OperationMultipleEval for polynomial basis functions (grad >= 2) using CUDA on grids without boundary nodes. More...

#include <OperationMultiEvalCuda.hpp>

Inheritance diagram for sgpp::datadriven::OperationMultiEvalCuda:

Public Member Functions
double	getDuration () override
	Returns time in s of last mult, multTransposed, multTransposedFMA. More...

void	mult (base::DataVector &source, base::DataVector &result) override
	Standard evaluation. More...

void	multTranspose (base::DataVector &source, base::DataVector &result) override
	Transposed evaluation. More...

void	multTransposeFMA (base::DataVector &source, base::DataVector &prev, double lambda, base::DataVector &result)
	Transposed evaluation with additional FMA. result = 1/M * Bsource + lambda prev. More...

	OperationMultiEvalCuda (base::Grid &grid, base::DataMatrix &dataset, const uint32_t &grad)
	Constructor. Autocheck dataset for Morton order and turn on permutation if not. More...

	OperationMultiEvalCuda (base::Grid &grid, base::DataMatrix &dataset, const uint32_t &grad, bool autocheck)
	Constructor. Autocheck for Morton order can be disabled. More...

void	prepare () override
	Does all preprocessing for given grid and dataset and copies data to the GPU. More...

	~OperationMultiEvalCuda ()
	Destructor. More...

Public Member Functions inherited from sgpp::base::OperationMultipleEval
void	eval (DataVector &alpha, DataVector &result)
	Evaluate multiple datapoints with the specified grid. More...

virtual std::string	getImplementationName ()
	Name of this implementation of the operation. More...

virtual void	mult (DataVector &alpha, DataVector &result, size_t startIndexData, size_t endIndexData)
	Multiplication of \(B^T\) with vector \(\alpha\). More...

virtual void	multTranspose (DataVector &source, DataVector &result, size_t startIndexGrid, size_t endIndexGrid)
	Multiplication of \(B\) with vector \(\alpha\). More...

	OperationMultipleEval (sgpp::base::Grid &grid, DataMatrix &dataset)
	Constructor. More...

virtual	~OperationMultipleEval ()
	Destructor. More...

Protected Attributes
uint32_t	_M

uint32_t	_N

OpMultiEvalCudaDetail::HostDevPtr< double >	alpha

OpMultiEvalCudaDetail::HostDevPtr< double >	data

uint32_t	DIM

double	duration

OpMultiEvalCudaDetail::HostDevPtr< uint32_t >	levellimit

uint32_t	M

uint32_t	maxlevel

base::SGppStopwatch	myTimer

uint32_t	N

OpMultiEvalCudaDetail::HostDevPtr< OpMultiEvalCudaDetail::gridnode_t >	node

uint32_t	polygrad

OpMultiEvalCudaDetail::HostDevPtr< double >	pos

OpMultiEvalCudaDetail::HostDevPtr< OpMultiEvalCudaDetail::limit_t >	streamlimit

OpMultiEvalCudaDetail::HostDevPtr< uint32_t >	subs

OpMultiEvalCudaDetail::MortonOrder *	zorder

Protected Attributes inherited from sgpp::base::OperationMultipleEval
DataMatrix &	dataset

Grid &	grid

bool	isPrepared

Detailed Description

OperationMultipleEval for polynomial basis functions (grad >= 2) using CUDA on grids without boundary nodes.

Constructor & Destructor Documentation

◆ OperationMultiEvalCuda() [1/2]

sgpp::datadriven::OperationMultiEvalCuda::OperationMultiEvalCuda	(	base::Grid &	grid,
		base::DataMatrix &	dataset,
		const uint32_t &	grad
	)

Constructor. Autocheck dataset for Morton order and turn on permutation if not.

Construct the class.

References polygrad, prepare(), and zorder.

◆ OperationMultiEvalCuda() [2/2]

sgpp::datadriven::OperationMultiEvalCuda::OperationMultiEvalCuda	(	base::Grid &	grid,
		base::DataMatrix &	dataset,
		const uint32_t &	grad,
		bool	autocheck
	)

Constructor. Autocheck for Morton order can be disabled.

