PardisoSchurSolver Class Reference

#include <PardisoSchurSolver.h>

Inheritance diagram for PardisoSchurSolver:

Public Member Functions
virtual void	firstCall ()
void	firstSolveCall (SparseGenMatrix &R, SparseGenMatrix &A, SparseGenMatrix &C)
	PardisoSchurSolver (SparseSymMatrix *sgm)
	PardisoSchurSolver (SparseSymMatrix *sgm, const int numOfNegEigVal_in)
virtual void	diagonalChanged (int idiag, int extent)
virtual int	matrixChanged ()
virtual void	solve (OoqpVector &rhs)
virtual void	solve (GenMatrix &rhs)
virtual void	schur_solve (SparseGenMatrix &R, SparseGenMatrix &A, SparseGenMatrix &C, DenseSymMatrix &SC)
Public Member Functions inherited from DoubleLinearSolver
	DoubleLinearSolver ()
virtual void	solveTrans (OoqpVector &x)
virtual void	Lsolve (OoqpVector &x)
virtual void	Dsolve (OoqpVector &x)
virtual void	Ltsolve (OoqpVector &x)
virtual	~DoubleLinearSolver ()

Protected Member Functions
	PardisoSchurSolver ()
void	SetUpPardisoSchurSolver (SparseSymMatrix *sgm)
virtual	~PardisoSchurSolver ()

Protected Attributes
SparseSymMatrix *	Msys
bool	first
bool	firstSolve
void *	pt [64]
int	iparm [64]
double	dparm [64]
int	num_threads
int	n
int	nSC
int	nnz
int *	rowptrAug
int *	colidxAug
double *	eltsAug
int *	index_LowTri
map< int, int >	diagMap
int *	elementMap_R
int *	elementMap_A
int *	elementMap_C
int	numOfNegEigVal
double *	nvec

Additional Inherited Members
Public Attributes inherited from DoubleLinearSolver
int	negEigVal
int	KryIter

Detailed Description

implements the linear solver class using the Pardiso SC solver

Constructor & Destructor Documentation

PardisoSchurSolver::PardisoSchurSolver ( )

inlineprotected

PardisoSchurSolver::PardisoSchurSolver

(

SparseSymMatrix *

sgm

)

sets mStorage to refer to the argument sgm

PardisoSchurSolver::PardisoSchurSolver	(	SparseSymMatrix *	sgm,
		const int	numOfNegEigVal_in
	)

PardisoSchurSolver::~PardisoSchurSolver ( )

protectedvirtual

Member Function Documentation

void PardisoSchurSolver::diagonalChanged ( int  idiag, int  extent  )

virtual

called if the diagonal elements of the matrix have changed. Triggers a refactorization of the matrix, if necessary.

Parameters

idiag	index of the first diagonal element that changed
extent	the number of diagonal element that changed.

Implements DoubleLinearSolver.

void PardisoSchurSolver::firstCall ( )

virtual

Reimplemented in PardisoSchur32Solver.

void PardisoSchurSolver::firstSolveCall	(	SparseGenMatrix &	R,
		SparseGenMatrix &	A,
		SparseGenMatrix &	C
	)

void PardisoSchurSolver::matrixChanged ( )

virtual

called if some elements of the matrix have changed. Triggers a refactorization of the matrix, if necessary.

Implements DoubleLinearSolver.

void PardisoSchurSolver::schur_solve ( SparseGenMatrix &  R, SparseGenMatrix &  A, SparseGenMatrix &  C, DenseSymMatrix &  SC  )

virtual

Functions specific to the Schur approach. The last argument is the Schur first stage matrix that will be updated.

1. schur_solver( rhs, SC)
   • this is the generic function
2. schur_solve(R,A,B, SC)
   • avoids forming the matrix rhs [R' A' B']'
   • assumes rhs block matrices does not change

void PardisoSchurSolver::SetUpPardisoSchurSolver ( SparseSymMatrix *  sgm )

protected

void PardisoSchurSolver::solve ( OoqpVector &  x )

virtual

solves a linear system.

Parameters

x	on entry the right hand side of the system to be solved. On exit, the solution.

Implements DoubleLinearSolver.

Reimplemented in PardisoSchur32Solver.

void PardisoSchurSolver::solve ( GenMatrix &  rhs )

virtual

Reimplemented from DoubleLinearSolver.

Member Data Documentation

int * PardisoSchurSolver::colidxAug

protected

map<int,int> PardisoSchurSolver::diagMap

protected

mapping from from the diagonals of the PIPS linear systems to the diagonal elements of the (1,1) block in the augmented system

double PardisoSchurSolver::dparm[64]

protected

int* PardisoSchurSolver::elementMap_A

protected

int* PardisoSchurSolver::elementMap_C

protected

int* PardisoSchurSolver::elementMap_R

protected

mapping from the element of the PIPS linear systems to the elements of the (1,1) block in the augmented system, hence we can update Jac

double* PardisoSchurSolver::eltsAug

protected

bool PardisoSchurSolver::first

protected

bool PardisoSchurSolver::firstSolve

protected

int* PardisoSchurSolver::index_LowTri

protected

int PardisoSchurSolver::iparm[64]

protected

SparseSymMatrix* PardisoSchurSolver::Msys

protected

int PardisoSchurSolver::n

protected

dimension of the PARDISO augmented system

int PardisoSchurSolver::nnz

protected

number of nonzeros in the PARDISO augmented matrix

int PardisoSchurSolver::nSC

protected

dimension of the Schur complement (# of rhs)

int PardisoSchurSolver::num_threads

protected

int PardisoSchurSolver::numOfNegEigVal

protected

double* PardisoSchurSolver::nvec

protected

void* PardisoSchurSolver::pt[64]

protected

int* PardisoSchurSolver::rowptrAug

protected

storage for the upper triangular (in row-major format)

---

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

• /home/michel/git/PIPS/PIPS-NLP/Core/LinearSolvers/PardisoSolver/PardisoSchurSolver.h
• /home/michel/git/PIPS/PIPS-NLP/Core/LinearSolvers/PardisoSolver/PardisoSchurSolver.C
• /home/michel/git/PIPS/PIPS-NLP/Core/LinearSolvers/PardisoSolver/testV.C