xgboost::linalg::Tensor< T, kDim > Class Template Reference

A tensor storage. More...

#include <linalg.h>

Public Types
using	ShapeT = size_t[kDim]
using	StrideT = ShapeT

Public Member Functions
template<typename I , int32_t D>
	Tensor (I const (&shape)[D], std::int32_t device, Order order=kC)
	Create a tensor with shape and device ordinal.
template<typename I , size_t D>
	Tensor (common::Span< I const, D > shape, std::int32_t device, Order order=kC)
template<typename It , typename I , int32_t D>
	Tensor (It begin, It end, I const (&shape)[D], std::int32_t device, Order order=kC)
	Initialize from 2 host iterators.
template<typename I , int32_t D>
	Tensor (std::initializer_list< T > data, I const (&shape)[D], std::int32_t device, Order order=kC)
template<typename... Index>
T &	operator() (Index &&...idx)
	Index operator.
template<typename... Index>
T const &	operator() (Index &&...idx) const
	Index operator.
TensorView< T, kDim >	View (int32_t device)
	Get a TensorView for this tensor.
TensorView< T const, kDim >	View (int32_t device) const
auto	HostView () const
auto	HostView ()
size_t	Size () const
auto	Shape () const
auto	Shape (size_t i) const
HostDeviceVector< T > *	Data ()
HostDeviceVector< T > const *	Data () const
template<typename Fn >
void	ModifyInplace (Fn &&fn)
	Visitor function for modification that changes shape and data.
template<typename... S, detail::EnableIfIntegral< S... > * = nullptr>
void	Reshape (S &&...s)
	Reshape the tensor.
template<size_t D>
void	Reshape (common::Span< size_t const, D > shape)
	Reshape the tensor.
template<size_t D>
void	Reshape (size_t(&shape)[D])
template<typename... S>
auto	Slice (S &&...slices) const
	Get a host view on the slice.
template<typename... S>
auto	Slice (S &&...slices)
	Get a host view on the slice.
void	SetDevice (int32_t device) const
	Set device ordinal for this tensor.
int32_t	DeviceIdx () const

Detailed Description

template<typename T, int32_t kDim = 5>
class xgboost::linalg::Tensor< T, kDim >

A tensor storage.

To use it for other functionality like slicing one needs to obtain a view first. This way we can use it on both host and device.

Constructor & Destructor Documentation

Tensor()

template<typename T , int32_t kDim = 5>

template<typename I , int32_t D>

xgboost::linalg::Tensor< T, kDim >::Tensor ( I const (&)  shape[D], std::int32_t  device, Order  order = kC  )

inlineexplicit

Create a tensor with shape and device ordinal.

The storage is initialized automatically.

See TensorView for parameters of this constructor.

Member Function Documentation

ModifyInplace()

template<typename T , int32_t kDim = 5>

template<typename Fn >

void xgboost::linalg::Tensor< T, kDim >::ModifyInplace ( Fn &&  fn )

inline

Visitor function for modification that changes shape and data.

Template Parameters

Fn	function that takes a pointer to HostDeviceVector and a static sized span as parameters.

operator()() [1/2]

template<typename T , int32_t kDim = 5>

template<typename... Index>

T & xgboost::linalg::Tensor< T, kDim >::operator() ( Index &&...  idx )

inline

Index operator.

Not thread safe, should not be used in performance critical region. For more efficient indexing, consider getting a view first.

operator()() [2/2]

template<typename T , int32_t kDim = 5>

template<typename... Index>

T const & xgboost::linalg::Tensor< T, kDim >::operator() ( Index &&...  idx ) const

inline

Index operator.

Not thread safe, should not be used in performance critical region. For more efficient indexing, consider getting a view first.

Reshape() [1/2]

template<typename T , int32_t kDim = 5>

template<size_t D>

void xgboost::linalg::Tensor< T, kDim >::Reshape ( common::Span< size_t const, D >  shape )

inline

Reshape the tensor.

If the total size is changed, then data in this tensor is no longer valid.

Reshape() [2/2]

template<typename T , int32_t kDim = 5>

template<typename... S, detail::EnableIfIntegral< S... > * = nullptr>

void xgboost::linalg::Tensor< T, kDim >::Reshape ( S &&...  s )

inline

Reshape the tensor.

If the total size is changed, then data in this tensor is no longer valid.

---

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

• External/xgboost/include/xgboost/linalg.h