Construct the class.

References polygrad, prepare(), and zorder.

◆ ~OperationMultiEvalCuda()

sgpp::datadriven::OperationMultiEvalCuda::~OperationMultiEvalCuda ( )

Destructor.

Destroy everything ;)

References zorder.

Member Function Documentation

◆ getDuration()

double sgpp::datadriven::OperationMultiEvalCuda::getDuration ( )

overridevirtual

Returns time in s of last mult, multTransposed, multTransposedFMA.

Returns last measured time.

Implements sgpp::base::OperationMultipleEval.

References duration.

◆ mult()

void sgpp::datadriven::OperationMultiEvalCuda::mult	(	base::DataVector &	source,
		base::DataVector &	result
	)

overridevirtual

Standard evaluation.

Implements sgpp::base::OperationMultipleEval.

References _M, _N, alpha, data, DIM, duration, python.statsfileInfo::i, python.utils.statsfile2gnuplot::j, levellimit, M, maxlevel, myTimer, N, node, pos, sgpp::base::SGppStopwatch::start(), sgpp::base::SGppStopwatch::stop(), subs, and zorder.

◆ multTranspose()

void sgpp::datadriven::OperationMultiEvalCuda::multTranspose	(	base::DataVector &	source,
		base::DataVector &	result
	)

overridevirtual

Transposed evaluation.

Implements sgpp::base::OperationMultipleEval.

References _M, _N, alpha, data, duration, python.statsfileInfo::i, python.utils.statsfile2gnuplot::j, M, myTimer, node, pos, sgpp::base::SGppStopwatch::start(), sgpp::base::SGppStopwatch::stop(), streamlimit, and zorder.

◆ multTransposeFMA()

void sgpp::datadriven::OperationMultiEvalCuda::multTransposeFMA	(	base::DataVector &	source,
		base::DataVector &	prev,
		double	lambda,
		base::DataVector &	result
	)

Transposed evaluation with additional FMA. result = 1/M * B*source + lambda * prev.

Transposed evaluation with additional FMA.

References _M, _N, alpha, chess::b, data, duration, python.statsfileInfo::i, python.utils.statsfile2gnuplot::j, lambda, M, myTimer, N, node, pos, sgpp::base::SGppStopwatch::start(), sgpp::base::SGppStopwatch::stop(), streamlimit, and zorder.

◆ prepare()

void sgpp::datadriven::OperationMultiEvalCuda::prepare ( )

overridevirtual

Does all preprocessing for given grid and dataset and copies data to the GPU.

Preprocessing.

Reimplemented from sgpp::base::OperationMultipleEval.

References _M, _N, alpha, data, sgpp::base::OperationMultipleEval::dataset, DIM, sgpp::base::DataMatrix::get(), sgpp::base::HashGridPoint::getIndex(), sgpp::base::HashGridPoint::getLeftChild(), sgpp::base::HashGridPoint::getLevel(), sgpp::base::DataMatrix::getNrows(), sgpp::base::HashGridPoint::getRightChild(), sgpp::base::Grid::getStorage(), sgpp::base::OperationMultipleEval::grid, python.statsfileInfo::i, sgpp::base::OperationMultipleEval::isPrepared, python.utils.statsfile2gnuplot::j, level, levellimit, M, maxlevel, N, node, polygrad, pos, create_scripts::s, streamlimit, subs, and zorder.

Referenced by OperationMultiEvalCuda().

Member Data Documentation

◆ _M

uint32_t sgpp::datadriven::OperationMultiEvalCuda::_M

protected

Referenced by mult(), multTranspose(), multTransposeFMA(), and prepare().

◆ _N

uint32_t sgpp::datadriven::OperationMultiEvalCuda::_N

protected

Referenced by mult(), multTranspose(), multTransposeFMA(), and prepare().

