Command group handler class. More...
#include <sycl/handler.hpp>
Classes | |
struct | ShouldEnableSetArg |
Public Types | |
template<typename T > | |
using | remove_cv_ref_t = typename detail::remove_cv_t< detail::remove_reference_t< T > > |
template<typename U , typename T > | |
using | is_same_type = std::is_same< remove_cv_ref_t< U >, remove_cv_ref_t< T > > |
Public Member Functions | |
handler (const handler &)=delete | |
handler (handler &&)=delete | |
handler & | operator= (const handler &)=delete |
handler & | operator= (handler &&)=delete |
template<auto & SpecName> | |
void | set_specialization_constant (typename std::remove_reference_t< decltype(SpecName)>::value_type Value) |
template<auto & SpecName> | |
std::remove_reference_t< decltype(SpecName)>::value_type | get_specialization_constant () const |
void | use_kernel_bundle (const kernel_bundle< bundle_state::executable > &ExecBundle) |
template<typename DataT , int Dims, access::mode AccMode, access::target AccTarget, access::placeholder isPlaceholder> | |
void | require (accessor< DataT, Dims, AccMode, AccTarget, isPlaceholder > Acc) |
Requires access to the memory object associated with the placeholder accessor. More... | |
void | depends_on (event Event) |
Registers event dependencies on this command group. More... | |
void | depends_on (const std::vector< event > &Events) |
Registers event dependencies on this command group. More... | |
template<typename T > | |
detail::enable_if_t< ShouldEnableSetArg< T >::value, void > | set_arg (int ArgIndex, T &&Arg) |
Sets argument for OpenCL interoperability kernels. More... | |
template<typename DataT , int Dims, access::mode AccessMode, access::target AccessTarget, access::placeholder IsPlaceholder> | |
void | set_arg (int ArgIndex, accessor< DataT, Dims, AccessMode, AccessTarget, IsPlaceholder > Arg) |
template<typename DataT , int Dims> | |
void | set_arg (int ArgIndex, local_accessor< DataT, Dims > Arg) |
template<typename... Ts> | |
void | set_args (Ts &&...Args) |
Sets arguments for OpenCL interoperability kernels. More... | |
template<typename KernelName = detail::auto_name, typename KernelType > | |
void | single_task (_KERNELFUNCPARAM(KernelFunc)) |
Defines and invokes a SYCL kernel function as a function object type. More... | |
template<typename KernelName = detail::auto_name, typename KernelType > | |
void | parallel_for (range< 1 > NumWorkItems, _KERNELFUNCPARAM(KernelFunc)) |
template<typename KernelName = detail::auto_name, typename KernelType > | |
void | parallel_for (range< 2 > NumWorkItems, _KERNELFUNCPARAM(KernelFunc)) |
template<typename KernelName = detail::auto_name, typename KernelType > | |
void | parallel_for (range< 3 > NumWorkItems, _KERNELFUNCPARAM(KernelFunc)) |
template<typename FuncT > | |
__SYCL_DEPRECATED ("run_on_host_intel() is deprecated, use host_task() instead") void run_on_host_intel(FuncT Func) | |
Defines and invokes a SYCL kernel on host device. More... | |
template<typename FuncT > | |
detail::enable_if_t< detail::check_fn_signature< detail::remove_reference_t< FuncT >, void()>::value||detail::check_fn_signature< detail::remove_reference_t< FuncT >, void(interop_handle)>::value > | host_task (FuncT &&Func) |
Enqueues a command to the SYCL runtime to invoke Func once. More... | |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims> | |
void | parallel_for (range< Dims > NumWorkItems, id< Dims > WorkItemOffset, _KERNELFUNCPARAM(KernelFunc)) |
Defines and invokes a SYCL kernel function for the specified range and offset. More... | |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims> | |
void | parallel_for_work_group (range< Dims > NumWorkGroups, _KERNELFUNCPARAM(KernelFunc)) |
Hierarchical kernel invocation method of a kernel defined as a lambda encoding the body of each work-group to launch. More... | |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims> | |
void | parallel_for_work_group (range< Dims > NumWorkGroups, range< Dims > WorkGroupSize, _KERNELFUNCPARAM(KernelFunc)) |
Hierarchical kernel invocation method of a kernel defined as a lambda encoding the body of each work-group to launch. More... | |
void | single_task (kernel Kernel) |
Invokes a SYCL kernel. More... | |
void | parallel_for (range< 1 > NumWorkItems, kernel Kernel) |
void | parallel_for (range< 2 > NumWorkItems, kernel Kernel) |
void | parallel_for (range< 3 > NumWorkItems, kernel Kernel) |
template<int Dims> | |
void | parallel_for (range< Dims > NumWorkItems, id< Dims > WorkItemOffset, kernel Kernel) |
Defines and invokes a SYCL kernel function for the specified range and offsets. More... | |
template<int Dims> | |
void | parallel_for (nd_range< Dims > NDRange, kernel Kernel) |
Defines and invokes a SYCL kernel function for the specified range and offsets. More... | |
template<typename KernelName = detail::auto_name, typename KernelType > | |
void | single_task (kernel Kernel, _KERNELFUNCPARAM(KernelFunc)) |
Defines and invokes a SYCL kernel function. More... | |
template<typename FuncT > | |
__SYCL_DEPRECATED ("interop_task() is deprecated, use host_task() instead") void interop_task(FuncT Func) | |
Invokes a lambda on the host. More... | |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims> | |
void | parallel_for (kernel Kernel, range< Dims > NumWorkItems, _KERNELFUNCPARAM(KernelFunc)) |
