Mesh-connectivity array: ParArray<index> whose operator[] yields an AdjacencyRow typed view. More...

#include <ArrayAdjacency.hpp>

Inheritance diagram for DNDS::ArrayAdjacency< _row_size, _row_max, _align >:

Collaboration diagram for DNDS::ArrayAdjacency< _row_size, _row_max, _align >:

Classes
class	iterator
	Row iterator for ArrayAdjacency, yielding AdjacencyRow per element. More...

Public Types
using	t_self = ArrayAdjacency< _row_size, _row_max, _align >

using	t_base = ParArray< index, _row_size, _row_max, _align >

template<DeviceBackend B>
using	t_deviceView = ArrayAdjacencyDeviceView< B, index, _row_size, _row_max, _align >

template<DeviceBackend B>
using	t_deviceViewConst = ArrayAdjacencyDeviceView< B, const index, _row_size, _row_max, _align >

Public Types inherited from DNDS::ParArray< T, _row_size, _row_max, _align >
using	TArray = Array< T, _row_size, _row_max, _align >

using	t_self = ParArray< T, _row_size, _row_max, _align >

using	t_pArray = ssp< TArray >

using	t_pRowSizes = typename TArray::t_pRowSizes

Public Types inherited from DNDS::Array< T, _row_size, _row_max, _align >
using	self_type = Array< T, _row_size, _row_max, _align >

using	t_Layout = ArrayLayout< T, _row_size, _row_max, _align >

using	t_View = ArrayView< T, _row_size, _row_max, _align >

using	t_Data = host_device_vector< value_type >

using	t_DataUncompressed = std::vector< std::vector< value_type > >

using	t_RowStart = host_device_vector< index >

using	t_pRowStart = ssp< t_RowStart >

using	t_RowSizes = host_device_vector< rowsize >

using	t_pRowSizes = ssp< t_RowSizes >

template<DeviceBackend B>
using	t_deviceView = ArrayDeviceView< B, T, _row_size, _row_max, _align >

template<DeviceBackend B>
using	t_deviceViewConst = ArrayDeviceView< B, const T, _row_size, _row_max, _align >

Public Types inherited from DNDS::ArrayLayout< T, _row_size, _row_max, _align >
using	value_type = T

Public Member Functions
	ArrayAdjacency (const t_self &R)=default

t_self &	operator= (const t_self &R)=default

void	clone (const t_self &R)
	Shallow copy (same semantics as assignment).

AdjacencyRow< index >	operator[] (index i)
	Typed row access; returns an AdjacencyRow bound to the row's pointer and width. Past-the-end queries are disabled.

AdjacencyRow< const index >	operator[] (index i) const
	Const typed row access.

index *	rowPtr (index i)
	Raw `index*` pointer to row `i` (bypasses the typed wrapper).

template<DeviceBackend B>
auto	deviceView ()

template<DeviceBackend B>
auto	deviceView () const

template<DeviceBackend B>
iterator< B >	begin ()

template<DeviceBackend B>
iterator< B >	end ()

Public Member Functions inherited from DNDS::ParArray< T, _row_size, _row_max, _align >
	ParArray (const t_self &R)=default

t_self &	operator= (const t_self &R)=default

void	clone (const t_self &R)
	Copy-assign from another ParArray. Shallow copy semantics (mirrors Array clone): shares structural/data buffers.

void	WriteSerializer (Serializer::SerializerBaseSSP serializerP, const std::string &name, Serializer::ArrayGlobalOffset offset)
	Serialize (write) the parallel array with MPI-aware metadata.

void	ReadSerializer (Serializer::SerializerBaseSSP serializerP, const std::string &name, Serializer::ArrayGlobalOffset &offset)
	Deserialize (read) the parallel array with MPI-aware metadata.

MPI_Datatype	getDataType ()
	MPI element datatype used for ghost exchange (deduced from `T`).

MPI_int	getTypeMult ()
	Per-element count multiplier that goes with getDataType.

void	setMPI (const MPIInfo &n_mpi)
	Install the MPI context after default construction.

MPIInfo &	getMPI ()
	Mutable MPI context accessor.

const MPIInfo &	getMPI () const
	Read-only MPI context accessor.

void	setDataType (MPI_Datatype n_dType, MPI_int n_TypeMult)
	Override the deduced MPI datatype and element multiplier (advanced; needed for custom compound element types).

