A handler gives user access to command group scope functionality, such as API calls. More...

#include <apis.h>

Public Member Functions
COMPUTECPP_TEST_VIRTUAL	~handler ()
	Destructor of the handler, implementation on the cpp file so the default_deleter can see the implementation of the internal transaction object. More...

template<typename T >
void	set_arg (int param_num, T &&param)
	Sets an argument when using interop kernels. More...

template<typename... Ts>
void	set_args (Ts &&... args)
	Set all the given kernel args arguments for an OpenCL kernel, as if set_arg() was used with each of them in the same order and increasing index always starting at 0. More...

void	single_task (kernel syclKernel)
	This function effectively just launches a single thread to execute the kernel in serial asynchronously to the host execution. More...

template<typename nameT = std::nullptr_t, typename functorT >
void	single_task (const functorT &functor)
	This function effectively just launches a single thread to execute the kernel in serial asynchronously to the host execution. More...

template<typename nameT = std::nullptr_t, typename functorT >
void	single_task (kernel syclKernel, const functorT &functor)
	This function effectively just launches a single thread to execute the kernel in serial asynchronously to the host execution. More...

template<int dimensions>
void	parallel_for (const nd_range< dimensions > &ndRange, kernel syclKernel)
	Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by ndRange. More...

template<int dimensions>
void	parallel_for (const range< dimensions > &range, kernel syclKernel)
	Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of global work items specified by range. More...

template<int dimensions>
void	parallel_for (const range< dimensions > &range, id< dimensions > offset, kernel syclKernel)
	Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of global work items specified by range. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for (const nd_range< dimensions > &ndRange, const functorT &functor)
	Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of local and global work items specified by ndRange. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for (kernel syclKernel, const nd_range< dimensions > &ndRange, const functorT &functor)
	Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of local and global work items specified by ndRange. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for (const range< dimensions > &range, const functorT &functor)
	Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of global work items specified by range. More...

template<typename nameT = std::nullptr_t, typename functorT >
void	parallel_for (const size_t range, const functorT &functor)
	Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of global work items specified by range. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for (const range< dimensions > &range, const id< dimensions > &offset, const functorT &functor)
	Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of global work items specified by range. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for (kernel syclKernel, const range< dimensions > &range, const functorT &functor)
	Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of global work items specified by range. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for (kernel syclKernel, const range< dimensions > &range, const id< dimensions > &offset, const functorT &functor)
	Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of global work items specified by range. More...

template<int dimensions>
void	parallel_for_work_group (const range< dimensions > &numGroups, kernel syclKernel)
	parallel_for_work_group will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by numGroups. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for_work_group (kernel syclKernel, const range< dimensions > &range, const functorT &functor)
	parallel_for_work_group will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by range. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for_work_group (const range< dimensions > &range, const functorT &functor)
	parallel_for_work_group will enqueue the kernel functor to be executed a number of instances working in parallel over the number of local and global work items specified by range. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for_work_group (const range< dimensions > &numGroups, const range< dimensions > &groupSize, const functorT &functor)
	parallel_for_work_group will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by numGroups and groupSize. More...

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>
void	parallel_for_work_group (kernel syclKernel, const range< dimensions > &numGroups, const range< dimensions > &groupSize, const functorT &functor)
	parallel_for_work_group will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by numGroups and groupSize. More...

template<typename elemT , int kDims, access::mode kMode, access::target kTarget, access::placeholder IsPlaceholder = access::placeholder::true_t>
void	require (const accessor< elemT, kDims, kMode, kTarget, IsPlaceholder > &acc)
	Function that registers a placeholder accessor with the handler. More...

template<typename elemT , int kDims, access::mode kMode, access::placeholder isPlaceholder>
void	register_for_dma (accessor< elemT, kDims, kMode, COMPUTECPP_ACCESS_TARGET_DEVICE, isPlaceholder > &acc, size_t stride)
	Registers a global memory accessor for DMA transfer. More...

template<typename elemT , int kDims, access::placeholder isPlaceholder>
void	register_for_dma (accessor< elemT, kDims, access::mode::read, COMPUTECPP_ACCESS_TARGET_DEVICE, isPlaceholder > &acc, size_t stride)
	Registers a constant memory accessor for DMA transfer. More...

template<typename elemT , int kDims, access::mode kMode, access::target kTarget>
	COMPUTECPP_DEPRECATED_API ("Deprecated Codeplay extension function: " "Bind the null accessor first, then call require()") void require(buffer< elemT
	Function that registers a placeholder accessor with the handler and the associated storage. More...

void	experimental_depends_on (cl::sycl::event e)
	Register a single event that this handler should wait for before running. More...

void	experimental_depends_on (const std::vector< cl::sycl::event > &v)
	Register a set of events that this handler should wait for before running. More...

void	depends_on (cl::sycl::event e)
	Register a single event that this handler should wait for before running. More...

void	depends_on (const std::vector< cl::sycl::event > &v)
	Register a set of events that this handler should wait for before running. More...

template<typename TAcc , typename THostPtr , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF( TAcc,(detail::can_copy_types< TAcc, THostPtr >::value && detail::is_read_mode< accessMode >::value)) >
void	copy (accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder > acc, shared_ptr_class< THostPtr > hostPtr)
	Copies the data from the device accessor to the host pointer. More...

template<typename TAcc , typename THostPtr , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF( TAcc,(detail::can_copy_types< THostPtr, TAcc >::value && detail::is_write_mode< accessMode >::value)) >
void	copy (shared_ptr_class< THostPtr > hostPtr, accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder > acc)
	Copies the data from the host pointer to the device accessor. More...