Defines and invokes a SYCL kernel function for the specified range. More... | |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims> | |
void | parallel_for (kernel Kernel, range< Dims > NumWorkItems, id< Dims > WorkItemOffset, _KERNELFUNCPARAM(KernelFunc)) |
Defines and invokes a SYCL kernel function for the specified range and offsets. More... | |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims> | |
void | parallel_for (kernel Kernel, nd_range< Dims > NDRange, _KERNELFUNCPARAM(KernelFunc)) |
Defines and invokes a SYCL kernel function for the specified range and offsets. More... | |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims> | |
void | parallel_for_work_group (kernel Kernel, range< Dims > NumWorkGroups, _KERNELFUNCPARAM(KernelFunc)) |
Hierarchical kernel invocation method of a kernel. More... | |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims> | |
void | parallel_for_work_group (kernel Kernel, range< Dims > NumWorkGroups, range< Dims > WorkGroupSize, _KERNELFUNCPARAM(KernelFunc)) |
Hierarchical kernel invocation method of a kernel. More... | |
template<typename KernelName = detail::auto_name, typename KernelType , typename PropertiesT > | |
std::enable_if_t< ext::oneapi::experimental::is_property_list< PropertiesT >::value > | single_task (PropertiesT, _KERNELFUNCPARAM(KernelFunc)) |
template<typename KernelName = detail::auto_name, typename KernelType , typename PropertiesT > | |
std::enable_if_t< ext::oneapi::experimental::is_property_list< PropertiesT >::value > | parallel_for (range< 1 > NumWorkItems, PropertiesT, _KERNELFUNCPARAM(KernelFunc)) |
template<typename KernelName = detail::auto_name, typename KernelType , typename PropertiesT > | |
std::enable_if_t< ext::oneapi::experimental::is_property_list< PropertiesT >::value > | parallel_for (range< 2 > NumWorkItems, PropertiesT, _KERNELFUNCPARAM(KernelFunc)) |
template<typename KernelName = detail::auto_name, typename KernelType , typename PropertiesT > | |
std::enable_if_t< ext::oneapi::experimental::is_property_list< PropertiesT >::value > | parallel_for (range< 3 > NumWorkItems, PropertiesT, _KERNELFUNCPARAM(KernelFunc)) |
template<typename KernelName = detail::auto_name, typename KernelType , typename PropertiesT , int Dims> | |
std::enable_if_t< ext::oneapi::experimental::is_property_list< PropertiesT >::value > | parallel_for (nd_range< Dims > Range, PropertiesT Properties, _KERNELFUNCPARAM(KernelFunc)) |
class | detail::queue_impl |
template<typename DataT , int Dims, access::mode AccMode, access::target AccTarget, access::placeholder isPlaceholder, typename PropertyListT > | |
class | accessor |
template<typename DataT , int Dimensions, access::mode AccessMode, access::target AccessTarget, access::placeholder IsPlaceholder> | |
class | detail::image_accessor |
class | stream |
class | detail::stream_impl |
template<typename T , class BinaryOperation , int Dims, size_t Extent, typename RedOutVar > | |
class | detail::reduction_impl_algo |
class | ::MockHandler |
device | detail::getDeviceFromHandler (handler &) |
void | detail::reduction::finalizeHandler (handler &CGH) |
template<class FunctorTy > | |
void | detail::reduction::withAuxHandler (handler &CGH, FunctorTy Func) |
template<typename KernelName , detail::reduction::strategy Strategy, int Dims, typename PropertiesT , typename... RestT> | |
void | detail::reduction_parallel_for (handler &CGH, range< Dims > NDRange, PropertiesT Properties, RestT... Rest) |
template<typename KernelName , detail::reduction::strategy Strategy, int Dims, typename PropertiesT , typename... RestT> | |
void | detail::reduction_parallel_for (handler &CGH, nd_range< Dims > NDRange, PropertiesT Properties, RestT... Rest) |
void | detail::associateWithHandler (handler &, detail::AccessorBaseHost *, access::target) |
template<typename KernelName = detail::auto_name, int Dims, typename PropertiesT , typename... RestT> | |
std::enable_if_t<(sizeof...(RestT) > 1) &&detail::AreAllButLastReductions< RestT... >::value &&ext::oneapi::experimental::is_property_list< PropertiesT >::value > | parallel_for (range< Dims > Range, PropertiesT Properties, RestT &&...Rest) |
Reductions. More... | |
template<typename KernelName = detail::auto_name, int Dims, typename... RestT> | |
std::enable_if_t< detail::AreAllButLastReductions< RestT... >::value > | parallel_for (range< Dims > Range, RestT &&...Rest) |
template<typename KernelName = detail::auto_name, int Dims, typename PropertiesT , typename... RestT> | |
std::enable_if_t<(sizeof...(RestT) > 1) &&detail::AreAllButLastReductions< RestT... >::value &&ext::oneapi::experimental::is_property_list< PropertiesT >::value > | parallel_for (nd_range< Dims > Range, PropertiesT Properties, RestT &&...Rest) |
template<typename KernelName = detail::auto_name, int Dims, typename... RestT> | |
std::enable_if_t< detail::AreAllButLastReductions< RestT... >::value > | parallel_for (nd_range< Dims > Range, RestT &&...Rest) |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims, typename PropertiesT > | |
void | parallel_for_work_group (range< Dims > NumWorkGroups, PropertiesT, _KERNELFUNCPARAM(KernelFunc)) |
}@ More... | |
template<typename KernelName = detail::auto_name, typename KernelType , int Dims, typename PropertiesT > | |
