Template Numerical Library: TNL::Containers::NDArrayView< Value, Device, Indexer

Simple data structure which provides a non-owning encapsulation of N-dimensional array data. More...

#include <TNL/Containers/NDArrayView.h>

Inheritance diagram for TNL::Containers::NDArrayView< Value, Device, Indexer >:

Collaboration diagram for TNL::Containers::NDArrayView< Value, Device, Indexer >:

Public Types
using	ConstViewType = NDArrayView< std::add_const_t< ValueType >, DeviceType, IndexerType >
	Compatible constant NDArrayView type.

using	DeviceType = Device
	Type of the device used for running operations on the array.

using	IndexerType = Indexer
	Type of the N-dimensional indexer, NDArrayIndexer.

using	IndexType = typename Indexer::IndexType
	Type of indices used for addressing the array elements.

using	OverlapsType = typename Indexer::OverlapsType
	Sequence of integers representing the overlaps in each dimension of a distributed N-dimensional array.

using	PermutationType = typename Indexer::PermutationType
	Permutation that is applied to indices when accessing the array elements.

using	SizesHolderType = typename Indexer::SizesHolderType
	Type of the underlying object which represents the sizes of the N-dimensional array.

using	StridesHolderType = typename Indexer::StridesHolderType
	Type of the base class which represents the strides of the N-dimensional array.

using	ValueType = Value
	Type of the values stored in the array.

using	ViewType = NDArrayView< ValueType, DeviceType, IndexerType >
	Compatible NDArrayView type.

Public Member Functions
__cuda_callable__	NDArrayView ()=default
	Constructs an array view with zero size.

__cuda_callable__	NDArrayView (const NDArrayView &)=default
	A shallow-copy copy-constructor.

__cuda_callable__	NDArrayView (NDArrayView &&) noexcept=default
	Move constructor for initialization from rvalues.

__cuda_callable__	NDArrayView (Value *data, IndexerType indexer)
	Constructs an array view initialized by a raw data pointer and an indexer.

__cuda_callable__	NDArrayView (Value *data, SizesHolderType sizes, StridesHolderType strides=StridesHolderType{}, OverlapsType overlaps=OverlapsType{})
	Constructs an array view initialized by a raw data pointer, sizes, strides and overlaps.

__cuda_callable__ void	bind (NDArrayView view)
	Re-binds (re-initializes) the array view to a different view.

__cuda_callable__ void	bind (Value *data)
	Re-binds (re-initializes) the array view to the given raw pointer and preserves the current indexer.

__cuda_callable__ void	bind (Value *data, IndexerType indexer)
	Re-binds (re-initializes) the array view to the given raw pointer and changes the indexer.

template<typename Device2 = DeviceType, typename Func >
void	forAll (Func f, const typename Device2::LaunchConfiguration &launch_configuration=typename Device2::LaunchConfiguration{}) const
	Evaluates the function `f` in parallel for all elements of the array view.

template<typename Device2 = DeviceType, typename SkipBegins , typename SkipEnds , typename Func >
void	forBoundary (const SkipBegins &skipBegins, const SkipEnds &skipEnds, Func f, const typename Device2::LaunchConfiguration &launch_configuration=typename Device2::LaunchConfiguration{}) const
	Evaluates the function `f` in parallel for all elements of the array view outside the N-dimensional interval `[skipBegins, skipEnds)`.

template<typename Device2 = DeviceType, typename Func >
void	forBoundary (Func f, const typename Device2::LaunchConfiguration &launch_configuration=typename Device2::LaunchConfiguration{}) const
	Evaluates the function `f` in parallel for all boundary elements of the array view.

template<typename Device2 = DeviceType, typename Begins , typename Ends , typename Func >
void	forInterior (const Begins &begins, const Ends &ends, Func f, const typename Device2::LaunchConfiguration &launch_configuration=typename Device2::LaunchConfiguration{}) const
	Evaluates the function `f` in parallel for all elements of the array view inside the N-dimensional interval `[begins, ends)`.

template<typename Device2 = DeviceType, typename Func >
void	forInterior (Func f, const typename Device2::LaunchConfiguration &launch_configuration=typename Device2::LaunchConfiguration{}) const
	Evaluates the function `f` in parallel for all internal elements of the array view.

__cuda_callable__ ConstViewType	getConstView () const
	Returns a non-modifiable view of the array.

__cuda_callable__ ValueType *	getData ()
	Returns a raw pointer to the data.

__cuda_callable__ std::add_const_t< ValueType > *	getData () const
	Returns a const-qualified raw pointer to the data.

__cuda_callable__ const IndexerType &	getIndexer () const
	Returns a const-qualified reference to the underlying indexer.

template<std::size_t... Dimensions, typename... IndexTypes>
__cuda_callable__ auto	getSubarrayView (IndexTypes &&... indices)
	Returns a modifiable view of a subarray.

