Condense the data format, and possibly offer a reduced representation of coordinates, along with work units detailing which systems should be imported by each block on a given GPU. More...

#include <condensate.h>

Public Member Functions
HybridFormat	getFormat () const
	Get the memory layout of the object.

PrecisionModel	getMode () const
	Get the compression mode.

StructureSource	getBasis () const
	Get the basis for the coordinates in the object: synthesis or series.

int	getSystemCount () const
	Get the number of systems found in the condensate.

bool	ownsCoordinates () const
	Get an indication of whether the Condensate keeps its own copy of the coordinates.

size_t	getAtomOffset (int system_index) const
	Get the starting index of atoms for one of the systems, using its index.

int	getAtomCount (int system_index) const
	Get the number of atoms in one of the systems, using its index.

size_t	getCoordinateSeriesTypeID () const
	Get the data type of the CoordinateSeries upon which this object is based.

template<typename T>
const CoordinateSeries< T > *	getSeriesPointer () const
	Get a const pointer to the original coordinate series. The developer must supply the template type for the CoordinateSeries in order to re-interpret the arbitrarily stored <int> type for the CoordinateSeries pointer.

const PhaseSpaceSynthesis *	getSynthesisPointer () const
	Get a const pointer to the original coordinate synthesis.

CoordinateFrame	exportCoordinateFrame (int system_index, HybridTargetLevel tier=HybridTargetLevel::HOST) const
	Export a single frame as double-precision coordinates.

std::vector< double >	getInterlacedCoordinates (int system_index, HybridTargetLevel tier=HybridTargetLevel::HOST) const
	Produce the interlaced coordinates of a specific system. As is the case with other coordinate objects, this will not include the box transformation matrices.

const Condensate *	getSelfPointer () const
	Get a pointer to the original object, useful if a pointer is needed but the object was passed into a function by const reference.

const CondensateBorders	borders (HybridTargetLevel tier=HybridTargetLevel::HOST) const
	Get the read-only summary of the system sizes. The data is never needed in as a collection of device viewable pointers to host-side data, as this information is constant as of the creation of the object and therefore consistent on both the CPU host and GPU device.


	Condensate (HybridFormat format_in=default_hpc_format)

	Condensate (const PhaseSpaceSynthesis *poly_ps_in, PrecisionModel mode_in, const GpuDetails &gpu=null_gpu)

	Condensate (const PhaseSpaceSynthesis &poly_ps_in, PrecisionModel mode_in, const GpuDetails &gpu=null_gpu)

template<typename T>
	Condensate (const CoordinateSeries< T > *cs_in, PrecisionModel mode_in, const GpuDetails &gpu=null_gpu)

template<typename T>
	Condensate (const CoordinateSeries< T > &cs_in, PrecisionModel mode_in, const GpuDetails &gpu=null_gpu)


	Condensate (const Condensate &original)
	Copy and move constructors as well as copy and move assignment operator overloads are defined in the code, but there is no copy constructor to create objects with a different memory format as the mechanics of the Condensate are more complex than those of other coordinate objects.

	Condensate (Condensate &&original)

Condensate &	operator= (const Condensate &original)

Condensate &	operator= (Condensate &&original)


const CondensateReader	data (HybridTargetLevel tier=HybridTargetLevel::HOST) const
	Get the appropriate abstract based on the const-ness of the object.

CondensateWriter	data (HybridTargetLevel tier=HybridTargetLevel::HOST)


void	rebuild (const PhaseSpaceSynthesis *poly_ps_in, PrecisionModel mode_in=PrecisionModel::SINGLE, const GpuDetails &gpu=null_gpu)
	Rebuild the object using a different PhaseSpaceSynthesis or CoordinateSeries, and possibly a new compression mode.

void	rebuild (const PhaseSpaceSynthesis &poly_ps_in, PrecisionModel mode_in=PrecisionModel::SINGLE, const GpuDetails &gpu=null_gpu)

template<typename T>
void	rebuild (const CoordinateSeries< T > *cs_in, const PrecisionModel mode_in=PrecisionModel::SINGLE, const GpuDetails &gpu=null_gpu)


void	update (HybridTargetLevel tier=HybridTargetLevel::HOST, const GpuDetails &gpu=null_gpu)
	Reload the coordinates based on an updated coordinate synthesis or series, without re-computing the block instructions or changing the compression level.

template<typename T>
void	update (const CoordinateSeries< T > *cs_basis, HybridTargetLevel tier=HybridTargetLevel::HOST, const GpuDetails &gpu=null_gpu)

template<typename T>
void	update (const CoordinateSeries< T > &cs_basis, HybridTargetLevel tier=HybridTargetLevel::HOST, const GpuDetails &gpu=null_gpu)

Detailed Description

Condense the data format, and possibly offer a reduced representation of coordinates, along with work units detailing which systems should be imported by each block on a given GPU.