void | parallel_for_work_group (range< Dims > NumWorkGroups, range< Dims > WorkGroupSize, PropertiesT, _KERNELFUNCPARAM(KernelFunc)) |
template<typename T_Src , typename T_Dst , int Dims, access::mode AccessMode, access::target AccessTarget, access::placeholder IsPlaceholder = access::placeholder::false_t> | |
void | copy (accessor< T_Src, Dims, AccessMode, AccessTarget, IsPlaceholder > Src, std::shared_ptr< T_Dst > Dst) |
Copies the content of memory object accessed by Src into the memory pointed by Dst. More... | |
template<typename T_Src , typename T_Dst , int Dims, access::mode AccessMode, access::target AccessTarget, access::placeholder IsPlaceholder = access::placeholder::false_t> | |
void | copy (std::shared_ptr< T_Src > Src, accessor< T_Dst, Dims, AccessMode, AccessTarget, IsPlaceholder > Dst) |
Copies the content of memory pointed by Src into the memory object accessed by Dst. More... | |
template<typename T_Src , typename T_Dst , int Dims, access::mode AccessMode, access::target AccessTarget, access::placeholder IsPlaceholder = access::placeholder::false_t> | |
void | copy (accessor< T_Src, Dims, AccessMode, AccessTarget, IsPlaceholder > Src, T_Dst *Dst) |
Copies the content of memory object accessed by Src into the memory pointed by Dst. More... | |
template<typename T_Src , typename T_Dst , int Dims, access::mode AccessMode, access::target AccessTarget, access::placeholder IsPlaceholder = access::placeholder::false_t> | |
void | copy (const T_Src *Src, accessor< T_Dst, Dims, AccessMode, AccessTarget, IsPlaceholder > Dst) |
Copies the content of memory pointed by Src into the memory object accessed by Dst. More... | |
template<typename T_Src , int Dims_Src, access::mode AccessMode_Src, access::target AccessTarget_Src, typename T_Dst , int Dims_Dst, access::mode AccessMode_Dst, access::target AccessTarget_Dst, access::placeholder IsPlaceholder_Src = access::placeholder::false_t, access::placeholder IsPlaceholder_Dst = access::placeholder::false_t> | |
void | copy (accessor< T_Src, Dims_Src, AccessMode_Src, AccessTarget_Src, IsPlaceholder_Src > Src, accessor< T_Dst, Dims_Dst, AccessMode_Dst, AccessTarget_Dst, IsPlaceholder_Dst > Dst) |
Copies the content of memory object accessed by Src to the memory object accessed by Dst. More... | |
template<typename T , int Dims, access::mode AccessMode, access::target AccessTarget, access::placeholder IsPlaceholder = access::placeholder::false_t> | |
void | update_host (accessor< T, Dims, AccessMode, AccessTarget, IsPlaceholder > Acc) |
Provides guarantees that the memory object accessed via Acc is updated on the host after command group object execution is complete. More... | |
template<typename T , int Dims, access::mode AccessMode, access::target AccessTarget, access::placeholder IsPlaceholder = access::placeholder::false_t, typename PropertyListT = property_list> | |
void | fill (accessor< T, Dims, AccessMode, AccessTarget, IsPlaceholder, PropertyListT > Dst, const T &Pattern) |
Fills memory pointed by accessor with the pattern given. More... | |
template<typename T > | |
void | fill (void *Ptr, const T &Pattern, size_t Count) |
Fills the specified memory with the specified pattern. More... | |
void | ext_oneapi_barrier () |
Prevents any commands submitted afterward to this queue from executing until all commands previously submitted to this queue have entered the complete state. More... | |
void | barrier () |
Prevents any commands submitted afterward to this queue from executing until all commands previously submitted to this queue have entered the complete state. More... | |
void | ext_oneapi_barrier (const std::vector< event > &WaitList) |
Prevents any commands submitted afterward to this queue from executing until all events in WaitList have entered the complete state. More... | |
void | barrier (const std::vector< event > &WaitList) |
Prevents any commands submitted afterward to this queue from executing until all events in WaitList have entered the complete state. More... | |
void | memcpy (void *Dest, const void *Src, size_t Count) |
Copies data from one memory region to another, each is either a host pointer or a pointer within USM allocation accessible on this handler's device. More... | |
template<typename T > | |
void | copy (const T *Src, T *Dest, size_t Count) |
Copies data from one memory region to another, each is either a host pointer or a pointer within USM allocation accessible on this handler's device. More... | |
void | memset (void *Dest, int Value, size_t Count) |
Fills the memory pointed by a USM pointer with the value specified. More... | |
void | prefetch (const void *Ptr, size_t Count) |
Provides hints to the runtime library that data should be made available on a device earlier than Unified Shared Memory would normally require it to be available. More... | |
void | mem_advise (const void *Ptr, size_t Length, int Advice) |
Provides additional information to the underlying runtime about how different allocations are used. More... | |
template<typename T = unsigned char, typename = std::enable_if_t<std::is_same_v<T, unsigned char>>> | |
void | ext_oneapi_memcpy2d (void *Dest, size_t DestPitch, const void *Src, size_t SrcPitch, size_t Width, size_t Height) |
Copies data from one 2D memory region to another, both pointed by USM pointers. More... | |
