A proxy class that can wrap and behave like a GmNumSolver object, adding the possibility for printing and / or saving the linear system matrices and vectors according to the given configuration. This feature was not included directly in the numeric solver API since numeric solvers can be shared between multiple solver processes with different print options. More...

#include <gmNumSolverPrinter.h>

Collaboration diagram for GmNumSolverPrinter:

[legend]

Public Types
enum	PrintAndSaveOptions { GM_NS_PRINT_MATRIX = 0x01 , GM_NS_PRINT_VECTOR = 0x02 , GM_NS_PRINT_RESULT = 0x04 , GM_NS_SAVE_MATRIX = 0x10 , GM_NS_SAVE_VECTOR = 0x20 , GM_NS_SAVE_RESULT = 0x40 }
	Options for printing and saving parts of the linear system Ax = b. More...

Public Member Functions
	GmNumSolverPrinter (GmNumSolver *solver)
	Constructor. Receives as parameter the wrapped solver object.

GmNumSolver *	solver () const
	Returns the wrapped numeric solver.

void	setPrintAndSaveOptions (int options, QString baseFileName, int fieldWidth=0, char format='g', int precision=-1, int filterAttr=-1, const QVector< double > &attrValues=QVector< double >())
	Set print and save options. Options is an or of PrintAndSaveOptions flags. If filterAttr is different from -1, the print or save operation will only be executed if the value of the given filter attribute (from GmRunProgress) is in attrValues (uses GmDoubleCmp::equal() for comparisons). The attribute should be a numeric attribute that increases with the simulation evolution. The attrValues list MUST be ordered.

bool	solve (GmSolverMatrix *A, const GmVector &b, GmVector &x, bool xIsGuess=false) const
	Solves the linear system Ax = b, filling the vector x with the calculated result. See comments on GmNumSolver::solve().

GmSolverMatrix *	newMatrix (int nlin, int ncol, const GmSolverMatrix *sharedLayout, int numThreads=-1) const

QString	id () const

QString	description () const

QString	pluginTypeName () const

const char *	pluginCategory () const

const char *	pluginName () const

const char *	pluginType () const

bool	setSolverTolerance (double tol)

QString	quickDescription () const

int	trackItemId () const

void	printParameters (const GmLogCategory &logger)

Private Member Functions
void	printAndSave (const GmSolverMatrix *A, const GmVector &b, QString filterStr) const
	Prints / saves the linear system A and b terms, depending on the configuration.

void	printAndSave (const GmVector &x, QString filterStr) const
	Prints / saves the linear system x term, depending on the configuration.

bool	saveMatrix (FILE f, const GmSolverMatrix A, bool extended) const
	Saves the matrix vector contents to the given file using Matrix market format. If extended is true, adjusts the header for saving a vector afterwards.

bool	saveVector (FILE *f, const GmVector &v, bool header) const
	Saves the given vector contents to the given file using Matrix market format. The header will only be saved if header is true.

bool	checkPrintAndSaveFilter (QString &filterStr) const
	Returns true if the current Linear system should be printed / saved. In that case, if there is a filter, fills filterStr with the file name suffix for the applied filter.

Private Attributes
GmNumSolver *	_solver
	The wrapped solver object.

int	_psOptions
	Print and save options. An or of PrintAndSaveOptions flags

int	_printFieldWidth
	Print format parameter: The matrix column minimum width.

char	_printFormat
	Print format parameter: The format type f' or 'g'.

int	_printPrecision
	Print format parameter: The number of decimal places.

QString	_saveBaseFileName
	The base file name for saving.

int	_saveIndex
	The unique index used to guarantee file name unicity.

int	_psFilterAttr
	The print and save filter attribute, -1 for no filtering.

QVector< double >	_psAttrValues
	The sorted list of attribute values for printing and saving when the filter is active.

QString	_psFilterSuffix
	The file suffix with the name of the active filter attribute (with : characters converted to -)

int	_nextAttrValIndex
	The index in _psAttrValues of the next attr value for the save / print filter.

Detailed Description

A proxy class that can wrap and behave like a GmNumSolver object, adding the possibility for printing and / or saving the linear system matrices and vectors according to the given configuration. This feature was not included directly in the numeric solver API since numeric solvers can be shared between multiple solver processes with different print options.

Member Enumeration Documentation

◆ PrintAndSaveOptions

enum GmNumSolverPrinter::PrintAndSaveOptions

Options for printing and saving parts of the linear system Ax = b.

Enumerator
GM_NS_PRINT_MATRIX	Should we print the A matrix before solving?
GM_NS_PRINT_VECTOR	Should we print the b vector before solving?
GM_NS_PRINT_RESULT	Should we print the x vector after solving?
GM_NS_SAVE_MATRIX	Should we save to a file the A matrix before solving?
GM_NS_SAVE_VECTOR	Should we save to a file the b vector before solving?
GM_NS_SAVE_RESULT	Should we save to a file the x vector after solving?

Member Function Documentation

◆ setPrintAndSaveOptions()

void GmNumSolverPrinter::setPrintAndSaveOptions	(	int	options,
		QString	baseFileName,
		int	fieldWidth = 0,
		char	format = 'g',
		int	precision = -1,
		int	filterAttr = -1,
		const QVector< double > &	attrValues = QVector<double>() )

Set print and save options. Options is an or of PrintAndSaveOptions flags. If filterAttr is different from -1, the print or save operation will only be executed if the value of the given filter attribute (from GmRunProgress) is in attrValues (uses GmDoubleCmp::equal() for comparisons). The attribute
should be a numeric attribute that increases with the simulation evolution. The attrValues list MUST be ordered.

When printing, the given fieldWidth, format and precision parameters control the format for printing each matrix / vector entry (for file saving, full precision is always used).

When saving, the matrix market format is used. If both A and b are selected for saving, they will be written to the same file in extended matrix market format. The result vector is always saved in a separate file. The base file name can have path attributes but should not contain an extension. The saved file name will be: base_count[_filter_val]["_result"].mtx where base is the baseFileName, count is an integer that increments at each saved file, filter is the filter attribute name and val is the attribute value. When saving the result vector, the "_result" suffix will be added.

◆ solve()

bool GmNumSolverPrinter::solve	(	GmSolverMatrix *	A,
		const GmVector &	b,
		GmVector &	x,
		bool	xIsGuess = false ) const

Solves the linear system Ax = b, filling the vector x with the calculated result. See comments on GmNumSolver::solve().

If configured through setPrintAndSaveOptions(), matrices will be printed and / or saved to disk.

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

inc/solution/gmNumSolverPrinter.h
src/solution/gmNumSolverPrinter.cpp

Public Types

Public Member Functions