Represents real host memory allocated for a Tensor/Blob per C type. More...

#include <ie_blob.h>

Inheritance diagram for InferenceEngine::TBlob< T, typename >:

Collaboration diagram for InferenceEngine::TBlob< T, typename >:

[legend]

Public Types
using	Ptr = std::shared_ptr< TBlob< T > >
	Smart Pointer to this TBlob object.

Public Types inherited from InferenceEngine::MemoryBlob
using	Ptr = std::shared_ptr< MemoryBlob >
	A smart pointer to the MemoryBlob object.

using	CPtr = std::shared_ptr< const MemoryBlob >
	A smart pointer to the const MemoryBlob object.

Public Types inherited from InferenceEngine::Blob
using	Ptr = std::shared_ptr< Blob >
	A smart pointer containing Blob object.

using	CPtr = std::shared_ptr< const Blob >
	A smart pointer to the const Blob object.

Public Member Functions
	TBlob (const TensorDesc &tensorDesc)
	Creates a TBlob object with the specified dimensions and layout but does not allocate the memory. More...

	TBlob (const TensorDesc &tensorDesc, T *ptr, size_t data_size=0)
	The constructor creates a TBlob object with the specified dimensions and layout on the pre-allocated memory. More...

	TBlob (const TensorDesc &tensorDesc, const std::shared_ptr< IAllocator > &alloc)
	Creates a TBlob object with the specified dimensions, layout and custom memory allocator but does not allocate the memory. More...

	TBlob (const TBlob< T > &blob)
	The copy constructor data is reallocated and copied from the source to the target blob. More...

	TBlob (TBlob< T > &&blob)
	A move constructor. More...

TBlob &	operator= (const TBlob &blob)
	Copy operator for the TBlob object. More...

virtual	~TBlob ()
	Virtual destructor.

size_t	element_size () const noexcept override
	Gets the size of the given type. More...

virtual LockedMemory< T >	data () noexcept
	Creates an new empty rvalue LockedMemory object. More...

virtual LockedMemory< const T >	readOnly () const noexcept
	Creates a new empty rvalue read-only LockedMemory object. More...

void	allocate () noexcept override
	Allocates or reallocates memory.

bool	deallocate () noexcept override
	Frees all allocated data.

LockedMemory< void >	buffer () noexcept override
	Creates a new LockedMemory instance holding void pointer. More...

LockedMemory< const void >	cbuffer () const noexcept override
	Creates a new LockedMemory instance holding constant void pointer. More...

LockedMemory< void >	rwmap () noexcept override
	Gets read/write access to the memory in virtual space of the process. The function returns object which retains mapped memory. The memory been addressed in the MemoryBlob in general case can be allocated on remote device. This function maps remote memory to the memory in the virtual process space and after destruction of the LockedMemory will upload changed content to the accelerator. More...

LockedMemory< const void >	rmap () const noexcept override
	Gets read only access to the memory in virtual space of the process. The function returns object which retains mapped memory. More...

LockedMemory< void >	wmap () noexcept override
	Gets "write only direction" access to the memory in virtual space of the process. The function returns object which retains memory to be uploaded on device. More...

Blob::Ptr	createROI (const ROI &roi) const override
	Creates a blob describing given ROI object based on the current blob with memory sharing. More...

details::BlobIterator< T >	begin ()
	Gets BlobIterator for the data. More...

details::BlobIterator< T >	end ()
	Gets BlobIterator for the end of data. More...

details::BlobIterator< const T >	begin () const
	Gets a const BlobIterator for the read-only data. More...

details::BlobIterator< const T >	end () const
	Gets a const BlobIterator for the end of read-only data. More...

Public Member Functions inherited from InferenceEngine::MemoryBlob
virtual	~MemoryBlob ()
	MemoryBlob virtual destructor.

	MemoryBlob (const TensorDesc &tensorDesc)
	Constructor. Creates an empty MemoryBlob object with the specified precision. More...

const TensorDesc &	getTensorDesc () const noexcept override
	Returns the tensor description.

TensorDesc &	getTensorDesc () noexcept override
	Returns the tensor description.

size_t	size () const noexcept override
	Returns the total number of elements, which is a product of all the dimensions. More...

size_t	byteSize () const noexcept override
	Returns the size of the current Blob in bytes calculated as `size() * element_size()`. More...