template<typename T > | |
void | ext_oneapi_copy2d (const T *Src, size_t SrcPitch, T *Dest, size_t DestPitch, size_t Width, size_t Height) |
Copies data from one 2D memory region to another, both pointed by USM pointers. More... | |
template<typename T = unsigned char, typename = std::enable_if_t<std::is_same_v<T, unsigned char>>> | |
void | ext_oneapi_memset2d (void *Dest, size_t DestPitch, int Value, size_t Width, size_t Height) |
Fills the memory pointed by a USM pointer with the value specified. More... | |
template<typename T > | |
void | ext_oneapi_fill2d (void *Dest, size_t DestPitch, const T &Pattern, size_t Width, size_t Height) |
Fills the memory pointed by a USM pointer with the value specified. More... | |
template<typename T , typename PropertyListT > | |
void | memcpy (ext::oneapi::experimental::device_global< T, PropertyListT > &Dest, const void *Src, size_t NumBytes=sizeof(T), size_t DestOffset=0) |
Copies data from a USM memory region to a device_global. More... | |
template<typename T , typename PropertyListT > | |
void | memcpy (void *Dest, const ext::oneapi::experimental::device_global< T, PropertyListT > &Src, size_t NumBytes=sizeof(T), size_t SrcOffset=0) |
Copies data from a device_global to USM memory. More... | |
template<typename T , typename PropertyListT > | |
void | copy (const std::remove_all_extents_t< T > *Src, ext::oneapi::experimental::device_global< T, PropertyListT > &Dest, size_t Count=sizeof(T)/sizeof(std::remove_all_extents_t< T >), size_t StartIndex=0) |
Copies elements of type std::remove_all_extents_t<T> from a USM memory region to a device_global. More... | |
template<typename T , typename PropertyListT > | |
void | copy (const ext::oneapi::experimental::device_global< T, PropertyListT > &Src, std::remove_all_extents_t< T > *Dest, size_t Count=sizeof(T)/sizeof(std::remove_all_extents_t< T >), size_t StartIndex=0) |
Copies elements of type std::remove_all_extents_t<T> from a device_global to a USM memory region. More... | |
Command group handler class.
Objects of the handler class collect information about command group, such as kernel, requirements to the memory, arguments for the kernel.
The command group can represent absolutely different operations. Depending on the operation we need to store different data. But, in most cases, it's impossible to say what kind of operation we need to perform until the very end. So, handler class contains all fields simultaneously, then during "finalization" it constructs CG object, that represents specific operation, passing fields that are required only.
Definition at line 313 of file handler.hpp.
using sycl::_V1::handler::is_same_type = std::is_same<remove_cv_ref_t<U>, remove_cv_ref_t<T> > |
Definition at line 1502 of file handler.hpp.
using sycl::_V1::handler::remove_cv_ref_t = typename detail::remove_cv_t<detail::remove_reference_t<T> > |
Definition at line 1499 of file handler.hpp.
|
delete |
|
delete |
|
inline |
Invokes a lambda on the host.
Dependencies are satisfied on the host.
Func | is a lambda that is executed on the host |
Definition at line 1784 of file handler.hpp.
|
inline |
Defines and invokes a SYCL kernel on host device.
Func | is a SYCL kernel function defined by lambda function or a named function object type. |
Definition at line 1582 of file handler.hpp.
|
inline |
Prevents any commands submitted afterward to this queue from executing until all commands previously submitted to this queue have entered the complete state.
Definition at line 2371 of file handler.hpp.
void sycl::_V1::handler::barrier | ( | const std::vector< event > & | WaitList | ) |
Prevents any commands submitted afterward to this queue from executing until all events in WaitList have entered the complete state.
If WaitList is empty, then the barrier has no effect.
WaitList | is a vector of valid SYCL events that need to complete before barrier command can be executed. |
Definition at line 673 of file handler.cpp.
References ext_oneapi_barrier().
|
inline |
Copies the content of memory object accessed by Src into the memory pointed by Dst.
Source must have at least as many bytes as the range accessed by Dst.
Src | is a source SYCL accessor. |
Dst | is a smart pointer to destination memory. |
Definition at line 2107 of file handler.hpp.
References sycl::_V1::AccessMode.
Referenced by sycl::_V1::queue::copy(), and sycl::_V1::detail::reduction_impl_algo< T, BinaryOperation, Dims, Extent, RedOutVar >::withInitializedMem().
|
inline |
Copies the content of memory object accessed by Src into the memory pointed by Dst.
Source must have at least as many bytes as the range accessed by Dst.
Src | is a source SYCL accessor. |
Dst | is a pointer to destination memory. |
Definition at line 2156 of file handler.hpp.
References sycl::_V1::AccessMode, and sycl::_V1::detail::getSyclObjImpl().
|
inline |
Copies the content of memory object accessed by Src to the memory object accessed by Dst.
Dst must have at least as many bytes as the range accessed by Src.
Src | is a source SYCL accessor. |
Dst | is a destination SYCL accessor. |
Definition at line 2236 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
|
inline |
Copies elements of type std::remove_all_extents_t<T>
from a device_global to a USM memory region.
