dsdpschurmat.h File Reference

Detailed Description

Methods of a Schur Matrix.

Go to the source code of this file.

Data Structures

struct DSDPSchurMat_C

Schur complement matrix whose solution is the Newton direction. More...

Typedefs

typedef struct DSDPSchurMat_C DSDPSchurMat

This object represents the Schur Matrix. Its structure is opaque to the DSDP solver, but it must implement the interface below and provide a structure of function pointers.

Functions

int DSDPSchurMatAddDiagonal (DSDPSchurMat, DSDPVec)

Add elements to a row of the Schur matrix.

int DSDPSchurMatAddDiagonalElement (DSDPSchurMat, int, double)

Determine with the cone should compute this diagonal element of M and RHS.

int DSDPSchurMatAddR (DSDPSchurMat, int, double)

Add an element to the Schur matrix correponding the variable r.

int DSDPSchurMatAddRow (DSDPSchurMat, int, double, DSDPVec)

Add elements to a row of the Schur matrix.

int DSDPSchurMatAssemble (DSDPSchurMat)

Final assembly of M.

int DSDPSchurMatDestroy (DSDPSchurMat *)

Free the memory in the data structure.

int DSDPSchurMatDiagonalScaling (DSDPSchurMat, DSDPVec)

Get the scaling and nonzero pattern of each diagonal element of the matrix.

int DSDPSchurMatFactor (DSDPSchurMat, DSDPTruth *)

Factor M.

int DSDPSchurMatInitialize (DSDPSchurMat *)

Initialize pointers to null.

int DSDPSchurMatInParallel (DSDPSchurMat, DSDPTruth *)

Determine whether M is computed in parallel.

int DSDPSchurMatMultiply (DSDPSchurMat, DSDPVec, DSDPVec)

Multiply M by a vector. y = M x.

int DSDPSchurMatReducePVec (DSDPSchurMat, DSDPVec)

Collect elements of the vector.

int DSDPSchurMatRowColumnScaling (DSDPSchurMat, int, DSDPVec, int *)

Get the scaling and nonzero pattern of each column in this row of the matrix.

int DSDPSchurMatRowScaling (DSDPSchurMat, DSDPVec)

Identify which rows on on this processor.

int DSDPSchurMatSetData (DSDPSchurMat *, struct DSDPSchurMat_Ops *, void *)

Set the Schur matrix with an opaque pointer and structure of function pointers.

int DSDPSchurMatSetR (DSDPSchurMat, double)

Set up the data structure.

int DSDPSchurMatSetup (DSDPSchurMat, DSDPVec)

Set up the data structure.

int DSDPSchurMatShiftDiagonal (DSDPSchurMat, double)

Add a scalar to each diagonal element of the matrix.

int DSDPSchurMatSolve (DSDPSchurMat, DSDPVec, DSDPVec)

Solve the linear system.

int DSDPSchurMatVariableCompute (DSDPSchurMat, int, double *)

Determine with the cone should compute this diagonal element of M and RHS.

int DSDPSchurMatVariableComputeR (DSDPSchurMat, double *)

Add an element to the Schur matrix correponding the variable r.

int DSDPSchurMatView (DSDPSchurMat)

Print the matrix.

int DSDPSchurMatZeroEntries (DSDPSchurMat)

Zero all element in the matrix.

Typedef Documentation

struct DSDPSchurMat_C DSDPSchurMat

This object represents the Schur Matrix. Its structure is opaque to the DSDP solver, but it must implement the interface below and provide a structure of function pointers.

Definition at line 49 of file dsdpschurmat.h.

Function Documentation

int DSDPSchurMatAddDiagonal	(	DSDPSchurMat	M,
		DSDPVec	D
	)

Add elements to a row of the Schur matrix.

Parameters:

	M	matrix
	D	diagonal elements.

Conic object call this routine when evaluating the Hessian of the barrier term.