template<typename TAcc , typename THostPtr , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF( TAcc,(detail::can_copy_types< TAcc, THostPtr >::value && detail::is_read_mode< accessMode >::value)) >
void	copy (accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder > acc, THostPtr *hostPtr)
	Copies the data from the device accessor to the host pointer. More...

template<typename TAcc , typename THostPtr , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF( TAcc,(detail::can_copy_types< THostPtr, TAcc >::value && detail::is_write_mode< accessMode >::value)) >
void	copy (const THostPtr *hostPtr, accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder > acc)
	Copies the data from the host pointer to the device accessor. More...

template<typename T , typename U , int dimsOrig, int dimsDest, access::mode accModeOrig, access::mode accModeDest, access::target accTargetOrig, access::target accTargetDest, access::placeholder isPlaceholderOrig, access::placeholder isPlaceholderDest, COMPUTECPP_ENABLE_IF(T,((detail::can_copy_types< T, U >::value) &&(detail::is_read_mode< accModeOrig >::value) &&(detail::is_write_mode< accModeDest >::value))) >
void	copy (accessor< T, dimsOrig, accModeOrig, accTargetOrig, isPlaceholderOrig > originAcc, accessor< U, dimsDest, accModeDest, accTargetDest, isPlaceholderDest > destinationAcc)
	Copies data associated with the origin accessor to the data associated with the destination accessor. More...

template<typename TAcc , typename T , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF(TAcc,(detail::can_copy_types< T, TAcc >::value && detail::is_write_mode< accessMode >::value)) >
void	fill (accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder > acc, T val)
	Fills the data associated with the accessor using the scalar value. More...

template<typename T , int dims, access::mode accessMode, access::target accessTarget, access::placeholder isPlaceholder>
void	update_host (accessor< T, dims, accessMode, accessTarget, isPlaceholder > acc)
	Update the memory object accessed by a given accessor on the host. More...

template<typename T >
void	fill (void *ptr, const T &pattern, size_t count)
	Fills the memory pointed by `ptr`. More...

void	memcpy (void dest, const void src, size_t size)
	Copies `count` bytes from `src` to . More...

void	experimental_prefetch (const void *ptr, size_t size)
	Hints to the SYCL runtime that the data is available earlier than when the USM model would require it. More...

void	prefetch (const void *ptr, size_t size)
	Hints to the SYCL runtime that the data is available earlier than when the USM model would require it. More...

Public Attributes
kDims &	bufObj

kDims const accessor< elemT, kDims, kMode, kTarget, access::placeholder::true_t > &	acc

Protected Member Functions
void	use_kernel_bundle_impl (const dkernelbundle_shptr execBundle)
	This command group will use device images from the given kernel bundle when invoking kernels. More...

void	update_device_data (const accessor_base &acc, shared_ptr_class< void > hostPtr, cl::sycl::access::mode accessMode, bool userProvidedPtr)
	Updates device data by copying to/from the device. More...

void	update_host_impl (const accessor_base &acc)
	Implementation for update_host. More...

void	copy_in_device (const accessor_base &originAcc, const accessor_base &destinationAcc)
	Copies data associated with the origin accessor to the data associated with the destination accessor. More...

void	fill (const accessor_base &acc, const void *patternData, const size_t patternSize)
	Fills the range of the accessor with value of hostScalarPtr[0]}. More...

void	fill (void ptr, const void patternData, size_t patternSize, size_t size)
	Fills the memory pointed by `ptr`. More...

void	interop_task_impl (const detail::interop_task_ptr &hostTaskCallable)
	Schedules a host task with an interop_handle object. More...

	handler (const dqueue_shptr &q, const dqueue_shptr &fallbackQueue=nullptr)
	Creates a handler for an specific queue. More...

context	get_context () const
	Returns the current context for the command group. More...

ddevice_wkptr	get_device_weak () const
	Returns the current device for the command group. More...

void	execute_kernel_single_task_ptr (const detail::nd_range_base &ndRange, const kernel &syclKernel, const detail::single_task_ptr &funcPtr, detail::enqueue_device_kernel_command *currentCommand)
	Creates the internal structures to execute the kernel. More...

void	execute_kernel_parallel_for_ptr (const detail::nd_range_base &ndRange, const kernel &syclKernel, const detail::parallel_for_ptr &funcPtr, detail::enqueue_device_kernel_command *currentCommand, int dimensions)

void	execute_kernel_parallel_for_id_ptr (const detail::nd_range_base &ndRange, const kernel &syclKernel, const detail::parallel_for_id_ptr &funcPtr, detail::enqueue_device_kernel_command *currentCommand, int dimensions)

void	execute_kernel_parallel_for_work_group_ptr (const detail::nd_range_base &ndRange, const kernel &syclKernel, const detail::parallel_for_work_group_ptr &funcPtr, detail::enqueue_device_kernel_command *currentCommand, int dimensions)

void	process_functor_arguments_impl (kernel syclKernel, detail::binary_address functorBuffer, const detail::functor_arg_descriptor &argDesc, detail::enqueue_device_kernel_command *currentCommand)
	Gets the parameters from a functor and sets them as OpenCL arguments. More...

void	require (const accessor_base &acc)
	Internal function that registers a placeholder accessor with the handler. More...

void	require (const accessor_base &acc, dmem_shptr memObj, access::mode mode, access::target target)
	Internal function that registers a placeholder accessor with the handler. More...

void	register_for_dma (accessor_base &acc, size_t strideBytes)
	Registers an accessor for DMA transfer. More...

detail::enqueue_device_kernel_command *	create_kernel_command_group (kernel &syclKernel)
	Creates a kernel command group Internal implementation. More...

detail::index_array	get_optimal_workgroup_size (const kernel &syclKernel)
	Gets the optimal workgroup size for the current device and the given kernel. More...

bool	has_kernel_bundle () const
	Returns true if use_kernel_bundle(bundle) has been called with the handler. More...

program	get_kernel_bundle_program () const
	Get the program to use from a kernel bundle. More...