Throws an exception if the copy operation intends to write outside the memory range
Src,as | specified through |
Count | and |
StartIndex. | |
Src | is the source device_global. |
Dest | is a USM pointer to copy to. |
Count | is a number of elements to copy. |
StartIndex | is the index of the first element in Src to copy from. |
Definition at line 2657 of file handler.hpp.
References sycl::_V1::detail::memcpy().
|
inline |
Copies elements of type std::remove_all_extents_t<T>
from a USM memory region to a device_global.
Throws an exception if the copy operation intends to write outside the memory range
Dest,as | specified through |
Count | and |
StartIndex. | |
Src | is a USM pointer to the source memory. |
Dest | is the destination device_glboal. |
Count | is a number of elements to copy. |
StartIndex | is the index of the first element in Dest to copy to. |
Definition at line 2637 of file handler.hpp.
References sycl::_V1::detail::memcpy().
|
inline |
Copies data from one memory region to another, each is either a host pointer or a pointer within USM allocation accessible on this handler's device.
No operations is done if
Count | is zero. An exception is thrown if either |
Dest | or |
Src | is nullptr. The behavior is undefined if any of the pointer parameters is invalid. |
Src | is a USM pointer to the source memory. |
Dest | is a USM pointer to the destination memory. |
Count | is a number of elements of type T to copy. |
Definition at line 2412 of file handler.hpp.
References sycl::_V1::detail::memcpy().
|
inline |
Copies the content of memory pointed by Src into the memory object accessed by Dst.
Source must have at least as many bytes as the range accessed by Dst.
Src | is a pointer to source memory. |
Dst | is a destination SYCL accessor. |
Definition at line 2195 of file handler.hpp.
References sycl::_V1::AccessMode, and sycl::_V1::detail::getSyclObjImpl().
|
inline |
Copies the content of memory pointed by Src into the memory object accessed by Dst.
Source must have at least as many bytes as the range accessed by Dst.
Src | is a smart pointer to source memory. |
Dst | is a destination SYCL accessor. |
Definition at line 2132 of file handler.hpp.
References sycl::_V1::AccessMode.
void sycl::_V1::handler::depends_on | ( | const std::vector< event > & | Events | ) |
Registers event dependencies on this command group.
Events | is a vector of valid SYCL events to wait on. |
Definition at line 793 of file handler.cpp.
References sycl::_V1::detail::getSyclObjImpl(), sycl::_V1::invalid, and sycl::_V1::make_error_code().
void sycl::_V1::handler::depends_on | ( | event | Event | ) |
Registers event dependencies on this command group.
Event | is a valid SYCL event to wait on. |
Definition at line 784 of file handler.cpp.
References sycl::_V1::detail::getSyclObjImpl(), sycl::_V1::invalid, and sycl::_V1::make_error_code().
Referenced by sycl::_V1::queue::ext_oneapi_copy2d(), sycl::_V1::queue::ext_oneapi_fill2d(), sycl::_V1::queue::ext_oneapi_memcpy2d(), sycl::_V1::queue::ext_oneapi_memset2d(), sycl::_V1::queue::fill(), sycl::_V1::detail::reduction_impl_algo< T, BinaryOperation, Dims, Extent, RedOutVar >::getGroupsCounterAccDiscrete(), sycl::_V1::queue::parallel_for(), sycl::_V1::queue::parallel_for_impl(), sycl::_V1::queue::prefetch(), sycl::_V1::queue::single_task(), and sycl::_V1::detail::reduction::withAuxHandler().
|
inline |
Prevents any commands submitted afterward to this queue from executing until all commands previously submitted to this queue have entered the complete state.
Definition at line 2362 of file handler.hpp.
Referenced by barrier(), and sycl::_V1::queue::ext_oneapi_submit_barrier().
void sycl::_V1::handler::ext_oneapi_barrier | ( | const std::vector< event > & | WaitList | ) |
Prevents any commands submitted afterward to this queue from executing until all events in WaitList have entered the complete state.
If WaitList is empty, then the barrier has no effect.
WaitList | is a vector of valid SYCL events that need to complete before barrier command can be executed. |
Definition at line 663 of file handler.cpp.
References sycl::_V1::detail::CG::BarrierWaitlist.
|
inline |
Copies data from one 2D memory region to another, both pointed by USM pointers.
No operations is done if
Width | or |
Height | is zero. An exception is thrown if either |
Dest | or |
Src | is nullptr or if |
Width | is strictly greater than either |
DestPitch | or |
SrcPitch. | The behavior is undefined if any of the pointer parameters is invalid. |
Src | is a USM pointer to the source memory. |
SrcPitch | is the pitch of the rows in |
Src. | |
Dest | is a USM pointer to the destination memory. |
DestPitch | is the pitch of the rows in |
Dest. | |
Width | is the width in number of elements of the 2D region to copy. |
Height | is the height in number of rows of the 2D region to copy. |
Definition at line 2496 of file handler.hpp.
References sycl::_V1::make_error_code().
Referenced by sycl::_V1::queue::ext_oneapi_copy2d().
|
inline |
Fills the memory pointed by a USM pointer with the value specified.
No operations is done if
Width | or |
Height | is zero. An exception is thrown if either |
Dest | or |
Src | is nullptr or if |
Width | is strictly greater than |
DestPitch. | The behavior is undefined if any of the pointer parameters is invalid. |
Dest | is a USM pointer to the destination memory. |
DestPitch | is the pitch of the rows in |
Dest. | |
Pattern | is the pattern to fill into the memory. T should be trivially copyable. |
Width | is the width in number of elements of the 2D region to fill. |
Height | is the height in number of rows of the 2D region to fill. |
Definition at line 2562 of file handler.hpp.