Public Member Functions inherited from InferenceEngine::Blob
virtual	~Blob ()
	Blob virtual destructor.

template<typename T , typename std::enable_if<!std::is_pointer< T >::value &&!std::is_reference< T >::value, int >::type = 0, typename std::enable_if< std::is_base_of< Blob, T >::value, int >::type = 0>
bool	is () noexcept
	Checks if the Blob object can be cast to the type T*. More...

template<typename T , typename std::enable_if<!std::is_pointer< T >::value &&!std::is_reference< T >::value, int >::type = 0, typename std::enable_if< std::is_base_of< Blob, T >::value, int >::type = 0>
bool	is () const noexcept
	Checks if the Blob object can be cast to the type const T*. More...

template<typename T , typename std::enable_if<!std::is_pointer< T >::value &&!std::is_reference< T >::value, int >::type = 0, typename std::enable_if< std::is_base_of< Blob, T >::value, int >::type = 0>
T *	as () noexcept
	Casts this Blob object to the type T*. More...

template<typename T , typename std::enable_if<!std::is_pointer< T >::value &&!std::is_reference< T >::value, int >::type = 0, typename std::enable_if< std::is_base_of< Blob, T >::value, int >::type = 0>
const T *	as () const noexcept
	Casts this Blob object to the type const T*. More...

	Blob (const TensorDesc &tensorDesc)
	Constructor. Creates an empty Blob object with the specified precision. More...

Protected Member Functions
void	copyFrom (const TBlob< T > &blob)
	Copies dimensions and data from the TBlob object. More...

template<class U >
void	moveFrom (TBlob< U > &blob)
	Swaps memory handlers between the current blob and the given one. More...

virtual bool	free ()
	Frees handler and cleans up the stored data.

template<class S >
LockedMemory< S >	lockme () const
	Creates a LockedMemory instance. More...

const std::shared_ptr< IAllocator > &	getAllocator () const noexcept override
	Gets an allocator or creates a default one. More...

void *	getHandle () const noexcept override
	Returns handle to the stored data.

	TBlob (const TBlob &origBlob, const ROI &roi)
	Creates a blob from the existing blob with a given ROI. More...

Protected Attributes
std::shared_ptr< IAllocator >	_allocator
	Local instance of IAllocator to manipulate memory.

std::shared_ptr< void >	_handle
	A handle for the stored memory returned from _allocator.alloc().

Protected Attributes inherited from InferenceEngine::Blob
TensorDesc	tensorDesc
	The tensor descriptor of the given blob.

Additional Inherited Members
Static Public Member Functions inherited from InferenceEngine::Blob
static Ptr	CreateFromData (const DataPtr &data)
	Creates a TBlob<> object from a Data node. More...

Static Protected Member Functions inherited from InferenceEngine::Blob
static size_t	product (const SizeVector &dims) noexcept
	Multiplies the dimension vector values. More...

Detailed Description

template<typename T, typename = std::enable_if<std::is_pod<T>::value>>
class InferenceEngine::TBlob< T, typename >

Represents real host memory allocated for a Tensor/Blob per C type.

Constructor & Destructor Documentation

◆ TBlob() [1/6]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

InferenceEngine::TBlob< T, typename >::TBlob ( const TensorDesc & tensorDesc )

inlineexplicit

Creates a TBlob object with the specified dimensions and layout but does not allocate the memory.

Use the allocate() method to allocate memory.

Parameters

tensorDesc Tensor description

◆ TBlob() [2/6]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

InferenceEngine::TBlob< T, typename >::TBlob	(	const TensorDesc &	tensorDesc,
		T *	ptr,
		size_t	data_size = `0`
	)

inline

The constructor creates a TBlob object with the specified dimensions and layout on the pre-allocated memory.

The allocate() call is not required.

Parameters

tensorDesc	Tensor description
ptr	Pointer to the pre-allocated memory
data_size	Length of the pre-allocated array. If not set, size is assumed equal to the dot product of dims.

◆ TBlob() [3/6]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

InferenceEngine::TBlob< T, typename >::TBlob	(	const TensorDesc &	tensorDesc,
		const std::shared_ptr< IAllocator > &	alloc
	)

inline

Creates a TBlob object with the specified dimensions, layout and custom memory allocator but does not allocate the memory.

Parameters

tensorDesc	Tensor description
alloc	An allocator

◆ TBlob() [4/6]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

InferenceEngine::TBlob< T, typename >::TBlob ( const TBlob< T > & blob )

inline

The copy constructor data is reallocated and copied from the source to the target blob.