Protected Attributes
dtrans_uptr	m_trans
	Internal transaction associated with the handler. More...

dqueue_shptr	m_queue
	Queue to which this handler was submitted to. More...

dqueue_shptr	m_fallbackQueue
	Pointer to the fallback queue (if any) More...

vector_class< add_param_func_t >	m_paramVec

unsigned	m_numKernels
	Number of kernels in the command group. More...

Detailed Description

A handler gives user access to command group scope functionality, such as API calls.

This simplifies the interface, as the command group class is not required anymore and the scope is explicit for accessors and API entries.

It is also used by accessors to get the current command group scope. Handlers can only be constructed from within queues. For the time being, the deprecated command group function also can create handlers.

The templated-side of the API entries is defined here. Some API entries are explicitly deleted to shield users from weird template errors caused by enable_if macros. In particular, if there is a pointer to a kernel instead of a kernel, the template deduction fails and causes a massive template error. However, using the deleted API entry the user sees an explicit error because they are using a non-valid interface.

Definition at line 210 of file apis.h.

Constructor & Destructor Documentation

◆ ~handler()

COMPUTECPP_TEST_VIRTUAL cl::sycl::handler::~handler ( )

Destructor of the handler, implementation on the cpp file so the default_deleter can see the implementation of the internal transaction object.

◆ handler()

cl::sycl::handler::handler	(	const dqueue_shptr &	q,
		const dqueue_shptr &	fallbackQueue = `nullptr`
	)

explicitprotected

Creates a handler for an specific queue.

Member Function Documentation

◆ COMPUTECPP_DEPRECATED_API()

template<typename elemT , int kDims, access::mode kMode, access::target kTarget>

cl::sycl::handler::COMPUTECPP_DEPRECATED_API	(	"Deprecated Codeplay extension function: " "Bind the null accessor	first,
		then call require()"
	)

Function that registers a placeholder accessor with the handler and the associated storage.

Defined in Codeplay Extension CP004. Will fail if accessor already associated with storage.

Parameters

buf	Buffer object
acc	Placeholder accessor

Deprecated:: Bind the null accessor first, then call require()

◆ copy() [1/5]

template<typename TAcc , typename THostPtr , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF( TAcc,(detail::can_copy_types< TAcc, THostPtr >::value && detail::is_read_mode< accessMode >::value)) >

void cl::sycl::handler::copy	(	accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder >	acc,
		shared_ptr_class< THostPtr >	hostPtr
	)

inline

Copies the data from the device accessor to the host pointer.

hostPtr must have enough space allocated to match the size of the accessor data.

The underlying type of the accessor and the host pointer must match

Accessor type can be const

At least one type is allowed to be void

Template Parameters

TAcc	Underlying type of the data associated with the accessor
THostPtr	Underlying type of the host pointer data
dims	Number of dimensions of the accessor
accessMode	Access mode of the accessor
accessTarget	Access target of the accessor
isPlaceholder	Whether the accessor is a placeholder
COMPUTECPP_ENABLE_IF	The function is only valid when the access mode includes read access

Parameters

acc	Accessor that is used to access the buffer or image
hostPtr	Host pointer that will be updated

Definition at line 1189 of file apis.h.

◆ copy() [2/5]

template<typename TAcc , typename THostPtr , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF( TAcc,(detail::can_copy_types< THostPtr, TAcc >::value && detail::is_write_mode< accessMode >::value)) >

void cl::sycl::handler::copy	(	shared_ptr_class< THostPtr >	hostPtr,
		accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder >	acc
	)

inline

Copies the data from the host pointer to the device accessor.

hostPtr must have enough space allocated to match the size of the accessor data.

The underlying type of the host pointer and the accessor must match

Host pointer type can be const

At least one type is allowed to be void

Template Parameters

THostPtr	Underlying type of the host pointer data
TAcc	Underlying type of the data associated with the accessor
dims	Number of dimensions of the accessor
accessMode	Access mode of the accessor
accessTarget	Access target of the accessor
isPlaceholder	Whether the accessor is a placeholder
COMPUTECPP_ENABLE_IF	The function is only valid when the access mode includes write access

Parameters

hostPtr	Host pointer that points to the new data
acc	Accessor that is used to access the buffer or image

Definition at line 1222 of file apis.h.

◆ copy() [3/5]

template<typename TAcc , typename THostPtr , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF( TAcc,(detail::can_copy_types< TAcc, THostPtr >::value && detail::is_read_mode< accessMode >::value)) >

void cl::sycl::handler::copy	(	accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder >	acc,
		THostPtr *	hostPtr
	)

inline

Copies the data from the device accessor to the host pointer.

hostPtr must have enough space allocated to match the size of the accessor data.

The underlying type of the accessor and the host pointer must match

Accessor type can be const

At least one type is allowed to be void

Template Parameters

TAcc	Underlying type of the data associated with the accessor
THostPtr	Underlying type of the host pointer data
dims	Number of dimensions of the accessor
accessMode	Access mode of the accessor
accessTarget	Access target of the accessor
isPlaceholder	Whether the accessor is a placeholder
COMPUTECPP_ENABLE_IF	The function is only valid when the access mode includes read access

Parameters

acc	Accessor that is used to access the buffer or image
hostPtr	Host pointer that will be updated

Definition at line 1257 of file apis.h.