References sycl::_V1::make_error_code().
Referenced by sycl::_V1::queue::ext_oneapi_fill2d().
|
inline |
Copies data from one 2D memory region to another, both pointed by USM pointers.
No operations is done if
Width | or |
Height | is zero. An exception is thrown if either |
Dest | or |
Src | is nullptr or if |
Width | is strictly greater than either |
DestPitch | or |
SrcPitch. | The behavior is undefined if any of the pointer parameters is invalid. |
NOTE: Function is dependent to prevent the fallback kernels from materializing without the use of the function.
Dest | is a USM pointer to the destination memory. |
DestPitch | is the pitch of the rows in |
Dest. | |
Src | is a USM pointer to the source memory. |
SrcPitch | is the pitch of the rows in |
Src. | |
Width | is the width in bytes of the 2D region to copy. |
Height | is the height in number of row of the 2D region to copy. |
Definition at line 2461 of file handler.hpp.
References sycl::_V1::make_error_code().
Referenced by sycl::_V1::queue::ext_oneapi_memcpy2d().
|
inline |
Fills the memory pointed by a USM pointer with the value specified.
No operations is done if
Width | or |
Height | is zero. An exception is thrown if either |
Dest | or |
Src | is nullptr or if |
Width | is strictly greater than |
DestPitch. | The behavior is undefined if any of the pointer parameters is invalid. |
NOTE: Function is dependent to prevent the fallback kernels from materializing without the use of the function.
Dest | is a USM pointer to the destination memory. |
DestPitch | is the pitch of the rows in |
Dest. | |
Value | is the value to fill into the region in |
Dest. | Value is cast as an unsigned char. |
Width | is the width in number of elements of the 2D region to fill. |
Height | is the height in number of rows of the 2D region to fill. |
Definition at line 2533 of file handler.hpp.
References sycl::_V1::make_error_code().
Referenced by sycl::_V1::queue::ext_oneapi_memset2d().
|
inline |
Fills memory pointed by accessor with the pattern given.
If the operation is submitted to queue associated with OpenCL device and accessor points to one dimensional memory object then use special type for filling. Otherwise fill using regular kernel.
Dst | is a destination SYCL accessor. |
Pattern | is a value to be used to fill the memory. |
Definition at line 2311 of file handler.hpp.
References sycl::_V1::AccessMode, sycl::_V1::detail::getSyclObjImpl(), and sycl::_V1::IsPlaceholder.
Referenced by sycl::_V1::queue::fill().
|
inline |
Fills the specified memory with the specified pattern.
Ptr | is the pointer to the memory to fill |
Pattern | is the pattern to fill into the memory. T should be trivially copyable. |
Count | is the number of times to fill Pattern into Ptr. |
Definition at line 2349 of file handler.hpp.
|
inline |
Definition at line 1454 of file handler.hpp.
References sycl::_V1::make_error_code().
|
inline |
Enqueues a command to the SYCL runtime to invoke Func
once.
Definition at line 1602 of file handler.hpp.
void sycl::_V1::handler::mem_advise | ( | const void * | Ptr, |
size_t | Length, | ||
int | Advice | ||
) |
Provides additional information to the underlying runtime about how different allocations are used.
Ptr | is a USM pointer to the allocation. |
Length | is a number of bytes in the allocation. |
Advice | is a device-defined advice for the specified allocation. |
Definition at line 715 of file handler.cpp.
References sycl::_V1::detail::CG::AdviseUSM.
|
inline |
Copies data from a USM memory region to a device_global.
Throws an exception if the copy operation intends to write outside the memory range
Dest,as | specified through |
NumBytes | and |
DestOffset. | |
Dest | is the destination device_glboal. |
Src | is a USM pointer to the source memory. |
NumBytes | is a number of bytes to copy. |
DestOffset | is the offset into |
Dest | to copy to. |
Definition at line 2590 of file handler.hpp.
References sycl::_V1::ext::oneapi::experimental::has_property(), and sycl::_V1::make_error_code().
|
inline |
Copies data from a device_global to USM memory.
Throws an exception if the copy operation intends to read outside the memory range
Src,as | specified through |
NumBytes | and |
SrcOffset. | |
Dest | is a USM pointer to copy to. |
Src | is the source device_global. |
NumBytes | is a number of bytes to copy. |
SrcOffset | is the offset into |
Src | to copy from. |
Definition at line 2613 of file handler.hpp.
References sycl::_V1::ext::oneapi::experimental::has_property(), and sycl::_V1::make_error_code().
void sycl::_V1::handler::memcpy | ( | void * | Dest, |
const void * | Src, | ||
size_t | Count | ||
) |
Copies data from one memory region to another, each is either a host pointer or a pointer within USM allocation accessible on this handler's device.
No operations is done if
Count | is zero. An exception is thrown if either |
Dest | or |
Src | is nullptr. The behavior is undefined if any of the pointer parameters is invalid. |
Dest | is a USM pointer to the destination memory. |
Src | is a USM pointer to the source memory. |
Count | is a number of bytes to copy. |
Definition at line 692 of file handler.cpp.