	ParArray ()=default
	Default-construct an uninitialised ParArray; call setMPI and Resize later.

	ParArray (const MPIInfo &n_mpi)
	Construct a ParArray bound to the given MPI context.

	ParArray (MPI_Datatype n_dType, MPI_int n_TypeMult, const MPIInfo &n_mpi)
	Construct with a custom (MPI datatype, multiplier) pair.

template<typename... Args>
	ParArray (ObjName objName, Args &&...args)
	Named constructor: sets the object name for tracing/debugging. All existing constructor overloads are supported via perfect forwarding. Inherited by derived classes (ArrayAdjacency, ArrayEigenVector, etc.) through `using t_base::t_base`.

void	AssertDataType ()
	Assert the MPI datatype matches `sizeof(T)` exactly.

bool	AssertConsistent ()
	Check array consistency across all ranks.

void	createGlobalMapping ()
	Collective: build the global offsets table.

index	globalSize () const
	Returns the total global size (sum of sizes across all ranks).

	Array ()=default
	Default-constructed array: empty, no storage.

	Array (ObjName objName)
	Named constructor: sets the object name for tracing/debugging. Delegates to the default constructor, then sets the name.

	Array (const self_type &R)
	Copy constructor (same semantics as clone).

Public Member Functions inherited from DNDS::Array< T, _row_size, _row_max, _align >
t_pRowStart	getRowStart ()
	Shared pointer to the row-start index (CSR layout only).

t_pRowSizes	getRowSizes ()
	Shared pointer to the per-row size vector (TABLE_Max / TABLE_StaticMax).

	Array ()=default
	Default-constructed array: empty, no storage.

	Array (ObjName objName)
	Named constructor: sets the object name for tracing/debugging. Delegates to the default constructor, then sets the name.

index	Size () const
	Number of rows currently stored. O(1).

rowsize	RowSize () const
	Uniform row width for fixed layouts (no row index needed).

rowsize	DataStride () const
	Number of T elements per row in flat storage (data stride). For TABLE_StaticMax/TABLE_Max, this is _row_max (padded), not RowSize (used). Not valid for CSR (variable stride per row).

rowsize	RowSize (index iRow) const
	Width used by row `iRow` in number of `T` elements.

rowsize	RowSizeMax () const
	Maximum allowed row width for `TABLE_Max` / `TABLE_StaticMax`.

rowsize	RowSizeField () const
	"Logical" row-field width used by derived (Eigen) arrays: max for padded layouts, uniform width for fixed layouts. Not valid for CSR.

rowsize	RowSizeField (index iRow) const
	Per-row "field" size for CSR (= actual row width). Invalid elsewhere.

bool	IfCompressed () const
	(CSR only) Whether the array is in packed / flat form.

void	CSRDecompress ()
	(CSR only) Switch to the uncompressed (nested vector) representation.

void	CSRCompress ()
	(CSR only) Pack the nested-vector representation into a flat buffer plus `_pRowStart`. No-op if already compressed.

void	Compress ()
	Layout-polymorphic compress: no-op for non-CSR, calls CSRCompress for CSR.

void	Decompress ()
	Layout-polymorphic decompress: no-op for non-CSR, calls CSRDecompress for CSR.

t_Data &	RawDataVector ()
	Access to the underlying flat buffer (`host_device_vector<T>`).

void	Resize (index nSize, rowsize nRow_size_dynamic)
	Resize the array, setting a uniform or maximum row width.

void	Resize (index nSize)
	Resize using only the row count (layouts with an implicit row width).

template<class TFRowSize >
void	Resize (index nSize, TFRowSize &&FRowSize)
	Resize a CSR array directly to the compressed form via a width functor.

void	ResizeRow (index iRow, rowsize nRowSize)
	Change the width of a single row.

void	ReserveRow (index iRow, rowsize nRowSize)
	Reserve capacity for a CSR decompressed row without changing its size.

t_View	view ()
	Produce a lightweight, device-agnostic view onto the array.

const T &	at (index iRow, rowsize iCol) const
	Bounds-checked element access.

T &	operator() (index iRow, rowsize iCol=0)
	Bounds-checked 2D element access (writable).

const T &	operator() (index iRow, rowsize iCol=0) const
	Bounds-checked 2D element access (read-only).