◆ copy() [4/5]

template<typename TAcc , typename THostPtr , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF( TAcc,(detail::can_copy_types< THostPtr, TAcc >::value && detail::is_write_mode< accessMode >::value)) >

void cl::sycl::handler::copy	(	const THostPtr *	hostPtr,
		accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder >	acc
	)

inline

Copies the data from the host pointer to the device accessor.

hostPtr must have enough space allocated to match the size of the accessor data.

The underlying type of the host pointer and the accessor must match

Host pointer type can be const

At least one type is allowed to be void

Template Parameters

THostPtr	Underlying type of the host pointer data
TAcc	Underlying type of the data associated with the accessor
dims	Number of dimensions of the accessor
accessMode	Access mode of the accessor
accessTarget	Access target of the accessor
isPlaceholder	Whether the accessor is a placeholder
COMPUTECPP_ENABLE_IF	The function is only valid when the access mode includes write access

Parameters

hostPtr	Host pointer that points to the new data
acc	Accessor that is used to access the buffer or image

Definition at line 1291 of file apis.h.

◆ copy() [5/5]

template<typename T , typename U , int dimsOrig, int dimsDest, access::mode accModeOrig, access::mode accModeDest, access::target accTargetOrig, access::target accTargetDest, access::placeholder isPlaceholderOrig, access::placeholder isPlaceholderDest, COMPUTECPP_ENABLE_IF(T,((detail::can_copy_types< T, U >::value) &&(detail::is_read_mode< accModeOrig >::value) &&(detail::is_write_mode< accModeDest >::value))) >

void cl::sycl::handler::copy	(	accessor< T, dimsOrig, accModeOrig, accTargetOrig, isPlaceholderOrig >	originAcc,
		accessor< U, dimsDest, accModeDest, accTargetDest, isPlaceholderDest >	destinationAcc
	)

inline

Copies data associated with the origin accessor to the data associated with the destination accessor.

There are a few restrictions on the accessors:

The underlying type and number of dimensions must match
- Origin type can be const
The origin accessor access mode must include read access
The destination accessor access mode must include write access

The size of the destination accessor data must be enough to hold the data from the origin accessor

Template Parameters

T	Underlying type of the data associated with the origin accessor
U	Underlying type of the data associated with the destination accessors
dimsOrig	Number of dimensions of the origin accessor
dimsDest	Number of dimensions of the destination accessor
accModeOrig	Access mode of the origin accessor
accModeDest	Access mode of the destination accessor
accTargetOrig	Access target of the origin accessor
accTargetDest	Access target of the destination accessor
isPlaceholder	Whether the origin accessor is a placeholder
isPlaceholder	Whether the destination accessor is a placeholder
COMPUTECPP_ENABLE_IF	Checks that the accessor types conform to the restrictions.

Parameters

originAcc	Accessor with the data that will be copied from
destinationAcc	Accessor with the data that will be copied to

Definition at line 1337 of file apis.h.

◆ copy_in_device()

void cl::sycl::handler::copy_in_device	(	const accessor_base &	originAcc,
		const accessor_base &	destinationAcc
	)

protected

Copies data associated with the origin accessor to the data associated with the destination accessor.

Parameters

originAcc	Accessor with the data that will be copied from
destinationAcc	Accessor with the data that will be copied to

◆ create_kernel_command_group()

detail::enqueue_device_kernel_command* cl::sycl::handler::create_kernel_command_group ( kernel & syclKernel )

protected

Creates a kernel command group Internal implementation.

Returns: Pointer to the newly created command

◆ depends_on() [1/2]

void cl::sycl::handler::depends_on ( cl::sycl::event e )

inline

Register a single event that this handler should wait for before running.

Parameters

e	The event that the handler should wait for before running.

Definition at line 1133 of file apis.h.

◆ depends_on() [2/2]

void cl::sycl::handler::depends_on ( const std::vector< cl::sycl::event > & v )

inline

Register a set of events that this handler should wait for before running.

Parameters

v	a vector of events.

Definition at line 1138 of file apis.h.

◆ execute_kernel_parallel_for_id_ptr()

void cl::sycl::handler::execute_kernel_parallel_for_id_ptr	(	const detail::nd_range_base &	ndRange,
		const kernel &	syclKernel,
		const detail::parallel_for_id_ptr &	funcPtr,
		detail::enqueue_device_kernel_command *	currentCommand,
		int	dimensions
	)

protected

◆ execute_kernel_parallel_for_ptr()

void cl::sycl::handler::execute_kernel_parallel_for_ptr	(	const detail::nd_range_base &	ndRange,
		const kernel &	syclKernel,
		const detail::parallel_for_ptr &	funcPtr,
		detail::enqueue_device_kernel_command *	currentCommand,
		int	dimensions
	)

protected

◆ execute_kernel_parallel_for_work_group_ptr()

void cl::sycl::handler::execute_kernel_parallel_for_work_group_ptr	(	const detail::nd_range_base &	ndRange,
		const kernel &	syclKernel,
		const detail::parallel_for_work_group_ptr &	funcPtr,
		detail::enqueue_device_kernel_command *	currentCommand,
		int	dimensions
	)

protected

◆ execute_kernel_single_task_ptr()

void cl::sycl::handler::execute_kernel_single_task_ptr	(	const detail::nd_range_base &	ndRange,
		const kernel &	syclKernel,
		const detail::single_task_ptr &	funcPtr,
		detail::enqueue_device_kernel_command *	currentCommand
	)

protected

Creates the internal structures to execute the kernel.

This function is explicitly instantiated on the cpp file for those FunctorPtr types supported.