Parameters

blob	Source blob

◆ TBlob() [5/6]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

InferenceEngine::TBlob< T, typename >::TBlob ( TBlob< T > && blob )

inline

A move constructor.

Parameters

blob	rvalue to make a move from

◆ TBlob() [6/6]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

InferenceEngine::TBlob< T, typename >::TBlob	(	const TBlob< T, typename > &	origBlob,
		const ROI &	roi
	)

inlineprotected

Creates a blob from the existing blob with a given ROI.

Parameters

origBlob	An original blob
roi	A ROI object

Member Function Documentation

◆ begin() [1/2]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

details::BlobIterator<T> InferenceEngine::TBlob< T, typename >::begin ( )

inline

Gets BlobIterator for the data.

Enables a ranged loop support for the TBlob object.

Returns: BlobIterator object of type T

◆ begin() [2/2]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

details::BlobIterator<const T> InferenceEngine::TBlob< T, typename >::begin ( ) const

inline

Gets a const BlobIterator for the read-only data.

Enables a ranged loop support for the TBlob object.

Returns: BlobIterator object of type const T

◆ buffer()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

LockedMemory<void> InferenceEngine::TBlob< T, typename >::buffer ( )

inlineoverridevirtualnoexcept

Creates a new LockedMemory instance holding void pointer.

Returns: LockedMemory instance holding void pointer

Implements InferenceEngine::MemoryBlob.

◆ cbuffer()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

LockedMemory<const void> InferenceEngine::TBlob< T, typename >::cbuffer ( ) const

inlineoverridevirtualnoexcept

Creates a new LockedMemory instance holding constant void pointer.

Returns: LockedMemory instance holding constant void pointer

Implements InferenceEngine::MemoryBlob.

◆ copyFrom()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

void InferenceEngine::TBlob< T, typename >::copyFrom ( const TBlob< T > & blob )

inlineprotected

Copies dimensions and data from the TBlob object.

Parameters

blob	object reference to copy from

◆ createROI()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

Blob::Ptr InferenceEngine::TBlob< T, typename >::createROI ( const ROI & roi ) const

inlineoverridevirtual

Creates a blob describing given ROI object based on the current blob with memory sharing.

Note: default implementation throws "not implemented" exception.

Parameters

roi	A ROI object inside of the current blob.

Returns: A shared pointer to the newly created ROI blob.

Reimplemented from InferenceEngine::Blob.

◆ data()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

virtual LockedMemory<T> InferenceEngine::TBlob< T, typename >::data ( )

inlinevirtualnoexcept

Creates an new empty rvalue LockedMemory object.

Returns: rvalue for the empty locked object of type T

◆ element_size()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

size_t InferenceEngine::TBlob< T, typename >::element_size ( ) const

inlineoverridevirtualnoexcept

Gets the size of the given type.

Returns: Size of the type

Implements InferenceEngine::MemoryBlob.

◆ end() [1/2]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

details::BlobIterator<T> InferenceEngine::TBlob< T, typename >::end ( )

inline

Gets BlobIterator for the end of data.

Enables a ranged loop support for the TBlob object.

Returns: BlobIterator object of type T representing end of the data

◆ end() [2/2]

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

details::BlobIterator<const T> InferenceEngine::TBlob< T, typename >::end ( ) const

inline

Gets a const BlobIterator for the end of read-only data.

Enables a ranged loop support for the TBlob object.

Returns: BlobIterator object of type const T representing end of data

◆ getAllocator()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

const std::shared_ptr<IAllocator>& InferenceEngine::TBlob< T, typename >::getAllocator ( ) const

inlineoverrideprotectedvirtualnoexcept

Gets an allocator or creates a default one.

Returns: IAllocator instance

Implements InferenceEngine::MemoryBlob.

◆ lockme()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

template<class S >

LockedMemory<S> InferenceEngine::TBlob< T, typename >::lockme ( ) const

inlineprotected

Creates a LockedMemory instance.

Template Parameters

S	Type of the LockedMemory to be created

Returns: A created instance of LockedMemory

◆ moveFrom()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

template<class U >

void InferenceEngine::TBlob< T, typename >::moveFrom ( TBlob< U > & blob )

inlineprotected

Swaps memory handlers between the current blob and the given one.