T *	operator[] (index iRow)
	Return a raw pointer to the start of row `iRow`.

const T *	operator[] (index iRow) const
	Const row pointer, see the non-const overload.

T *	data (DeviceBackend B=DeviceBackend::Unknown)
	Raw pointer to the flat data buffer.

size_t	DataSize () const
	Total number of `T` elements currently stored in the flat buffer.

size_t	DataSizeBytes () const
	Flat buffer size in bytes (= `DataSize() * sizeof(T)`).

void	CopyRowFrom (index dstRow, const self_type &src, index srcRow)
	Copy raw row data from another Array of the same type. Works for all layouts (StaticFixed, Fixed, CSR, etc.) and is not hidden by derived types that make operator()/operator[] private.

template<class TRowSizeFunc >
void	ResizeRowsAndCompress (TRowSizeFunc &&rowSizeFunc)
	Set per-row sizes from a source, applying an index mapping, then compress. For CSR layout only. rowSizeFunc(i) returns the desired row size for row i. Not hidden by derived types that make ResizeRow private.

size_t	FullSizeBytes () const
	Total footprint in bytes including structural arrays.

std::size_t	hash ()
	Combined hash of size, structural arrays, and data.

constexpr DataLayout	GetDataLayout ()
	Runtime accessor for the layout tag (constexpr-folded).

void	clone (const self_type &R)
	Shallow clone: copies all metadata and shares structural/data storage.

void	CopyData (const self_type &R)
	Deep copy alias. Currently delegates to clone; kept for API compatibility and to allow a future true deep-copy implementation.

self_type &	operator= (const self_type &R)
	Copy-assignment; implemented via clone with self-assign guard.

	Array (const self_type &R)
	Copy constructor (same semantics as clone).

void	SwapData (self_type &R)
	Swap the storage of two arrays in-place.

void	__WriteSerializerData (const Serializer::SerializerBaseSSP &serializerP, Serializer::ArrayGlobalOffset offset)

void	__ReadSerializerData (const Serializer::SerializerBaseSSP &serializerP, Serializer::ArrayGlobalOffset &offset)

void	WriteSerializer (Serializer::SerializerBaseSSP serializerP, const std::string &name, Serializer::ArrayGlobalOffset offset, Serializer::ArrayGlobalOffset dataOffset=Serializer::ArrayGlobalOffset_Unknown)
	Serialize (write) array data to a serializer.

void	ReadSerializer (Serializer::SerializerBaseSSP serializerP, const std::string &name, Serializer::ArrayGlobalOffset &offset)
	Convenience overload that discards the dataOffset output.

void	ReadSerializer (Serializer::SerializerBaseSSP serializerP, const std::string &name, Serializer::ArrayGlobalOffset &offset, Serializer::ArrayGlobalOffset &dataOffset)
	Deserialize (read) array data from a serializer.

ReadSerializerMetaResult	ReadSerializerMeta (Serializer::SerializerBaseSSP serializerP, const std::string &name)
	Reads only metadata from a serialized array without reading data.

void	__ReadSerializerStructuralAndResolveDataOffset (const Serializer::SerializerBaseSSP &serializerP, Serializer::ArrayGlobalOffset &offset, Serializer::ArrayGlobalOffset &dataOffset)
	Reads structural data (pRowStart / pRowSizes) and resolves dataOffset.

void	__ReadSerializerDataAndPropagateOffset (const Serializer::SerializerBaseSSP &serializerP, Serializer::ArrayGlobalOffset &offset, Serializer::ArrayGlobalOffset &dataOffset)
	Reads flat data and propagates resolved offsets back to the caller.

void	to_host ()
	Mirror the flat/structural buffers back to host memory.

void	to_device (DeviceBackend backend=DeviceBackend::Host)
	Mirror the flat/structural buffers to a target device (e.g. CUDA).

void	clear_device ()
	Release any device-side mirror of this array's buffers.

template<DeviceBackend B>
t_deviceView< B >	deviceView ()
	Mutable device-callable view (`Eigen::Map`-style row access on GPU).

template<DeviceBackend B>
t_deviceViewConst< B >	deviceView () const
	Const device-callable view. See non-const overload.

DeviceBackend	device () const
	Current device backend the data is mirrored on, or Unknown if host-only.

template<DeviceBackend B>
iterator< B >	begin ()
	Iterator to the first row, viewed on device backend `B`.

template<DeviceBackend B>
iterator< B >	end ()
	Iterator one past the last row, viewed on device backend `B`.