Parameters

nd_range_base	The given nd range
syclKernel	The SYCL device kernel
FunctorPtr	The std::function pointer to the user functor
currentCommand	The command currently being executing

◆ experimental_depends_on() [1/2]

void cl::sycl::handler::experimental_depends_on ( cl::sycl::event e )

Register a single event that this handler should wait for before running.

Parameters

e	The event that the handler should wait for before running.

◆ experimental_depends_on() [2/2]

void cl::sycl::handler::experimental_depends_on ( const std::vector< cl::sycl::event > & v )

Register a set of events that this handler should wait for before running.

Parameters

v	a vector of events.

◆ experimental_prefetch()

void cl::sycl::handler::experimental_prefetch	(	const void *	ptr,
		size_t	size
	)

Hints to the SYCL runtime that the data is available earlier than when the USM model would require it.

Can only be overlapped with kernel execution when Concurrent or System USM is available.

Parameters

ptr	Pointer to the memory to be prefetched to the device
size	Number of bytes requested to be prefetched

◆ fill() [1/4]

template<typename TAcc , typename T , int dims, cl::sycl::access::mode accessMode, cl::sycl::access::target accessTarget, access::placeholder isPlaceholder, COMPUTECPP_ENABLE_IF(TAcc,(detail::can_copy_types< T, TAcc >::value && detail::is_write_mode< accessMode >::value)) >

void cl::sycl::handler::fill	(	accessor< TAcc, dims, accessMode, accessTarget, isPlaceholder >	acc,
		T	val
	)

inline

Fills the data associated with the accessor using the scalar value.

Special case of copy from host to device where the origin is a scalar value that will be replicated across the range of the accessor.

Template Parameters

TAcc	Underlying type of the data associated with the accessor
T	Underlying type of the host scalar
dims	Number of dimensions of the accessor
accessMode	Access mode of the accessor
accessTarget	Access target of the accessor
isPlaceholder	Whether the accessor is a placeholder
COMPUTECPP_ENABLE_IF	The function is only valid when the access mode includes read access

Parameters

acc	Accessor with the data that will be filled
val	Scalar used to fill the device data with

Definition at line 1365 of file apis.h.

◆ fill() [2/4]

template<typename T >

void cl::sycl::handler::fill	(	void *	ptr,
		const T &	pattern,
		size_t	count
	)

inline

Fills the memory pointed by ptr.

Template Parameters

T	The type of the element.

Parameters

ptr	Pointer object to fill.
pattern	The pattern to fill each element of `ptr`
count	The number of elements of type T to fill.

Definition at line 1402 of file apis.h.

◆ fill() [3/4]

void cl::sycl::handler::fill	(	const accessor_base &	acc,
		const void *	patternData,
		const size_t	patternSize
	)

protected

Fills the range of the accessor with value of hostScalarPtr[0]}.

Parameters

acc	Accessor with the data that will be filled
patternData	Host data used to fill the device data with
patternSize	Size of the host data

◆ fill() [4/4]

void cl::sycl::handler::fill	(	void *	ptr,
		const void *	patternData,
		size_t	patternSize,
		size_t	size
	)

protected

Fills the memory pointed by ptr.

Parameters

ptr	Pointer object to fill.
patternData	Pointer to the memory that contains the pattern to use when filling `ptr`.
patternSize	The size in bytes of the pattern.
size	The number of bytes of `ptr` to fill with `patternData`.

◆ get_context()

context cl::sycl::handler::get_context ( ) const

protected

Returns the current context for the command group.

◆ get_device_weak()

ddevice_wkptr cl::sycl::handler::get_device_weak ( ) const

protected

Returns the current device for the command group.

◆ get_kernel_bundle_program()

program cl::sycl::handler::get_kernel_bundle_program ( ) const

protected

Get the program to use from a kernel bundle.

has_kernel_bundle() == true is a precondition.

Returns: A program.

◆ get_optimal_workgroup_size()

detail::index_array cl::sycl::handler::get_optimal_workgroup_size ( const kernel & syclKernel )

protected

Gets the optimal workgroup size for the current device and the given kernel.

Internal implementation.

Parameters

syclKernel Kernel to which we need to compute the workgroup size.

◆ has_kernel_bundle()

bool cl::sycl::handler::has_kernel_bundle ( ) const

protected

Returns true if use_kernel_bundle(bundle) has been called with the handler.

◆ interop_task_impl()

void cl::sycl::handler::interop_task_impl ( const detail::interop_task_ptr & hostTaskCallable )

protected

Schedules a host task with an interop_handle object.

Parameters

hostTaskCallable Callable object that will be executed

◆ memcpy()

void cl::sycl::handler::memcpy	(	void *	dest,
		const void *	src,
		size_t	size
	)

Copies count bytes from src to .

Parameters

dest	Pointer to the memory location to copy to.
src	Pointer to the memory location to copy from.
size	The number of bytes to copy.

◆ parallel_for() [1/10]

template<int dimensions>

void cl::sycl::handler::parallel_for	(	const nd_range< dimensions > &	ndRange,
		kernel	syclKernel
	)

inline

Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by ndRange.

Template Parameters

dimensions Number of dimensions of the kernel

Parameters

ndRange	Dimensions of the global and local work groups
syclKernel	The precompiled kernel to be enqueued

Definition at line 467 of file apis.h.

◆ parallel_for() [2/10]

template<int dimensions>

void cl::sycl::handler::parallel_for	(	const range< dimensions > &	range,
		kernel	syclKernel
	)

inline

Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of global work items specified by range.

Template Parameters

dimensions Number of dimensions of the kernel

Parameters

range	Dimensions of the global work group
syclKernel	The precompiled kernel to be enqueued

Definition at line 482 of file apis.h.