References sycl::_V1::detail::CG::CopyUSM.
void sycl::_V1::handler::memset | ( | void * | Dest, |
int | Value, | ||
size_t | Count | ||
) |
Fills the memory pointed by a USM pointer with the value specified.
No operations is done if
Count | is zero. An exception is thrown if |
Dest | is nullptr. The behavior is undefined if |
Dest | is invalid. |
Dest | is a USM pointer to the memory to fill. |
Value | is a value to be set. Value is cast as an unsigned char. |
Count | is a number of bytes to fill. |
Definition at line 700 of file handler.cpp.
References sycl::_V1::detail::CG::FillUSM.
|
inline |
Defines and invokes a SYCL kernel function for the specified range and offsets.
Kernel | is a SYCL kernel that is executed on a SYCL device (except for the host device). |
NDRange | is a ND-range defining global and local sizes as well as offset. |
KernelFunc | is a lambda that is used if device, queue is bound to, is a host device. |
Definition at line 1876 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
|
inline |
Defines and invokes a SYCL kernel function for the specified range.
Kernel | is a SYCL kernel that is executed on a SYCL device (except for the host device). |
NumWorkItems | is a range defining indexing space. |
KernelFunc | is a lambda that is used if device, queue is bound to, is a host device. |
Definition at line 1800 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
|
inline |
Defines and invokes a SYCL kernel function for the specified range and offsets.
Kernel | is a SYCL kernel that is executed on a SYCL device (except for the host device). |
NumWorkItems | is a range defining indexing space. |
WorkItemOffset | is an offset to be applied to each work item index. |
KernelFunc | is a lambda that is used if device, queue is bound to, is a host device. |
Definition at line 1838 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
|
inline |
Defines and invokes a SYCL kernel function for the specified range and offsets.
The SYCL kernel function is defined as SYCL kernel object.
NDRange | is a ND-range defining global and local sizes as well as offset. |
Kernel | is a SYCL kernel function. |
Definition at line 1738 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
|
inline |
Definition at line 2025 of file handler.hpp.
|
inline |
Definition at line 2058 of file handler.hpp.
|
inline |
Definition at line 2066 of file handler.hpp.
|
inline |
Definition at line 1563 of file handler.hpp.
Referenced by sycl::_V1::detail::reduAuxCGFunc(), sycl::_V1::detail::reduCGFuncMulti(), sycl::_V1::detail::NDRangeReduction< reduction::strategy::local_atomic_and_atomic_cross_wg >::run(), sycl::_V1::detail::NDRangeReduction< reduction::strategy::group_reduce_and_last_wg_detection >::run(), sycl::_V1::detail::NDRangeReduction< reduction::strategy::range_basic >::run(), sycl::_V1::detail::NDRangeReduction< reduction::strategy::group_reduce_and_atomic_cross_wg >::run(), sycl::_V1::detail::NDRangeReduction< reduction::strategy::local_mem_tree_and_atomic_cross_wg >::run(), sycl::_V1::detail::NDRangeReduction< reduction::strategy::group_reduce_and_multiple_kernels >::run(), and sycl::_V1::detail::NDRangeReduction< reduction::strategy::basic >::run().
Definition at line 1697 of file handler.hpp.
|
inline |
Definition at line 1995 of file handler.hpp.
|
inline |
Definition at line 1568 of file handler.hpp.
Definition at line 1701 of file handler.hpp.
|
inline |
Definition at line 2005 of file handler.hpp.
|
inline |
Definition at line 1573 of file handler.hpp.
Definition at line 1705 of file handler.hpp.
|
inline |
Definition at line 2015 of file handler.hpp.
|
inline |
Defines and invokes a SYCL kernel function for the specified range and offset.
The SYCL kernel function is defined as a lambda function or a named function object type and given an id or item for indexing in the indexing space defined by range. If it is a named function object and the function object type is globally visible, there is no need for the developer to provide a kernel name for it.
NumWorkItems | is a range defining indexing space. |
WorkItemOffset | is an offset to be applied to each work item index. |
KernelFunc | is a SYCL kernel function. |
Definition at line 1622 of file handler.hpp.
|
inline |
Defines and invokes a SYCL kernel function for the specified range and offsets.
The SYCL kernel function is defined as SYCL kernel object.
NumWorkItems | is a range defining indexing space. |
WorkItemOffset | is an offset to be applied to each work item index. |
Kernel | is a SYCL kernel function. |
Definition at line 1719 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
|
inline |
Reductions.
Definition at line 2038 of file handler.hpp.
|
inline |
Definition at line 2046 of file handler.hpp.
|
inline |
Hierarchical kernel invocation method of a kernel.
This version of parallel_for_work_group
takes two parameters representing the same kernel. The first one - Kernel
- is a compiled form of the second one - kernelFunc
, which is the source form of the kernel. The same source kernel can be compiled multiple times yielding multiple kernel class objects accessible via the program
class interface.
Kernel | is a compiled SYCL kernel. |
NumWorkGroups | is a range describing the number of work-groups in each dimension. |
KernelFunc | is a lambda representing kernel. |
Definition at line 1918 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
|
inline |
Hierarchical kernel invocation method of a kernel.
This version of parallel_for_work_group
takes two parameters representing the same kernel. The first one - Kernel
- is a compiled form of the second one - kernelFunc
, which is the source form of the kernel. The same source kernel can be compiled multiple times yielding multiple kernel class objects accessible via the program
class interface.
