#include <opencv2/core/ocl.hpp>

Public Member Functions
	Kernel ()

	Kernel (const char *kname, const Program &prog)

	Kernel (const char kname, const ProgramSource &prog, const String &buildopts=String(), String errmsg=0)

	Kernel (const Kernel &k)

	~Kernel ()

template<typename... _Tps>
Kernel &	args (const _Tps &... kernel_args)
	Setup OpenCL Kernel arguments. Avoid direct using of set(i, ...) methods. More...

bool	compileWorkGroupSize (size_t wsz[]) const

bool	create (const char *kname, const Program &prog)

bool	create (const char kname, const ProgramSource &prog, const String &buildopts, String errmsg=0)

bool	empty () const

size_t	localMemSize () const

Kernel &	operator= (const Kernel &k)

size_t	preferedWorkGroupSizeMultiple () const

void *	ptr () const

bool	run (int dims, size_t globalsize[], size_t localsize[], bool sync, const Queue &q=Queue())
	Run the OpenCL kernel. More...

int64	runProfiling (int dims, size_t globalsize[], size_t localsize[], const Queue &q=Queue())
	Similar to synchronized run() call with returning of kernel execution time Separate OpenCL command queue may be used (with CL_QUEUE_PROFILING_ENABLE) More...

bool	runTask (bool sync, const Queue &q=Queue())

int	set (int i, const void *value, size_t sz)

int	set (int i, const Image2D &image2D)

int	set (int i, const UMat &m)

int	set (int i, const KernelArg &arg)

template<typename _Tp >
int	set (int i, const _Tp &value)

size_t	workGroupSize () const

Protected Member Functions
template<typename _Tp0 >
int	set_args_ (int i, const _Tp0 &a0)

template<typename _Tp0 , typename... _Tps>
int	set_args_ (int i, const _Tp0 &a0, const _Tps &... rest_args)

Protected Attributes
Impl *	p

Constructor & Destructor Documentation

◆ Kernel() [1/4]

cv::ocl::Kernel::Kernel ( )

◆ Kernel() [2/4]

cv::ocl::Kernel::Kernel	(	const char *	kname,
		const Program &	prog
	)

◆ Kernel() [3/4]

cv::ocl::Kernel::Kernel	(	const char *	kname,
		const ProgramSource &	prog,
		const String &	buildopts = `String()`,
		String *	errmsg = `0`
	)

◆ ~Kernel()

cv::ocl::Kernel::~Kernel ( )

◆ Kernel() [4/4]

cv::ocl::Kernel::Kernel ( const Kernel & k )

Member Function Documentation

◆ args()

template<typename... _Tps>

Kernel& cv::ocl::Kernel::args ( const _Tps &... kernel_args )

inline

Setup OpenCL Kernel arguments. Avoid direct using of set(i, ...) methods.

bool ok = kernel
    .args(
        srcUMat, dstUMat,
        (float)some_float_param
    ).run(ndims, globalSize, localSize);
if (!ok) return false;

◆ compileWorkGroupSize()

bool cv::ocl::Kernel::compileWorkGroupSize ( size_t wsz[] ) const

◆ create() [1/2]

bool cv::ocl::Kernel::create	(	const char *	kname,
		const Program &	prog
	)

◆ create() [2/2]

bool cv::ocl::Kernel::create	(	const char *	kname,
		const ProgramSource &	prog,
		const String &	buildopts,
		String *	errmsg = `0`
	)

◆ empty()

bool cv::ocl::Kernel::empty ( ) const

◆ localMemSize()

size_t cv::ocl::Kernel::localMemSize ( ) const

◆ operator=()

Kernel& cv::ocl::Kernel::operator= ( const Kernel & k )

◆ preferedWorkGroupSizeMultiple()

size_t cv::ocl::Kernel::preferedWorkGroupSizeMultiple ( ) const

◆ ptr()

void* cv::ocl::Kernel::ptr ( ) const

◆ run()

bool cv::ocl::Kernel::run	(	int	dims,
		size_t	globalsize[],
		size_t	localsize[],
		bool	sync,
		const Queue &	q = `Queue()`
	)

Run the OpenCL kernel.

Parameters

dims	the work problem dimensions. It is the length of globalsize and localsize. It can be either 1, 2 or 3.
globalsize	work items for each dimension. It is not the final globalsize passed to OpenCL. Each dimension will be adjusted to the nearest integer divisible by the corresponding value in localsize. If localsize is NULL, it will still be adjusted depending on dims. The adjusted values are greater than or equal to the original values.
localsize	work-group size for each dimension.
sync	specify whether to wait for OpenCL computation to finish before return.
q	command queue

◆ runProfiling()

int64 cv::ocl::Kernel::runProfiling	(	int	dims,
		size_t	globalsize[],
		size_t	localsize[],
		const Queue &	q = `Queue()`
	)

Similar to synchronized run() call with returning of kernel execution time Separate OpenCL command queue may be used (with CL_QUEUE_PROFILING_ENABLE)

Returns: Execution time in nanoseconds or negative number on error

◆ runTask()

bool cv::ocl::Kernel::runTask	(	bool	sync,
		const Queue &	q = `Queue()`
	)

◆ set() [1/5]

int cv::ocl::Kernel::set	(	int	i,
		const void *	value,
		size_t	sz
	)

◆ set() [2/5]

int cv::ocl::Kernel::set	(	int	i,
		const Image2D &	image2D
	)

◆ set() [3/5]

int cv::ocl::Kernel::set	(	int	i,
		const UMat &	m
	)

◆ set() [4/5]

int cv::ocl::Kernel::set	(	int	i,
		const KernelArg &	arg
	)

◆ set() [5/5]

template<typename _Tp >

int cv::ocl::Kernel::set	(	int	i,
		const _Tp &	value
	)

inline

◆ set_args_() [1/2]

template<typename _Tp0 >

int cv::ocl::Kernel::set_args_	(	int	i,
		const _Tp0 &	a0
	)

inlineprotected

◆ set_args_() [2/2]

template<typename _Tp0 , typename... _Tps>

int cv::ocl::Kernel::set_args_	(	int	i,
		const _Tp0 &	a0,
		const _Tps &...	rest_args
	)