__cuda_callable__ ViewType	getView ()
	Returns a modifiable view of the array.

TNL_NVCC_HD_WARNING_DISABLE __cuda_callable__ bool	operator!= (const NDArrayView &other) const
	Compares the array view with another N-dimensional array view.

template<typename... IndexTypes>
__cuda_callable__ ValueType &	operator() (IndexTypes &&... indices)
	Accesses an element of the array.

template<typename... IndexTypes>
__cuda_callable__ const ValueType &	operator() (IndexTypes &&... indices) const
	Accesses an element of the array.

TNL_NVCC_HD_WARNING_DISABLE __cuda_callable__ NDArrayView &	operator= (const NDArrayView &other)
	Copy-assignment operator for deep-copying data from another array.

template<typename OtherView >
TNL_NVCC_HD_WARNING_DISABLE __cuda_callable__ NDArrayView &	operator= (const OtherView &other)
	Templated copy-assignment operator for deep-copying data from another array.

TNL_NVCC_HD_WARNING_DISABLE __cuda_callable__ bool	operator== (const NDArrayView &other) const
	Compares the array view with another N-dimensional array view.

__cuda_callable__ ValueType &	operator[] (IndexType &&index)
	Accesses an element in a one-dimensional array.

__cuda_callable__ const ValueType &	operator[] (IndexType index) const
	Accesses an element in a one-dimensional array.

__cuda_callable__ void	reset ()
	Resets the array view to the empty state.

Protected Attributes
Value *	array = nullptr
	Pointer to the first element of the bound N-dimensional array.

IndexerType	indexer
	Object which transforms the multi-dimensional indices to a one-dimensional index.

Detailed Description

template<typename Value, typename Device, typename Indexer>
class TNL::Containers::NDArrayView< Value, Device, Indexer >

Simple data structure which provides a non-owning encapsulation of N-dimensional array data.

Template Parameters

Value	Type of the values stored in the array.
Device	Type of the device that will be used for running operations on the array.
Type	of the N-dimensional indexer, NDArrayIndexer.

Constructor & Destructor Documentation

◆ NDArrayView()

template<typename Value , typename Device , typename Indexer >

__cuda_callable__ TNL::Containers::NDArrayView< Value, Device, Indexer >::NDArrayView ( const NDArrayView< Value, Device, Indexer > & )

default

A shallow-copy copy-constructor.

This allows views to be passed-by-value into CUDA kernels and be captured-by-value in cuda_callable lambda functions.

Member Function Documentation

◆ forAll()

template<typename Value , typename Device , typename Indexer >

template<typename Device2 = DeviceType, typename Func >

void TNL::Containers::NDArrayView< Value, Device, Indexer >::forAll	(	Func	f,
		const typename Device2::LaunchConfiguration &	launch_configuration = typename Device2::LaunchConfiguration{} ) const

inline

Evaluates the function f in parallel for all elements of the array view.

The function f is called as f(indices...), where indices... are substituted by the actual indices of all array elements. For example, given a 3D array view a with int as IndexType, the function can be defined and used as follows:

auto setter = [&a] ( int i, int j, int k )
{
   a( i, j, k ) = 1;
};
a.forAll( setter );

◆ forBoundary() [1/2]

template<typename Value , typename Device , typename Indexer >

template<typename Device2 = DeviceType, typename SkipBegins , typename SkipEnds , typename Func >

void TNL::Containers::NDArrayView< Value, Device, Indexer >::forBoundary	(	const SkipBegins &	skipBegins,
		const SkipEnds &	skipEnds,
		Func	f,
		const typename Device2::LaunchConfiguration &	launch_configuration = typename Device2::LaunchConfiguration{} ) const

inline

Evaluates the function f in parallel for all elements of the array view outside the N-dimensional interval [skipBegins, skipEnds).

The function f is called as f(indices...), where indices... are substituted by the actual indices of all array view elements.

◆ forBoundary() [2/2]

template<typename Value , typename Device , typename Indexer >

template<typename Device2 = DeviceType, typename Func >

void TNL::Containers::NDArrayView< Value, Device, Indexer >::forBoundary	(	Func	f,
		const typename Device2::LaunchConfiguration &	launch_configuration = typename Device2::LaunchConfiguration{} ) const

inline

Evaluates the function f in parallel for all boundary elements of the array view.