Template Parameters

U	Type of the blob to move from

Parameters

blob	TBlob instance to move from

◆ operator=()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

TBlob& InferenceEngine::TBlob< T, typename >::operator= ( const TBlob< T, typename > & blob )

inline

Copy operator for the TBlob object.

Parameters

blob	object reference to copy from

Returns: Newly copied object

◆ readOnly()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

virtual LockedMemory<const T> InferenceEngine::TBlob< T, typename >::readOnly ( ) const

inlinevirtualnoexcept

Creates a new empty rvalue read-only LockedMemory object.

Returns: rvalue for the empty locked const object of type T.

◆ rmap()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

LockedMemory<const void> InferenceEngine::TBlob< T, typename >::rmap ( ) const

inlineoverridevirtualnoexcept

Gets read only access to the memory in virtual space of the process. The function returns object which retains mapped memory.

The memory been addressed in the MemoryBlob in general case can be allocated on remote device. This function copies remote memory to the memory in the virtual process space and after destruction of the LockedMemory it will not upload host memory back, because it is expected that content is not changed.

To have an ability change content, you can use rwmap() and wmap() functions.

In case of memory originally allocated on the host, this function returns LockedMemory which will transparently refer to original memory address. No extra copy will happen

In general case, pointer received from that LockedMemory becomes invalid just after destruction of LockedMemory instance. Keep Locked memory alive while you need to address memory in the process on the host.

Abstract method.

Returns: A LockedMemory object

Implements InferenceEngine::MemoryBlob.

◆ rwmap()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

LockedMemory<void> InferenceEngine::TBlob< T, typename >::rwmap ( )

inlineoverridevirtualnoexcept

Gets read/write access to the memory in virtual space of the process. The function returns object which retains mapped memory. The memory been addressed in the MemoryBlob in general case can be allocated on remote device. This function maps remote memory to the memory in the virtual process space and after destruction of the LockedMemory will upload changed content to the accelerator.

To avoid extra copy of data, you can use rmap() and wmap() functions.

In case of memory originally allocated on the host, this function returns LockedMemory which will transparently refer to original memory address. No extra copy will happen

In general case, pointer received from that LockedMemory becomes invalid just after destruction of LockedMemory instance. Keep Locked memory alive while you need to address memory in the process on the host.

Abstract method.

Returns: A LockedMemory object

Implements InferenceEngine::MemoryBlob.

◆ wmap()

template<typename T , typename = std::enable_if<std::is_pod<T>::value>>

LockedMemory<void> InferenceEngine::TBlob< T, typename >::wmap ( )

inlineoverridevirtualnoexcept

Gets "write only direction" access to the memory in virtual space of the process. The function returns object which retains memory to be uploaded on device.

The memory been addressed in the MemoryBlob in general case can be allocated on remote device. This function does not copy of the content from the device to the memory in the virtual process space, the content of the memory just after calling of this function is not specified. After destruction of the LockedMemory, content will be upload host memory. In the same time there is no abilities to restrict reading from the memory, you need to care of reading from memory got by wmap(), it might have sense in some cases like filling of content and before uploading to device

To access data stored in the blob, you can use rwmap() and rmap() functions.

In case of memory originally allocated on the host, this function returns LockedMemory which will transparently refer to original memory address. No extra copy will happen

In general case, pointer received from that LockedMemory becomes invalid just after destruction of LockedMemory instance. Keep Locked memory alive while you need to address memory in the process on the host.

Abstract method.

Returns: A LockedMemory object

Implements InferenceEngine::MemoryBlob.

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

ie_blob.h

Public Types

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

template<typename T, typename = std::enable_if<std::is_pod<T>::value>> class InferenceEngine::TBlob< T, typename >

Constructor & Destructor Documentation

◆ TBlob() [1/6]

◆ TBlob() [2/6]

◆ TBlob() [3/6]

◆ TBlob() [4/6]

◆ TBlob() [5/6]

◆ TBlob() [6/6]

Member Function Documentation

◆ begin() [1/2]

◆ begin() [2/2]

◆ buffer()

◆ cbuffer()

◆ copyFrom()

◆ createROI()

◆ data()

◆ element_size()

◆ end() [1/2]

◆ end() [2/2]

◆ getAllocator()

◆ lockme()

◆ moveFrom()

◆ operator=()

◆ readOnly()

◆ rmap()

◆ rwmap()

◆ wmap()

template<typename T, typename = std::enable_if<std::is_pod<T>::value>>
class InferenceEngine::TBlob< T, typename >