◆ parallel_for() [3/10]

template<int dimensions>

void cl::sycl::handler::parallel_for	(	const range< dimensions > &	range,
		id< dimensions >	offset,
		kernel	syclKernel
	)

inline

Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of global work items specified by range.

Template Parameters

dimensions Number of dimensions of the kernel

Parameters

range	Dimensions of the global work group
offset	The offset into the data being executed
syclKernel	The precompiled kernel to be enqueued

Definition at line 499 of file apis.h.

◆ parallel_for() [4/10]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for	(	const nd_range< dimensions > &	ndRange,
		const functorT &	functor
	)

inline

Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of local and global work items specified by ndRange.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

ndRange	Dimensions of the global and local work groups
functor	The kernel being enqueued

Definition at line 579 of file apis.h.

◆ parallel_for() [5/10]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for	(	kernel	syclKernel,
		const nd_range< dimensions > &	ndRange,
		const functorT &	functor
	)

inline

Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of local and global work items specified by ndRange.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

syclKernel	The precompiled kernel to be enqueued
ndRange	Dimensions of the global and local work groups
functor	The kernel being enqueued

Definition at line 601 of file apis.h.

◆ parallel_for() [6/10]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for	(	const range< dimensions > &	range,
		const functorT &	functor
	)

inline

Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of global work items specified by range.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

range	Dimensions of the global work group
functor	The kernel being enqueued

Definition at line 755 of file apis.h.

◆ parallel_for() [7/10]

template<typename nameT = std::nullptr_t, typename functorT >

void cl::sycl::handler::parallel_for	(	const size_t	range,
		const functorT &	functor
	)

inline

Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of global work items specified by range.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler

Parameters

range	Size of the global work group
functor	The kernel being enqueued

Definition at line 773 of file apis.h.

◆ parallel_for() [8/10]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for	(	const range< dimensions > &	range,
		const id< dimensions > &	offset,
		const functorT &	functor
	)

inline

Parallel_for will enqueue the kernel functor to be executed a number of instances working in parallel over the number of global work items specified by range.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

range	Dimensions of the global work group
offset	The offset into the data being executed
functor	The kernel being enqueued

Definition at line 794 of file apis.h.

◆ parallel_for() [9/10]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for	(	kernel	syclKernel,
		const range< dimensions > &	range,
		const functorT &	functor
	)

inline

Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of global work items specified by range.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

syclKernel	The precompiled kernel which is being run
range	Dimensions of the global work group
functor	The kernel being enqueued

Definition at line 815 of file apis.h.

◆ parallel_for() [10/10]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for	(	kernel	syclKernel,
		const range< dimensions > &	range,
		const id< dimensions > &	offset,
		const functorT &	functor
	)

inline

Parallel_for will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of global work items specified by range.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

syclKernel	The precompiled kernel which is being run
range	Dimensions of the global work group
offset	The offset into the data being executed
functor	The kernel being enqueued

Definition at line 838 of file apis.h.

◆ parallel_for_work_group() [1/5]

template<int dimensions>

void cl::sycl::handler::parallel_for_work_group	(	const range< dimensions > &	numGroups,
		kernel	syclKernel
	)

inline

parallel_for_work_group will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by numGroups.

Template Parameters

dimensions Number of dimensions of the kernel

Parameters

numGroups	Dimensions of the global and local work groups
syclKernel	The precompiled kernel which is being run

Definition at line 857 of file apis.h.

◆ parallel_for_work_group() [2/5]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for_work_group	(	kernel	syclKernel,
		const range< dimensions > &	range,
		const functorT &	functor
	)

inline

parallel_for_work_group will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by range.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

syclKernel	The precompiled kernel which is being run
range	Dimensions of the global work groups
functor	The kernel being enqueued

Definition at line 975 of file apis.h.

◆ parallel_for_work_group() [3/5]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for_work_group	(	const range< dimensions > &	range,
		const functorT &	functor
	)

inline

parallel_for_work_group will enqueue the kernel functor to be executed a number of instances working in parallel over the number of local and global work items specified by range.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

range	Dimensions of the global and local work groups
functor	The kernel being enqueued

Definition at line 996 of file apis.h.

◆ parallel_for_work_group() [4/5]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for_work_group	(	const range< dimensions > &	numGroups,
		const range< dimensions > &	groupSize,
		const functorT &	functor
	)

inline

parallel_for_work_group will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by numGroups and groupSize.

Template Parameters

functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

syclKernel	The precompiled kernel which is being run
numGroups	dimensions of the work groups being launched
groupSize	each work group will launch work-items of dimension of groupSize

Template Parameters

nameT The name of the kernel being enqueued

Parameters

functor The kernel being enqueued

Definition at line 1019 of file apis.h.

◆ parallel_for_work_group() [5/5]

template<typename nameT = std::nullptr_t, typename functorT , int dimensions>

void cl::sycl::handler::parallel_for_work_group	(	kernel	syclKernel,
		const range< dimensions > &	numGroups,
		const range< dimensions > &	groupSize,
		const functorT &	functor
	)

inline

parallel_for_work_group will enqueue the precompiled kernel syclKernel to be executed a number of instances working in parallel over the number of local and global work items specified by numGroups and groupSize.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler
dimensions	Number of dimensions of the kernel

Parameters

syclKernel	The precompiled kernel which is being run
numGroups	dimensions of the work groups being launched
groupSize	each work group will launch work-items of dimension of groupSize
functor	The kernel being enqueued

Definition at line 1044 of file apis.h.

