Data Structures

Classes
class	cv::cuda::BufferPool
	BufferPool for use with CUDA streams. More...
class	cv::cuda::Event
struct	cv::cuda::EventAccessor
	Class that enables getting cudaEvent_t from cuda::Event. More...
struct	cv::cuda::GpuData
class	cv::cuda::GpuMat
	Base storage class for GPU memory with reference counting. More...
class	cv::cuda::GpuMatND
class	cv::cuda::HostMem
	Class with reference counting wrapping special memory type allocation functions from CUDA. More...
class	cv::cuda::Stream
	This class encapsulates a queue of asynchronous calls. More...
struct	cv::cuda::StreamAccessor
	Class that enables getting cudaStream_t from cuda::Stream. More...

Functions
void	cv::cuda::createContinuous (int rows, int cols, int type, OutputArray arr)
	Creates a continuous matrix.
GpuMat	cv::cuda::createGpuMatFromCudaMemory (int rows, int cols, int type, size_t cudaMemoryAddress, size_t step=Mat::AUTO_STEP)
	Bindings overload to create a GpuMat from existing GPU memory.
GpuMat	cv::cuda::createGpuMatFromCudaMemory (Size size, int type, size_t cudaMemoryAddress, size_t step=Mat::AUTO_STEP)
void	cv::cuda::ensureSizeIsEnough (int rows, int cols, int type, OutputArray arr)
	Ensures that the size of a matrix is big enough and the matrix has a proper type.
void	cv::cuda::registerPageLocked (Mat &m)
	Page-locks the memory of matrix and maps it for the device(s).
void	cv::cuda::setBufferPoolConfig (int deviceId, size_t stackSize, int stackCount)
void	cv::cuda::setBufferPoolUsage (bool on)
	BufferPool management (must be called before Stream creation)
void	cv::cuda::unregisterPageLocked (Mat &m)
	Unmaps the memory of matrix and makes it pageable again.
Stream	cv::cuda::wrapStream (size_t cudaStreamMemoryAddress)
	Bindings overload to create a Stream object from the address stored in an existing CUDA Runtime API stream pointer (cudaStream_t).

Detailed Description

Function Documentation

createContinuous()

void cv::cuda::createContinuous	(	int	rows,
		int	cols,
		int	type,
		OutputArray	arr
	)

#include <opencv2/core/cuda.hpp>

Creates a continuous matrix.

Parameters

rows	Row count.
cols	Column count.
type	Type of the matrix.
arr	Destination matrix. This parameter changes only if it has a proper type and area ( \(\texttt{rows} \times \texttt{cols}\) ).

Matrix is called continuous if its elements are stored continuously, that is, without gaps at the end of each row.

createGpuMatFromCudaMemory() [1/2]

GpuMat cv::cuda::createGpuMatFromCudaMemory ( int  rows, int  cols, int  type, size_t  cudaMemoryAddress, size_t  step = Mat::AUTO_STEP  )

inline

#include <opencv2/core/cuda.hpp>

Bindings overload to create a GpuMat from existing GPU memory.

Parameters

rows	Row count.
cols	Column count.
type	Type of the matrix.
cudaMemoryAddress	Address of the allocated GPU memory on the device. This does not allocate matrix data. Instead, it just initializes the matrix header that points to the specified cudaMemoryAddress, which means that no data is copied. This operation is very efficient and can be used to process external data using OpenCV functions. The external data is not automatically deallocated, so you should take care of it.
step	Number of bytes each matrix row occupies. The value should include the padding bytes at the end of each row, if any. If the parameter is missing (set to Mat::AUTO_STEP ), no padding is assumed and the actual step is calculated as cols*elemSize(). See GpuMat::elemSize.

Note

Overload for generation of bindings only, not exported or intended for use internally from C++.

createGpuMatFromCudaMemory() [2/2]

GpuMat cv::cuda::createGpuMatFromCudaMemory ( Size  size, int  type, size_t  cudaMemoryAddress, size_t  step = Mat::AUTO_STEP  )

inline

#include <opencv2/core/cuda.hpp>

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

size	2D array size: Size(cols, rows). In the Size() constructor, the number of rows and the number of columns go in the reverse order.
type	Type of the matrix.
cudaMemoryAddress	Address of the allocated GPU memory on the device. This does not allocate matrix data. Instead, it just initializes the matrix header that points to the specified cudaMemoryAddress, which means that no data is copied. This operation is very efficient and can be used to process external data using OpenCV functions. The external data is not automatically deallocated, so you should take care of it.
step	Number of bytes each matrix row occupies. The value should include the padding bytes at the end of each row, if any. If the parameter is missing (set to Mat::AUTO_STEP ), no padding is assumed and the actual step is calculated as cols*elemSize(). See GpuMat::elemSize.

Note

Overload for generation of bindings only, not exported or intended for use internally from C++.

ensureSizeIsEnough()

void cv::cuda::ensureSizeIsEnough	(	int	rows,
		int	cols,
		int	type,
		OutputArray	arr
	)

#include <opencv2/core/cuda.hpp>

Ensures that the size of a matrix is big enough and the matrix has a proper type.

Parameters

rows	Minimum desired number of rows.
cols	Minimum desired number of columns.
type	Desired matrix type.
arr	Destination matrix.

The function does not reallocate memory if the matrix has proper attributes already.

registerPageLocked()

void cv::cuda::registerPageLocked

(

Mat &

m

)

#include <opencv2/core/cuda.hpp>

Page-locks the memory of matrix and maps it for the device(s).

Parameters

m	Input matrix.

setBufferPoolConfig()

void cv::cuda::setBufferPoolConfig	(	int	deviceId,
		size_t	stackSize,
		int	stackCount
	)

#include <opencv2/core/cuda.hpp>

setBufferPoolUsage()

void cv::cuda::setBufferPoolUsage

(

bool

on

)

#include <opencv2/core/cuda.hpp>

BufferPool management (must be called before Stream creation)

unregisterPageLocked()

void cv::cuda::unregisterPageLocked

(

Mat &

m

)

#include <opencv2/core/cuda.hpp>

Unmaps the memory of matrix and makes it pageable again.

Parameters

m	Input matrix.

wrapStream()

Stream cv::cuda::wrapStream

(

size_t

cudaStreamMemoryAddress

)

#include <opencv2/core/cuda.hpp>

Bindings overload to create a Stream object from the address stored in an existing CUDA Runtime API stream pointer (cudaStream_t).

Parameters

cudaStreamMemoryAddress

Memory address stored in a CUDA Runtime API stream pointer (cudaStream_t). The created Stream object does not perform any allocation or deallocation and simply wraps existing raw CUDA Runtime API stream pointer.

Note

Overload for generation of bindings only, not exported or intended for use internally from C++.