Constructor & Destructor Documentation

◆ Condensate() [1/2]

stormm::synthesis::Condensate::Condensate ( HybridFormat format_in = default_hpc_format )

The constructor requires only a complete coordinate synthesis to build the entire object. The object can be rebuilt post-creation by providing a new PhaseSpaceSynthesis, or essentially erased by providing a nullptr to conserve memory. The memory format of a Condensate is determined by the PhaseSpaceSynthesis or CoordinateSeries object it serves.

Parameters

poly_ps_in	The coordinate synthesis
mode_in	Compression mode to use
gpu	Details of the available GPU

◆ Condensate() [2/2]

stormm::synthesis::Condensate::Condensate ( const Condensate & original )

Copy and move constructors as well as copy and move assignment operator overloads are defined in the code, but there is no copy constructor to create objects with a different memory format as the mechanics of the Condensate are more complex than those of other coordinate objects.

Parameters

original	The original object to copy or move
other	Another object to fill the assignment statement's right-hand side

Member Function Documentation

◆ borders()

const CondensateBorders stormm::synthesis::Condensate::borders ( HybridTargetLevel tier = HybridTargetLevel::HOST ) const

Get the read-only summary of the system sizes. The data is never needed in as a collection of device viewable pointers to host-side data, as this information is constant as of the creation of the object and therefore consistent on both the CPU host and GPU device.

Parameters

tier	The level (host or device) at which to get the set of pointers

◆ data()

const CondensateReader stormm::synthesis::Condensate::data ( HybridTargetLevel tier = HybridTargetLevel::HOST ) const

Get the appropriate abstract based on the const-ness of the object.

Overloaded:

Get a reader for a const Condensate
Get a writer for a non-const Condensate (the writer can be quickly converted to a reader)

Parameters

tier	Obtain pointers at the level of the CPU host or the GPU device

◆ exportCoordinateFrame()

CoordinateFrame stormm::synthesis::Condensate::exportCoordinateFrame	(	int	system_index,
		HybridTargetLevel	tier = HybridTargetLevel::HOST ) const

Export a single frame as double-precision coordinates.

Parameters

system_index	Index of the system of interest
tier	Obtain information for the frame on the CPU host or the GPU device

◆ getAtomCount()

int stormm::synthesis::Condensate::getAtomCount ( int system_index ) const

Get the number of atoms in one of the systems, using its index.

Parameters

system_index Index of the system of interest

◆ getAtomOffset()

size_t stormm::synthesis::Condensate::getAtomOffset ( int system_index ) const

Get the starting index of atoms for one of the systems, using its index.

Parameters

system_index Index of the system of interest

◆ getInterlacedCoordinates()

std::vector< double > stormm::synthesis::Condensate::getInterlacedCoordinates	(	int	system_index,
		HybridTargetLevel	tier = HybridTargetLevel::HOST ) const

Produce the interlaced coordinates of a specific system. As is the case with other coordinate objects, this will not include the box transformation matrices.

Parameters

system_index	Index of the system of interest
tier	Obtain information for the frame on the CPU host or the GPU device

◆ rebuild()

void stormm::synthesis::Condensate::rebuild	(	const PhaseSpaceSynthesis *	poly_ps_in,
		PrecisionModel	mode_in = PrecisionModel::SINGLE,
		const GpuDetails &	gpu = null_gpu )

Rebuild the object using a different PhaseSpaceSynthesis or CoordinateSeries, and possibly a new compression mode.

Parameters

poly_ps_in	The new coordinate synthesis
cs_in	The new coordinate series
mode_in	Compression mode to use (data for any other mode in use will be cleared)
gpu	Details of the available GPU

◆ update()

void stormm::synthesis::Condensate::update	(	HybridTargetLevel	tier = HybridTargetLevel::HOST,
		const GpuDetails &	gpu = null_gpu )

Reload the coordinates based on an updated coordinate synthesis or series, without re-computing the block instructions or changing the compression level.

Overloaded::

Rely on internally stored pointers to the original coordinate objects
Supply a pointer or reference to the CoordinateSeries object, if applicable

Parameters

cs_basis	The original coordinate series (will be checked for validity)
tier	Obtain pointers at the level of the CPU host or the GPU device
gpu	Details of the available GPU

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

src/Synthesis/condensate.h
src/Synthesis/condensate.cpp

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ Condensate() [1/2]

◆ Condensate() [2/2]

Member Function Documentation

◆ borders()

◆ data()

◆ exportCoordinateFrame()

◆ getAtomCount()

◆ getAtomOffset()

◆ getInterlacedCoordinates()

◆ rebuild()

◆ update()