◆ prefetch()

void cl::sycl::handler::prefetch	(	const void *	ptr,
		size_t	size
	)

inline

Hints to the SYCL runtime that the data is available earlier than when the USM model would require it.

Can only be overlapped with kernel execution when Concurrent or System USM is available.

Parameters

ptr	Pointer to the memory to be prefetched to the device
size	Number of bytes requested to be prefetched

Definition at line 1444 of file apis.h.

◆ process_functor_arguments_impl()

void cl::sycl::handler::process_functor_arguments_impl	(	kernel	syclKernel,
		detail::binary_address	functorBuffer,
		const detail::functor_arg_descriptor &	argDesc,
		detail::enqueue_device_kernel_command *	currentCommand
	)

protected

Gets the parameters from a functor and sets them as OpenCL arguments.

Internal implementation.

Parameters

syclKernel	Kernel to which the functor is associated
functorBuffer	Functor buffer casted down to a binary array
SI	Kernel struct iterator
SE	Kernel struct iterator
OI	Kernel struct iterator

◆ register_for_dma() [1/3]

template<typename elemT , int kDims, access::mode kMode, access::placeholder isPlaceholder>

void cl::sycl::handler::register_for_dma	(	accessor< elemT, kDims, kMode, COMPUTECPP_ACCESS_TARGET_DEVICE, isPlaceholder > &	acc,
		size_t	stride
	)

inline

Registers a global memory accessor for DMA transfer.

Parameters

acc	Accessor to use in a DMA transfer
stride	DMA transfer stride, in number of elements

Definition at line 1078 of file apis.h.

◆ register_for_dma() [2/3]

template<typename elemT , int kDims, access::placeholder isPlaceholder>

void cl::sycl::handler::register_for_dma	(	accessor< elemT, kDims, access::mode::read, COMPUTECPP_ACCESS_TARGET_DEVICE, isPlaceholder > &	acc,
		size_t	stride
	)

inline

Registers a constant memory accessor for DMA transfer.

Parameters

acc	Accessor to use in a DMA transfer
stride	DMA transfer stride, in number of elements

Definition at line 1091 of file apis.h.

◆ register_for_dma() [3/3]

void cl::sycl::handler::register_for_dma	(	accessor_base &	acc,
		size_t	strideBytes
	)

protected

Registers an accessor for DMA transfer.

Parameters

acc	Accessor to use in a DMA transfer
stride	DMA transfer stride, in bytes

◆ require() [1/3]

template<typename elemT , int kDims, access::mode kMode, access::target kTarget, access::placeholder IsPlaceholder = access::placeholder::true_t>

void cl::sycl::handler::require ( const accessor< elemT, kDims, kMode, kTarget, IsPlaceholder > & acc )

inline

Function that registers a placeholder accessor with the handler.

Defined in Codeplay Extension CP004.

Parameters

acc	Placeholder accessor

Definition at line 1067 of file apis.h.

◆ require() [2/3]

void cl::sycl::handler::require ( const accessor_base & acc )

protected

Internal function that registers a placeholder accessor with the handler.

Parameters

acc	Placeholder accessor

◆ require() [3/3]

void cl::sycl::handler::require	(	const accessor_base &	acc,
		dmem_shptr	memObj,
		access::mode	mode,
		access::target	target
	)

protected

Internal function that registers a placeholder accessor with the handler.

Parameters

acc	Placeholder accessor

Deprecated:: Bind the null accessor first, then call require()

◆ set_arg()

template<typename T >

void cl::sycl::handler::set_arg	(	int	param_num,
		T &&	param
	)

inline

Sets an argument when using interop kernels.

Definition at line 356 of file apis.h.

◆ set_args()

template<typename... Ts>

void cl::sycl::handler::set_args ( Ts &&... args )

inline

Set all the given kernel args arguments for an OpenCL kernel, as if set_arg() was used with each of them in the same order and increasing index always starting at 0.

Template Parameters

Ts	Types of the parameters passed to the OpenCL kernel

Parameters

args	Parameters passed to the OpenCL kernel

Definition at line 375 of file apis.h.

◆ single_task() [1/3]

void cl::sycl::handler::single_task ( kernel syclKernel )

This function effectively just launches a single thread to execute the kernel in serial asynchronously to the host execution.

This function takes in a precompiled kernel syclKernel previously created using build_with_kernel_type or compile_with_kernel_type

Parameters

syclKernel The precompiled kernel to be enqueued

◆ single_task() [2/3]

template<typename nameT = std::nullptr_t, typename functorT >

void cl::sycl::handler::single_task ( const functorT & functor )

inline

This function effectively just launches a single thread to execute the kernel in serial asynchronously to the host execution.

Template Parameters

nameT	The name of the kernel being enqueued
functorT	This is the type of the kernel. It will be automatically deduced by the compiler

Parameters

functor The kernel being enqueued

Definition at line 430 of file apis.h.

◆ single_task() [3/3]

template<typename nameT = std::nullptr_t, typename functorT >

void cl::sycl::handler::single_task	(	kernel	syclKernel,
		const functorT &	functor
	)

inline

This function effectively just launches a single thread to execute the kernel in serial asynchronously to the host execution.

This function takes in a precompiled kernel syclKernel previously created using build_with_kernel_type or compile_with_kernel_type

Template Parameters

nameT The name of the kernel being enqueued

Parameters

syclKernel The precompiled kernel to be enqueued

Template Parameters

functorT This is the type of the kernel. It will be automatically deduced by the compiler

Parameters

functor The kernel being enqueued

Definition at line 448 of file apis.h.