Public Member Functions inherited from DNDS::ObjectNaming
	ObjectNaming ()=default

	ObjectNaming (const ObjectNaming &)=default

	ObjectNaming (ObjectNaming &&)=default

ObjectNaming &	operator= (const ObjectNaming &)=default

ObjectNaming &	operator= (ObjectNaming &&)=default

	~ObjectNaming ()=default

void	setObjectName (const std::string &name)

const std::string &	getObjectName () const

std::string	getObjectIdentity (const std::string &sig) const

Additional Inherited Members
Static Public Member Functions inherited from DNDS::Array< T, _row_size, _row_max, _align >
static constexpr bool	IsCSR ()

static constexpr DataLayout	GetDataLayoutStatic ()
	Compile-time layout tag (one of `TABLE_StaticFixed`, `TABLE_Fixed`, `TABLE_StaticMax`, `TABLE_Max`, `CSR`).

Static Public Member Functions inherited from DNDS::ArrayLayout< T, _row_size, _row_max, _align >
static constexpr DataLayout	_GetDataLayout ()
	Deduce the DataLayout tag from the template parameters.

static std::string	GetArrayName ()
	Human-readable type identifier including element typeid, sizes, and alignment.

static std::string	GetArraySignature ()
	Compiler-independent identifier used by serializers to tag an array.

static std::string	GetArraySignatureRelaxed ()
	Signature with `_row_size` / `_row_max` replaced by DynamicSize.

static std::tuple< int, int, int, int >	ParseArraySignatureTuple (const std::string &v)
	Parse a signature string into `(sizeof_T, row_size, row_max, align)`.

static bool	ArraySignatureIsCompatible (const std::string &v)
	Whether a stored signature can be read into this array type.

Public Attributes inherited from DNDS::ParArray< T, _row_size, _row_max, _align >
t_pLGlobalMapping	pLGlobalMapping
	Shared pointer to the global-offsets table. Populated by createGlobalMapping; may be pointed at an existing table to skip the collective setup.

MPIInfo	mpi
	MPI context associated with this array (must be set before collectives).

Static Public Attributes inherited from DNDS::ParArray< T, _row_size, _row_max, _align >
static const DataLayout	_dataLayout = TArray::_dataLayout

Static Public Attributes inherited from DNDS::ArrayLayout< T, _row_size, _row_max, _align >
static const rowsize	al = _align

static const rowsize	rs = _row_size

static const rowsize	rm = _row_max

static const size_t	sizeof_T = sizeof(value_type)

static const rowsize	s_T = al == NoAlign ? sizeof_T : (sizeof_T / al + 1) * al

static const DataLayout	_dataLayout = _GetDataLayout()

static const bool	isCSR = _dataLayout == CSR

Protected Member Functions inherited from DNDS::Array< T, _row_size, _row_max, _align >
bool	IfCompressed_ () const

Protected Attributes inherited from DNDS::Array< T, _row_size, _row_max, _align >
t_pRowStart	_pRowStart

t_pRowSizes	_pRowSizes

t_Data	_data

DeviceBackend	deviceBackend = DeviceBackend::Unknown

t_DataUncompressed	_dataUncompressed

index	_size = 0

rowsize	_row_size_dynamic = 0

Detailed Description

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>
class DNDS::ArrayAdjacency< _row_size, _row_max, _align >

Mesh-connectivity array: ParArray<index> whose operator[] yields an AdjacencyRow typed view.

Used for cell-to-node, cell-to-cell, face-to-cell, ... every mesh connectivity table in DNDSR. Variable row widths (triangle vs quad vs hex) fall out naturally when instantiated with NonUniformSize / NonUniformSize template parameters (CSR layout); fixed-width connectivity (e.g., edges = 2 nodes) uses ArrayAdjacency<2>.

The typed operator[] returns an AdjacencyRow<index>, a lightweight span that supports size queries, range-based for, conversion to std::vector, and assignment from std::vector.

Template Parameters

_row_size	1 or a small fixed edge/face size; NonUniformSize for mixed-element connectivity.
_row_max	Ignored unless `_row_size == NonUniformSize` (see Array).

Definition at line 41 of file ArrayAdjacency.hpp.

