nvneural::ScopedMemoryAllocation Class Reference

ScopedMemoryAllocation is a helper function for allocating memory blocks from the backend while adding the security and ease of use of the RAII idiom. More...

#include <ScopedMemoryAllocation.h>

Public Member Functions
void *	asPointer ()
	Returns the pointer to the allocated memory block. More...
MemoryHandle	handle () const
	Returns the memory handle of the allocated memory block. More...
	operator bool () const
	Returns true if the memory allocation is valid (non-null handle).
ScopedMemoryAllocation &	operator= (const ScopedMemoryAllocation &copyFrom)=delete
ScopedMemoryAllocation &	operator= (ScopedMemoryAllocation &&moveFrom)
	Takes ownership of an allocation from another ScopedMemoryAllocation. More...
	ScopedMemoryAllocation (const ScopedMemoryAllocation &copyFrom)=delete
	ScopedMemoryAllocation (INetworkBackend *pBackend, size_t bufferSize)
	Creates a temporary memory allocation using a network backend. More...
	ScopedMemoryAllocation (INetworkBackend *pBackend, size_t bufferSize, const char *pTrackingKey)
	Creates a temporary memory allocation, using a network backend, of the requested size and allows tracking of the memory block using a user-defined key. More...
	ScopedMemoryAllocation (ScopedMemoryAllocation &&moveFrom)
	This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.
	~ScopedMemoryAllocation ()
	Frees the allocation.

Detailed Description

ScopedMemoryAllocation is a helper function for allocating memory blocks from the backend while adding the security and ease of use of the RAII idiom.

Usage notes: This is meant to be used only temporarily within a particular block of code. The class doesn't take a reference to the network backend. Check handle() for nullptr, which is the failure of allocation. Memory is freed automatically once out of scope.

Constructor & Destructor Documentation

ScopedMemoryAllocation() [1/2]

nvneural::ScopedMemoryAllocation::ScopedMemoryAllocation ( INetworkBackend *  pBackend, size_t  bufferSize  )

inline

Creates a temporary memory allocation using a network backend.

The memory allocated is in the backend's address space (cuDevicePtr for CUDA).

Parameters

pBackend	Backend performing the allocation
bufferSize	Size of buffer to request, in bytes

ScopedMemoryAllocation() [2/2]

nvneural::ScopedMemoryAllocation::ScopedMemoryAllocation ( INetworkBackend *  pBackend, size_t  bufferSize, const char *  pTrackingKey  )

inline

Creates a temporary memory allocation, using a network backend, of the requested size and allows tracking of the memory block using a user-defined key.

It is best practice to use the layer's name at this key. The memory allocated is in the backend's address space (cuDevicePtr for CUDA).

Parameters

pBackend	Backend performing the allocation
bufferSize	Size of buffer to request, in bytes
pTrackingKey	A string key, usually the layer's name, which allows tracking of this allocation's information, nullptr to disable

Member Function Documentation

asPointer()

void* nvneural::ScopedMemoryAllocation::asPointer ( )

inline

Returns the pointer to the allocated memory block.

Returns

A void pointer to the memory block.

handle()

MemoryHandle nvneural::ScopedMemoryAllocation::handle ( ) const

inline

Returns the memory handle of the allocated memory block.

Returns

The memory handle, nullptr indicates a failed allocation.

operator=()

ScopedMemoryAllocation& nvneural::ScopedMemoryAllocation::operator= ( ScopedMemoryAllocation &&  moveFrom )

inline

Takes ownership of an allocation from another ScopedMemoryAllocation.

Parameters

moveFrom

rvalue reference to the source allocation object

---

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

• Helpers/Inc/ScopedMemoryAllocation.h