VectorImpl< T, VectorType > Class Template Reference

Base class template for various DataVector and related types. More...

#include <VectorImpl.hpp>

Public Types
using	value_type = T
using	size_type = size_t
using	difference_type = std::ptrdiff_t
using	BaseType = blaze::CustomVector< T, blaze::AlignmentFlag::unaligned, blaze::PaddingFlag::unpadded, blaze::defaultTransposeFlag, blaze_default_group, VectorType >
using	ElementType = T
using	TransposeType = VectorImpl< T, VectorType >
using	CompositeType = const VectorImpl< T, VectorType > &
using	iterator = typename BaseType::Iterator
using	const_iterator = typename BaseType::ConstIterator

Public Member Functions
	VectorImpl (size_t set_size) noexcept
	Create with the given size. In debug mode, the vector is initialized to 'NaN' by default. If not initialized to 'NaN', the memory is allocated but not initialized. More...
	VectorImpl (size_t set_size, T value) noexcept
	Create with the given size and value. More...
	VectorImpl (T *start, size_t set_size) noexcept
	Create a non-owning VectorImpl that points to start
template<class U , Requires< std::is_same_v< U, T > > = nullptr>
	VectorImpl (std::initializer_list< U > list) noexcept
	Create from an initializer list of T.
	VectorImpl ()=default
	Empty VectorImpl.
VectorImpl &	operator= (const T &rhs) noexcept
void	destructive_resize (const size_t new_size) noexcept
	A common operation for checking the size and resizing a memory buffer if needed to ensure that it has the desired size. This operation is not permitted on a non-owning vector. More...
bool	is_owning () const noexcept
	Returns true if the class owns the data.
void	pup (PUP::er &p) noexcept
	Serialization for Charm++.
template<typename T , typename VectorType >
	VectorImpl (const VectorImpl< T, VectorType > &rhs) noexcept
template<typename T , typename VectorType >
	VectorImpl (VectorImpl< T, VectorType > &&rhs) noexcept
const BaseType &	operator* () const noexcept
	Upcast to BaseType More...
BaseType &	operator* () noexcept
	Upcast to BaseType More...
void	set_data_ref (gsl::not_null< VectorType * > rhs) noexcept
	Set the VectorImpl to be a reference to another VectorImpl object.
void	set_data_ref (T *const start, const size_t set_size) noexcept
	Set the VectorImpl to be a reference to another VectorImpl object.

Static Public Attributes
static constexpr bool	transpose_flag = blaze::defaultTransposeFlag

Protected Member Functions
void	reset_pointer_vector (const size_t set_size) noexcept

Protected Attributes
std::unique_ptr< value_type[]>	owned_data_ {}
bool	owning_ {true}

Detailed Description

template<typename T, typename VectorType>
class VectorImpl< T, VectorType >

Base class template for various DataVector and related types.

Details

The VectorImpl class is the generic parent class for vectors representing collections of related function values, such as DataVectors for contiguous data over a computational domain.

The VectorImpl does not itself define any particular mathematical operations on the contained values. The VectorImpl template class and the macros defined in VectorImpl.hpp assist in the construction of various derived classes supporting a chosen set of mathematical operations.

In addition, the equivalence operator == is inherited from the underlying blaze::CustomVector type, and returns true if and only if the size and contents of the two compared vectors are equivalent.

Template parameters:

• T is the underlying stored type, e.g. double, std::complex<double>, float, etc.
• VectorType is the type that should be associated with the VectorImpl during mathematical computations. In most cases, inherited types should have themselves as the second template argument, e.g.

  class DataVector : VectorImpl<double, DataVector> {

  The second template parameter communicates arithmetic type restrictions to the underlying Blaze framework. For example, if VectorType is DataVector, then the underlying architecture will prevent addition with a vector type whose ResultType (which is aliased to its VectorType) is ModalVector. Since DataVectors and ModalVectors represent data in different spaces, we wish to forbid several operations between them. This vector-type-tracking through an expression prevents accidental mixing of vector types in math expressions.

Note

• If created with size 0, then data() will return nullptr
• If either SPECTRE_DEBUG or SPECTRE_NAN_INIT are defined, then the VectorImpl is default initialized to signaling_NaN(). Otherwise, the vector is filled with uninitialized memory for performance.

Constructor & Destructor Documentation

VectorImpl() [1/2]

template<typename T , typename VectorType >

VectorImpl< T, VectorType >::VectorImpl ( size_t  set_size )

inlineexplicitnoexcept

Create with the given size. In debug mode, the vector is initialized to 'NaN' by default. If not initialized to 'NaN', the memory is allocated but not initialized.

• set_size number of values

VectorImpl() [2/2]

template<typename T , typename VectorType >

VectorImpl< T, VectorType >::VectorImpl ( size_t  set_size, T  value  )

inlinenoexcept

Create with the given size and value.

• set_size number of values
• value the value to initialize each element

Member Function Documentation

destructive_resize()

template<typename T , typename VectorType >

void VectorImpl< T, VectorType >::destructive_resize ( const size_t  new_size )

inlinenoexcept

A common operation for checking the size and resizing a memory buffer if needed to ensure that it has the desired size. This operation is not permitted on a non-owning vector.

Note

This utility should NOT be used when it is anticipated that the supplied buffer will typically be the wrong size (in that case, suggest either manual checking or restructuring so that resizing is less common). This uses UNLIKELY to perform the check most quickly when the buffer needs no resizing, but will be slower when resizing is common.

operator*() [1/2]

template<typename T , typename VectorType >

const BaseType & VectorImpl< T, VectorType >::operator* ( ) const

inlinenoexcept

Upcast to BaseType

Attention

upcast should only be used when implementing a derived vector type, not in calling code

operator*() [2/2]

template<typename T , typename VectorType >

BaseType & VectorImpl< T, VectorType >::operator* ( )

inlinenoexcept

Upcast to BaseType

Attention

upcast should only be used when implementing a derived vector type, not in calling code