Member Typedef Documentation

◆ t_base

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

using DNDS::ArrayAdjacency< _row_size, _row_max, _align >::t_base = ParArray<index, _row_size, _row_max, _align>

Definition at line 45 of file ArrayAdjacency.hpp.

◆ t_deviceView

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

template<DeviceBackend B>

using DNDS::ArrayAdjacency< _row_size, _row_max, _align >::t_deviceView = ArrayAdjacencyDeviceView<B, index, _row_size, _row_max, _align>

Definition at line 80 of file ArrayAdjacency.hpp.

◆ t_deviceViewConst

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

template<DeviceBackend B>

using DNDS::ArrayAdjacency< _row_size, _row_max, _align >::t_deviceViewConst = ArrayAdjacencyDeviceView<B, const index, _row_size, _row_max, _align>

Definition at line 83 of file ArrayAdjacency.hpp.

◆ t_self

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

using DNDS::ArrayAdjacency< _row_size, _row_max, _align >::t_self = ArrayAdjacency<_row_size, _row_max, _align>

Definition at line 44 of file ArrayAdjacency.hpp.

Constructor & Destructor Documentation

◆ ArrayAdjacency()

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

DNDS::ArrayAdjacency< _row_size, _row_max, _align >::ArrayAdjacency ( const t_self & R )

default

Member Function Documentation

◆ begin()

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

template<DeviceBackend B>

iterator< B > DNDS::ArrayAdjacency< _row_size, _row_max, _align >::begin ( )

inline

Definition at line 126 of file ArrayAdjacency.hpp.

◆ clone()

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

void DNDS::ArrayAdjacency< _row_size, _row_max, _align >::clone ( const t_self & R )

inline

Shallow copy (same semantics as assignment).

Definition at line 54 of file ArrayAdjacency.hpp.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ deviceView() [1/2]

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

template<DeviceBackend B>

auto DNDS::ArrayAdjacency< _row_size, _row_max, _align >::deviceView ( )

inline

Definition at line 86 of file ArrayAdjacency.hpp.

◆ deviceView() [2/2]

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

template<DeviceBackend B>

auto DNDS::ArrayAdjacency< _row_size, _row_max, _align >::deviceView ( ) const

inline

Definition at line 92 of file ArrayAdjacency.hpp.

◆ end()

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

template<DeviceBackend B>

iterator< B > DNDS::ArrayAdjacency< _row_size, _row_max, _align >::end ( )

inline

Definition at line 132 of file ArrayAdjacency.hpp.

Here is the call graph for this function:

◆ operator=()

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

t_self & DNDS::ArrayAdjacency< _row_size, _row_max, _align >::operator= ( const t_self & R )

default

Here is the caller graph for this function:

◆ operator[]() [1/2]

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

AdjacencyRow< index > DNDS::ArrayAdjacency< _row_size, _row_max, _align >::operator[] ( index i )

inline

Typed row access; returns an AdjacencyRow bound to the row's pointer and width. Past-the-end queries are disabled.

disable past-end input

Definition at line 61 of file ArrayAdjacency.hpp.

Here is the call graph for this function:

◆ operator[]() [2/2]

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

AdjacencyRow< const index > DNDS::ArrayAdjacency< _row_size, _row_max, _align >::operator[] ( index i ) const

inline

Const typed row access.

Definition at line 71 of file ArrayAdjacency.hpp.

◆ rowPtr()

template<rowsize _row_size = 1, rowsize _row_max = _row_size, rowsize _align = NoAlign>

index * DNDS::ArrayAdjacency< _row_size, _row_max, _align >::rowPtr ( index i )

inline

Raw index* pointer to row i (bypasses the typed wrapper).

Definition at line 77 of file ArrayAdjacency.hpp.

Here is the call graph for this function:

Here is the caller graph for this function:

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

/home/runner/work/DNDSR/DNDSR/src/DNDS/ArrayDerived/ArrayAdjacency.hpp

Classes

Public Types

Public Member Functions

Additional Inherited Members

Detailed Description

Member Typedef Documentation

◆ t_base

◆ t_deviceView

◆ t_deviceViewConst

◆ t_self

Constructor & Destructor Documentation

◆ ArrayAdjacency()

Member Function Documentation

◆ begin()

◆ clone()

◆ deviceView() [1/2]

◆ deviceView() [2/2]

◆ end()

◆ operator=()

◆ operator[]() [1/2]

◆ operator[]() [2/2]

◆ rowPtr()