The function f is called as f(indices...), where indices... are substituted by the actual indices of all array view elements. For example, given a 3D array view a with int as IndexType, the function can be defined and used as follows:

auto setter = [&a] ( int i, int j, int k )
{
   a( i, j, k ) = 1;
};
a.forBoundary( setter );

◆ forInterior() [1/2]

template<typename Value , typename Device , typename Indexer >

template<typename Device2 = DeviceType, typename Begins , typename Ends , typename Func >

void TNL::Containers::NDArrayView< Value, Device, Indexer >::forInterior	(	const Begins &	begins,
		const Ends &	ends,
		Func	f,
		const typename Device2::LaunchConfiguration &	launch_configuration = typename Device2::LaunchConfiguration{} ) const

inline

Evaluates the function f in parallel for all elements of the array view inside the N-dimensional interval [begins, ends).

The function f is called as f(indices...), where indices... are substituted by the actual indices of all array view elements.

◆ forInterior() [2/2]

template<typename Value , typename Device , typename Indexer >

template<typename Device2 = DeviceType, typename Func >

void TNL::Containers::NDArrayView< Value, Device, Indexer >::forInterior	(	Func	f,
		const typename Device2::LaunchConfiguration &	launch_configuration = typename Device2::LaunchConfiguration{} ) const

inline

Evaluates the function f in parallel for all internal elements of the array view.

auto setter = [&a] ( int i, int j, int k )
{
   a( i, j, k ) = 1;
};
a.forInterior( setter );

◆ getSubarrayView()

template<typename Value , typename Device , typename Indexer >

template<std::size_t... Dimensions, typename... IndexTypes>

__cuda_callable__ auto TNL::Containers::NDArrayView< Value, Device, Indexer >::getSubarrayView ( IndexTypes &&... indices )

inlinenodiscard

Returns a modifiable view of a subarray.

Template Parameters

Dimensions Sequence of integers representing the dimensions of the array which selects the subset of dimensions to appear in the subarray.

Parameters

indices Indices of the origin of the subarray in the whole array. The number of indices supplied must be equal to N, i.e. getDimension().

Returns: NDArrayView instantiated for ValueType, DeviceType and an NDArrayIndexer matching the specified dimensions and subarray sizes.

◆ operator()() [1/2]

template<typename Value , typename Device , typename Indexer >

template<typename... IndexTypes>

__cuda_callable__ ValueType & TNL::Containers::NDArrayView< Value, Device, Indexer >::operator() ( IndexTypes &&... indices )

inlinenodiscard

Accesses an element of the array.

Parameters

indices Indices of the element in the N-dimensional array. The number of indices supplied must be equal to N, i.e. getDimension().

Returns: Reference to the array element.

◆ operator()() [2/2]

template<typename Value , typename Device , typename Indexer >

template<typename... IndexTypes>

__cuda_callable__ const ValueType & TNL::Containers::NDArrayView< Value, Device, Indexer >::operator() ( IndexTypes &&... indices ) const

inlinenodiscard

Accesses an element of the array.

Parameters

indices Indices of the element in the N-dimensional array. The number of indices supplied must be equal to N, i.e. getDimension().

Returns: Constant reference to the array element.

◆ operator=()

template<typename Value , typename Device , typename Indexer >

TNL_NVCC_HD_WARNING_DISABLE __cuda_callable__ NDArrayView & TNL::Containers::NDArrayView< Value, Device, Indexer >::operator= ( const NDArrayView< Value, Device, Indexer > & other )

inline

Copy-assignment operator for deep-copying data from another array.

This is just like the operator on a regular array, but the sizes must match (i.e. copy-assignment cannot resize).

Note that there is no move-assignment operator, so expressions like a = b.getView() are resolved as copy-assignment.

◆ operator[]() [1/2]

template<typename Value , typename Device , typename Indexer >

__cuda_callable__ ValueType & TNL::Containers::NDArrayView< Value, Device, Indexer >::operator[] ( IndexType && index )

inlinenodiscard

Accesses an element in a one-dimensional array.

Warning: This function can be used only when the dimension of the array is equal to 1.

Parameters

index Index of the element in the one-dimensional array.

Returns: Reference to the array element.

◆ operator[]() [2/2]

template<typename Value , typename Device , typename Indexer >

__cuda_callable__ const ValueType & TNL::Containers::NDArrayView< Value, Device, Indexer >::operator[] ( IndexType index ) const

inlinenodiscard

Accesses an element in a one-dimensional array.

Warning: This function can be used only when the dimension of the array is equal to 1.

Parameters

index Index of the element in the one-dimensional array.

Returns: Constant reference to the array element.

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

src/TNL/Containers/NDArrayView.h

Public Types

Public Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

◆ NDArrayView()

Member Function Documentation

◆ forAll()

◆ forBoundary() [1/2]

◆ forBoundary() [2/2]

◆ forInterior() [1/2]

◆ forInterior() [2/2]

◆ getSubarrayView()

◆ operator()() [1/2]

◆ operator()() [2/2]

◆ operator=()

◆ operator[]() [1/2]

◆ operator[]() [2/2]