See also:: DSDPSchurMatDiagonalScaling()

Definition at line 272 of file dsdpschurmatadd.c.

int DSDPSchurMatAddDiagonalElement	(	DSDPSchurMat	M,
		int	row,
		double	dd
	)

Determine with the cone should compute this diagonal element of M and RHS.

Parameters:

	M	matrix
	row	correponding the y variable
	dd	zero or one..

See also:: DSDPSchurMatVariableCompute()

Definition at line 157 of file dsdpschurmatadd.c.

int DSDPSchurMatAddR	(	DSDPSchurMat	M,
		int	row,
		double	dd
	)

Add an element to the Schur matrix correponding the variable r.

Parameters:

	M	matrix
	row	corresponding to variable y.
	dd	element

Definition at line 198 of file dsdpschurmatadd.c.

int DSDPSchurMatAddRow	(	DSDPSchurMat	M,
		int	row,
		double	alpha,
		DSDPVec	R
	)

Add elements to a row of the Schur matrix.

Parameters:

	M	matrix
	row	correponding to variable y
	alpha	multiply elements in R by this scalar.
	R	a row of elements.

Conic object call this routine when evaluating the Hessian of the barrier term. This routine is to be used with DSDPSchurMatRowColumnScaling().

See also:: DSDPSchurMatRowColumnScaling()

Definition at line 76 of file dsdpschurmatadd.c.

Referenced by DSDPAddSchurRow(), and SDPConeComputeHessian().

int DSDPSchurMatAssemble ( DSDPSchurMat M )

Final assembly of M.

Parameters:

M

matrix

Important in parallel implementation.

Definition at line 174 of file dsdpschurmat.c.

Referenced by DSDPComputeHessian().

int DSDPSchurMatDestroy ( DSDPSchurMat * M )

Free the memory in the data structure.

Parameters:

M

matrix

Definition at line 414 of file dsdpschurmat.c.

Referenced by DSDPTakeDown().

int DSDPSchurMatDiagonalScaling	(	DSDPSchurMat	M,
		DSDPVec	D
	)

Get the scaling and nonzero pattern of each diagonal element of the matrix.

Parameters:

	M	matrix
	D	multply each element of the diagonal by this quantity.

Conic object call this routine when evaluating the Hessian of the barrier term. The elements will be a 0 or a 1. Important for parallel version.

See also:: DSDPSchurMatAddDiagonal()

Definition at line 235 of file dsdpschurmatadd.c.

Referenced by DSDPSchurMatRowScaling().

int DSDPSchurMatFactor	(	DSDPSchurMat	M,
		DSDPTruth *	successful
	)

Factor M.

Parameters:

	M	matrix
	successful	false if factorization failed.

Definition at line 196 of file dsdpschurmat.c.

Referenced by DSDPComputeDualStepDirections().

int DSDPSchurMatInitialize ( DSDPSchurMat * M )

Initialize pointers to null.

Parameters:

M

Schur matrix object

Definition at line 79 of file dsdpschurmat.c.

Referenced by DSDPCreate().

int DSDPSchurMatInParallel	(	DSDPSchurMat	M,
		DSDPTruth *	flag
	)

Determine whether M is computed in parallel.

Parameters:

	M	matrix
	flag	true or false

Important in parallel implementation.

Definition at line 149 of file dsdpschurmat.c.

Referenced by DSDPSchurMatDiagonalScaling(), DSDPSchurMatReducePVec(), and DSDPSchurMatVariableCompute().

int DSDPSchurMatMultiply	(	DSDPSchurMat	M,
		DSDPVec	x,
		DSDPVec	y
	)

Multiply M by a vector. y = M x.

Parameters:

	M	matrix
	x	in vector.
	y	product.

Definition at line 231 of file dsdpschurmat.c.

int DSDPSchurMatReducePVec	(	DSDPSchurMat	M,
		DSDPVec	x
	)

Collect elements of the vector.