Kernel | is a compiled SYCL kernel. |
NumWorkGroups | is a range describing the number of work-groups in each dimension. |
WorkGroupSize | is a range describing the size of work-groups in each dimension. |
KernelFunc | is a lambda representing kernel. |
Definition at line 1957 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl(), and sycl::_V1::ext::oneapi::experimental::detail::WorkGroupSize.
|
inline |
Hierarchical kernel invocation method of a kernel defined as a lambda encoding the body of each work-group to launch.
Lambda may contain multiple calls to parallel_for_work_item(...) methods representing the execution on each work-item. Launches NumWorkGroups work-groups of runtime-defined size.
NumWorkGroups | is a range describing the number of work-groups in each dimension. |
KernelFunc | is a lambda representing kernel. |
Definition at line 1652 of file handler.hpp.
|
inline |
}@
Definition at line 2076 of file handler.hpp.
|
inline |
Hierarchical kernel invocation method of a kernel defined as a lambda encoding the body of each work-group to launch.
Lambda may contain multiple calls to parallel_for_work_item(...) methods representing the execution on each work-item. Launches NumWorkGroups work-groups of WorkGroupSize size.
NumWorkGroups | is a range describing the number of work-groups in each dimension. |
WorkGroupSize | is a range describing the size of work-groups in each dimension. |
KernelFunc | is a lambda representing kernel. |
Definition at line 1671 of file handler.hpp.
References sycl::_V1::ext::oneapi::experimental::detail::WorkGroupSize.
|
inline |
Definition at line 2084 of file handler.hpp.
References sycl::_V1::ext::oneapi::experimental::detail::WorkGroupSize.
void sycl::_V1::handler::prefetch | ( | const void * | Ptr, |
size_t | Count | ||
) |
Provides hints to the runtime library that data should be made available on a device earlier than Unified Shared Memory would normally require it to be available.
Ptr | is a USM pointer to the memory to be prefetched to the device. |
Count | is a number of bytes to be prefetched. |
Definition at line 708 of file handler.cpp.
References sycl::_V1::detail::CG::PrefetchUSM.
Referenced by sycl::_V1::queue::prefetch().
|
inline |
Requires access to the memory object associated with the placeholder accessor.
Calling this function with a non-placeholder accessor has no effect.
The command group has a requirement to gain access to the given memory object before executing.
Acc | is a SYCL accessor describing required memory region. |
Definition at line 1482 of file handler.hpp.
References sycl::_V1::detail::associateWithHandler().
Referenced by sycl::_V1::queue::copy(), sycl::_V1::queue::fill(), and sycl::_V1::queue::update_host().
|
inline |
Definition at line 1531 of file handler.hpp.
|
inline |
Definition at line 1537 of file handler.hpp.
|
inline |
Sets argument for OpenCL interoperability kernels.
Registers Arg passed as argument # ArgIndex.
ArgIndex | is a positional number of argument to be set. |
Arg | is an argument value to be set. |
Definition at line 1524 of file handler.hpp.
|
inline |
Sets arguments for OpenCL interoperability kernels.
Registers pack of arguments(Args) with indexes starting from 0.
Args | are argument values to be set. |
Definition at line 1546 of file handler.hpp.
|
inline |
Definition at line 1439 of file handler.hpp.
|
inline |
Defines and invokes a SYCL kernel function as a function object type.
If it is a named function object and the function object type is globally visible, there is no need for the developer to provide a kernel name for it.
KernelFunc | is a SYCL kernel function. |
Definition at line 1558 of file handler.hpp.
Referenced by sycl::_V1::detail::reduSaveFinalResultToUserMem(), and sycl::_V1::detail::reduction_impl_algo< T, BinaryOperation, Dims, Extent, RedOutVar >::withInitializedMem().
|
inline |
Invokes a SYCL kernel.
Executes exactly once. The kernel invocation method has no functors and cannot be called on host.
Kernel | is a SYCL kernel object. |
Definition at line 1684 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
|
inline |
Defines and invokes a SYCL kernel function.
Kernel | is a SYCL kernel that is executed on a SYCL device (except for the host device). |
KernelFunc | is a lambda that is used if device, queue is bound to, is a host device. |
Definition at line 1755 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
|
inline |
Definition at line 1987 of file handler.hpp.
|
inline |
Provides guarantees that the memory object accessed via Acc is updated on the host after command group object execution is complete.
Acc | is a SYCL accessor that needs to be updated on host. |
Definition at line 2284 of file handler.hpp.
References sycl::_V1::detail::getSyclObjImpl().
Referenced by sycl::_V1::queue::update_host().
void sycl::_V1::handler::use_kernel_bundle | ( | const kernel_bundle< bundle_state::executable > & | ExecBundle | ) |
Definition at line 760 of file handler.cpp.
References sycl::_V1::kernel_bundle< State >::get_context(), sycl::_V1::detail::getSyclObjImpl(), sycl::_V1::invalid, and sycl::_V1::make_error_code().
|
friend |
Definition at line 2764 of file handler.hpp.
|
friend |
Definition at line 2727 of file handler.hpp.
|
friend |
|
friend |
|
friend |
Definition at line 2732 of file handler.hpp.
|
friend |
Definition at line 2722 of file handler.hpp.
|
friend |
|
friend |
|
friend |
Definition at line 2740 of file handler.hpp.
|
friend |
|
friend |
|
friend |
Definition at line 2735 of file handler.hpp.
|
friend |
Definition at line 2734 of file handler.hpp.