◆ update_device_data()

void cl::sycl::handler::update_device_data	(	const accessor_base &	acc,
		shared_ptr_class< void >	hostPtr,
		cl::sycl::access::mode	accessMode,
		bool	userProvidedPtr
	)

protected

Updates device data by copying to/from the device.

Parameters

acc	Accessor that is used to access the buffer or image
hostPtr	Pointer that points to data on the host
accessMode	Operation indicator (read -> copy_from_device, write -> copy_to_device)
userProvidedPtr	Indicated whether the host pointer was provided by the user or whether to use the internal one

◆ update_host()

template<typename T , int dims, access::mode accessMode, access::target accessTarget, access::placeholder isPlaceholder>

void cl::sycl::handler::update_host ( accessor< T, dims, accessMode, accessTarget, isPlaceholder > acc )

inline

Update the memory object accessed by a given accessor on the host.

The host copy of the memory object will be up to date once this has executed.

Template Parameters

T	The accessor element type.
dims	The accessor's dimensionality
accessMode	The access mode of the accessor
accessTarget	The access target.
isPlaceholder	Is the accessor a placeholder?

Parameters

acc	An accessor for the memory object to update.

Note: This does not yet automatically update the host pointer given to the initial buffer. For this to be updated, construct the buffer using the property property::buffer::use_host_ptr.

Definition at line 1389 of file apis.h.

◆ update_host_impl()

void cl::sycl::handler::update_host_impl ( const accessor_base & acc )

protected

Implementation for update_host.

Parameters

acc	An accessor for the memory object to update.

Note: This does not yet automatically update the host pointer given to the initial buffer. For this to be updated, construct the buffer using the property property::buffer::use_host_ptr.

◆ use_kernel_bundle_impl()

void cl::sycl::handler::use_kernel_bundle_impl ( const dkernelbundle_shptr execBundle )

protected

This command group will use device images from the given kernel bundle when invoking kernels.

Parameters

execBundle The bundle to use kernels from.

Member Data Documentation

◆ acc

kDims const accessor<elemT, kDims, kMode, kTarget, access::placeholder::true_t>& cl::sycl::handler::acc

Initial value:

{
    this->require(static_cast<const accessor_base&>(acc), bufObj.get_impl(),
 kMode, kTarget)

Definition at line 1114 of file apis.h.

◆ bufObj

kDims& cl::sycl::handler::bufObj

Definition at line 1112 of file apis.h.

◆ m_fallbackQueue

dqueue_shptr cl::sycl::handler::m_fallbackQueue

protected

Pointer to the fallback queue (if any)

Definition at line 1661 of file apis.h.

◆ m_numKernels

unsigned cl::sycl::handler::m_numKernels

protected

Number of kernels in the command group.

Definition at line 1671 of file apis.h.

◆ m_paramVec

vector_class<add_param_func_t> cl::sycl::handler::m_paramVec

protected

Definition at line 1667 of file apis.h.

◆ m_queue

dqueue_shptr cl::sycl::handler::m_queue

protected

Queue to which this handler was submitted to.

Definition at line 1657 of file apis.h.

◆ m_trans

dtrans_uptr cl::sycl::handler::m_trans

protected

Internal transaction associated with the handler.

Definition at line 1653 of file apis.h.

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

apis.h

menu_bookReference

Public Member Functions

Public Attributes

Protected Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

◆ ~handler()

◆ handler()

Member Function Documentation

◆ COMPUTECPP_DEPRECATED_API()

◆ copy() [1/5]

◆ copy() [2/5]

◆ copy() [3/5]

◆ copy() [4/5]

◆ copy() [5/5]

◆ copy_in_device()

◆ create_kernel_command_group()

◆ depends_on() [1/2]

◆ depends_on() [2/2]

◆ execute_kernel_parallel_for_id_ptr()

◆ execute_kernel_parallel_for_ptr()

◆ execute_kernel_parallel_for_work_group_ptr()

◆ execute_kernel_single_task_ptr()

◆ experimental_depends_on() [1/2]

◆ experimental_depends_on() [2/2]

◆ experimental_prefetch()

◆ fill() [1/4]

◆ fill() [2/4]

◆ fill() [3/4]

◆ fill() [4/4]

◆ get_context()

◆ get_device_weak()

◆ get_kernel_bundle_program()

◆ get_optimal_workgroup_size()

◆ has_kernel_bundle()

◆ interop_task_impl()

◆ memcpy()

◆ parallel_for() [1/10]

◆ parallel_for() [2/10]

◆ parallel_for() [3/10]

◆ parallel_for() [4/10]

◆ parallel_for() [5/10]

◆ parallel_for() [6/10]

◆ parallel_for() [7/10]

◆ parallel_for() [8/10]

◆ parallel_for() [9/10]

◆ parallel_for() [10/10]

◆ parallel_for_work_group() [1/5]

◆ parallel_for_work_group() [2/5]

◆ parallel_for_work_group() [3/5]

◆ parallel_for_work_group() [4/5]

◆ parallel_for_work_group() [5/5]

◆ prefetch()

◆ process_functor_arguments_impl()

◆ register_for_dma() [1/3]

◆ register_for_dma() [2/3]

◆ register_for_dma() [3/3]

◆ require() [1/3]

◆ require() [2/3]

◆ require() [3/3]

◆ set_arg()

◆ set_args()

◆ single_task() [1/3]

◆ single_task() [2/3]

◆ single_task() [3/3]

◆ update_device_data()

◆ update_host()

◆ update_host_impl()

◆ use_kernel_bundle_impl()

Member Data Documentation

◆ acc

◆ bufObj

◆ m_fallbackQueue

◆ m_numKernels

◆ m_paramVec

◆ m_queue

◆ m_trans

Select a Product

oneAPI