Parameters:

	M	matrix
	x	vector.

Important in parallel implementation. Otherwise does nothing.

Definition at line 307 of file dsdpschurmat.c.

Referenced by DSDPComputeG(), DSDPComputeHessian(), and DSDPHessianMultiplyAdd().

int DSDPSchurMatRowColumnScaling	(	DSDPSchurMat	M,
		int	row,
		DSDPVec	V,
		int *	nzcols
	)

Get the scaling and nonzero pattern of each column in this row of the matrix.

Parameters:

	M	matrix
	row	correponding to variable y
	V	multply each element of the row by this quantity.
	nzcols	how many nonzeros. Check for a 0! Conic object call this routine when evaluating the Hessian of the barrier term. The vector V identifies sparsity, whether its using upper or lower half of Schur, and also used to distribute rows over processros The elements will be a 0 or a 1. This routine is to be used with DSDPSchurMatAddRow().

See also:: DSDPSchurMatAddRow()

Definition at line 33 of file dsdpschurmatadd.c.

Referenced by SDPConeComputeHessian().

int DSDPSchurMatRowScaling	(	DSDPSchurMat	M,
		DSDPVec	D
	)

Identify which rows on on this processor.

Parameters:

	M	matrix
	D	Variables marked with 0 or 1.

Definition at line 399 of file dsdpschurmat.c.

Referenced by DSDPComputeG(), and DSDPHessianMultiplyAdd().

int DSDPSchurMatSetData	(	DSDPSchurMat *	M,
		struct DSDPSchurMat_Ops *	ops,
		void *	data
	)

Set the Schur matrix with an opaque pointer and structure of function pointers.

Parameters:

	M	Schur matrix object
	ops	address of a structure of function pointers.
	data	opaque pointer to its internal data structure.

Definition at line 28 of file dsdpschurmat.c.

Referenced by DSDPSchurMatDestroy(), DSDPSchurMatInitialize(), and DSDPSetSchurMatOps().

int DSDPSchurMatSetR	(	DSDPSchurMat	M,
		double	rr
	)

Set up the data structure.

Parameters:

	M	matrix
	rr	residual

Definition at line 338 of file dsdpschurmat.c.

Referenced by DSDPComputeG(), DSDPComputeHessian(), and DSDPComputeNewY().

int DSDPSchurMatSetup	(	DSDPSchurMat	M,
		DSDPVec	Y
	)

Set up the data structure.

Parameters:

	M	matrix
	Y	variable vector.

Definition at line 352 of file dsdpschurmat.c.

Referenced by DSDPSetup().

int DSDPSchurMatShiftDiagonal	(	DSDPSchurMat	M,
		double	dd
	)

Add a scalar to each diagonal element of the matrix.

Parameters:

	M	matrix
	dd	diagonal shift

Definition at line 120 of file dsdpschurmat.c.

Referenced by DSDPComputeDualStepDirections().

int DSDPSchurMatSolve	(	DSDPSchurMat	M,
		DSDPVec	b,
		DSDPVec	x
	)

Solve the linear system.

Parameters:

	M	matrix
	b	the right-hand side
	x	solution

Definition at line 466 of file dsdpschurmat.c.

Referenced by DSDPCGSolve().

int DSDPSchurMatVariableCompute	(	DSDPSchurMat	M,
		int	row,
		double *	rcv
	)

Determine with the cone should compute this diagonal element of M and RHS.

Parameters:

	M	matrix
	row	correponding the y variable
	rcv	Used to evaluate M. Important in parallel implementation.

See also:: DSDPSchurMatAddDiagonalElement()

Definition at line 120 of file dsdpschurmatadd.c.

Referenced by DSDPObjectiveGH().

int DSDPSchurMatVariableComputeR	(	DSDPSchurMat	M,
		double *	rcv
	)

Add an element to the Schur matrix correponding the variable r.

