next level (used in recursive ApplyToVec).

---

Updated on 19 November 2024 at 10:20:48 CET