◆ alpha

OpMultiEvalCudaDetail::HostDevPtr<double> sgpp::datadriven::OperationMultiEvalCuda::alpha

protected

Referenced by mult(), multTranspose(), multTransposeFMA(), prepare(), and python.learner.Classifier.Classifier::refineGrid().

◆ data

OpMultiEvalCudaDetail::HostDevPtr<double> sgpp::datadriven::OperationMultiEvalCuda::data

protected

Referenced by mult(), multTranspose(), multTransposeFMA(), and prepare().

◆ DIM

uint32_t sgpp::datadriven::OperationMultiEvalCuda::DIM

protected

Referenced by mult(), and prepare().

◆ duration

double sgpp::datadriven::OperationMultiEvalCuda::duration

protected

Referenced by getDuration(), mult(), multTranspose(), and multTransposeFMA().

◆ levellimit

OpMultiEvalCudaDetail::HostDevPtr<uint32_t> sgpp::datadriven::OperationMultiEvalCuda::levellimit

protected

Referenced by mult(), and prepare().

◆ M

uint32_t sgpp::datadriven::OperationMultiEvalCuda::M

protected

Referenced by mult(), multTranspose(), multTransposeFMA(), and prepare().

◆ maxlevel

uint32_t sgpp::datadriven::OperationMultiEvalCuda::maxlevel

protected

Referenced by mult(), and prepare().

◆ myTimer

base::SGppStopwatch sgpp::datadriven::OperationMultiEvalCuda::myTimer

protected

Referenced by mult(), multTranspose(), and multTransposeFMA().

◆ N

uint32_t sgpp::datadriven::OperationMultiEvalCuda::N

protected

Referenced by mult(), multTransposeFMA(), and prepare().

◆ node

OpMultiEvalCudaDetail::HostDevPtr<OpMultiEvalCudaDetail::gridnode_t> sgpp::datadriven::OperationMultiEvalCuda::node

protected

Referenced by mult(), multTranspose(), multTransposeFMA(), and prepare().

◆ polygrad

uint32_t sgpp::datadriven::OperationMultiEvalCuda::polygrad

protected

Referenced by OperationMultiEvalCuda(), and prepare().

◆ pos

OpMultiEvalCudaDetail::HostDevPtr<double> sgpp::datadriven::OperationMultiEvalCuda::pos

protected

Referenced by mult(), multTranspose(), multTransposeFMA(), and prepare().

◆ streamlimit

OpMultiEvalCudaDetail::HostDevPtr<OpMultiEvalCudaDetail::limit_t> sgpp::datadriven::OperationMultiEvalCuda::streamlimit

protected

Referenced by multTranspose(), multTransposeFMA(), and prepare().

◆ subs

OpMultiEvalCudaDetail::HostDevPtr<uint32_t> sgpp::datadriven::OperationMultiEvalCuda::subs

protected

Referenced by mult(), and prepare().

◆ zorder

OpMultiEvalCudaDetail::MortonOrder* sgpp::datadriven::OperationMultiEvalCuda::zorder

protected

Referenced by mult(), multTranspose(), multTransposeFMA(), OperationMultiEvalCuda(), prepare(), and ~OperationMultiEvalCuda().

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

datadriven/src/sgpp/datadriven/operation/hash/OperationMultiEvalCuda/OperationMultiEvalCuda.hpp
datadriven/src/sgpp/datadriven/operation/hash/OperationMultiEvalCuda/OperationMultiEvalCuda.cpp

Public Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

◆ OperationMultiEvalCuda() [1/2]

◆ OperationMultiEvalCuda() [2/2]

◆ ~OperationMultiEvalCuda()

Member Function Documentation

◆ getDuration()

◆ mult()

◆ multTranspose()

◆ multTransposeFMA()

◆ prepare()

Member Data Documentation

◆ _M

◆ _N

◆ alpha

◆ data

◆ DIM

◆ duration

◆ levellimit

◆ M

◆ maxlevel

◆ myTimer

◆ N

◆ node

◆ polygrad

◆ pos

◆ streamlimit

◆ subs

◆ zorder