Parameters:

	M	matrix
	*rcv	zero or one

Definition at line 181 of file dsdpschurmatadd.c.

int DSDPSchurMatView ( DSDPSchurMat M )

Print the matrix.

Parameters:

M

matrix

Definition at line 376 of file dsdpschurmat.c.

Referenced by DSDPComputeDualStepDirections().

int DSDPSchurMatZeroEntries ( DSDPSchurMat M )

Zero all element in the matrix.

Parameters:

M

matrix

Definition at line 97 of file dsdpschurmat.c.

Referenced by DSDPComputeHessian().


Data Structures
struct	DSDPSchurMat_C
	Schur complement matrix whose solution is the Newton direction. More...
Typedefs
typedef struct DSDPSchurMat_C	DSDPSchurMat
	This object represents the Schur Matrix. Its structure is opaque to the DSDP solver, but it must implement the interface below and provide a structure of function pointers.
Functions
int	DSDPSchurMatAddDiagonal (DSDPSchurMat, DSDPVec)
	Add elements to a row of the Schur matrix.
int	DSDPSchurMatAddDiagonalElement (DSDPSchurMat, int, double)
	Determine with the cone should compute this diagonal element of M and RHS.
int	DSDPSchurMatAddR (DSDPSchurMat, int, double)
	Add an element to the Schur matrix correponding the variable r.
int	DSDPSchurMatAddRow (DSDPSchurMat, int, double, DSDPVec)
	Add elements to a row of the Schur matrix.
int	DSDPSchurMatAssemble (DSDPSchurMat)
	Final assembly of M.
int	DSDPSchurMatDestroy (DSDPSchurMat *)
	Free the memory in the data structure.
int	DSDPSchurMatDiagonalScaling (DSDPSchurMat, DSDPVec)
	Get the scaling and nonzero pattern of each diagonal element of the matrix.
int	DSDPSchurMatFactor (DSDPSchurMat, DSDPTruth *)
	Factor M.
int	DSDPSchurMatInitialize (DSDPSchurMat *)
	Initialize pointers to null.
int	DSDPSchurMatInParallel (DSDPSchurMat, DSDPTruth *)
	Determine whether M is computed in parallel.
int	DSDPSchurMatMultiply (DSDPSchurMat, DSDPVec, DSDPVec)
	Multiply M by a vector. y = M x.
int	DSDPSchurMatReducePVec (DSDPSchurMat, DSDPVec)
	Collect elements of the vector.
int	DSDPSchurMatRowColumnScaling (DSDPSchurMat, int, DSDPVec, int *)
	Get the scaling and nonzero pattern of each column in this row of the matrix.
int	DSDPSchurMatRowScaling (DSDPSchurMat, DSDPVec)
	Identify which rows on on this processor.
int	DSDPSchurMatSetData (DSDPSchurMat , struct DSDPSchurMat_Ops , void *)
	Set the Schur matrix with an opaque pointer and structure of function pointers.
int	DSDPSchurMatSetR (DSDPSchurMat, double)
	Set up the data structure.
int	DSDPSchurMatSetup (DSDPSchurMat, DSDPVec)
	Set up the data structure.
int	DSDPSchurMatShiftDiagonal (DSDPSchurMat, double)
	Add a scalar to each diagonal element of the matrix.
int	DSDPSchurMatSolve (DSDPSchurMat, DSDPVec, DSDPVec)
	Solve the linear system.
int	DSDPSchurMatVariableCompute (DSDPSchurMat, int, double *)
	Determine with the cone should compute this diagonal element of M and RHS.
int	DSDPSchurMatVariableComputeR (DSDPSchurMat, double *)
	Add an element to the Schur matrix correponding the variable r.
int	DSDPSchurMatView (DSDPSchurMat)
	Print the matrix.
int	DSDPSchurMatZeroEntries (DSDPSchurMat)
	Zero all element in the matrix.