ESIMD APIs to access memory via accessors, USM pointers, perform per-element atomic operations. More...

Collaboration diagram for Memory access API.:

Modules
	Atomic memory access.
	Memory access functions which perform per-lane atomic update using given operation. "Per-lane" means that the atomicity guarantees of a vector atomic operation are the same as of N independent scalar atomic operations per lane (N is number of lanes).

	Shared local memory access functions.

	LSC-specific memory access APIs.
	This group combines types and functions specific to LSC, which is available in Intel GPUs starting from PVC and ACM.

	Named barrier APIs.

	HW thread .

Namespaces
	sycl::_V1::ext::intel::esimd::detail

Enumerations
enum	sycl::_V1::ext::intel::esimd::fence_mask : uint8_t { sycl::_V1::ext::intel::esimd::global_coherent_fence = 0x1, sycl::_V1::ext::intel::esimd::l3_flush_instructions = 0x2, sycl::_V1::ext::intel::esimd::l3_flush_texture_data = 0x4, sycl::_V1::ext::intel::esimd::l3_flush_constant_data = 0x8, sycl::_V1::ext::intel::esimd::l3_flush_rw_data = 0x10, sycl::_V1::ext::intel::esimd::local_barrier = 0x20, sycl::_V1::ext::intel::esimd::l1_flush_ro_data = 0x40, sycl::_V1::ext::intel::esimd::sw_barrier = 0x80 }
	Represetns a bit mask to control behavior of esimd::fence. More...

Functions
template<typename AccessorTy >
__ESIMD_API SurfaceIndex	sycl::_V1::ext::intel::esimd::get_surface_index (AccessorTy acc)
	Get surface index corresponding to a SYCL accessor. More...

template<typename Tx , int N, typename Toffset >
__ESIMD_API simd< Tx, N >	sycl::_V1::ext::intel::esimd::gather (const Tx *p, simd< Toffset, N > offsets, simd_mask< N > mask=1)
	Loads ("gathers") elements from different memory locations and returns a vector of them. More...

template<typename Tx , int N, typename Toffset , typename RegionTy = region1d_t<Toffset, N, 1>>
__ESIMD_API simd< Tx, N >	sycl::_V1::ext::intel::esimd::gather (const Tx *p, simd_view< Toffset, RegionTy > offsets, simd_mask< N > mask=1)
	A variation of `gather` API with `offsets` represented as `simd_view` object. More...

template<typename Tx , int N, typename Toffset >
__ESIMD_API std::enable_if_t< std::is_integral_v< Toffset >, simd< Tx, N > >	sycl::_V1::ext::intel::esimd::gather (const Tx *p, Toffset offset, simd_mask< N > mask=1)
	A variation of `gather` API with `offsets` represented as scalar. More...

template<typename Tx , int N, typename Toffset >
__ESIMD_API void	sycl::_V1::ext::intel::esimd::scatter (Tx *p, simd< Toffset, N > offsets, simd< Tx, N > vals, simd_mask< N > mask=1)
	Writes ("scatters") elements of the input vector to different memory locations. More...

template<typename Tx , int N, typename Toffset , typename RegionTy = region1d_t<Toffset, N, 1>>
__ESIMD_API void	sycl::_V1::ext::intel::esimd::scatter (Tx *p, simd_view< Toffset, RegionTy > offsets, simd< Tx, N > vals, simd_mask< N > mask=1)
	A variation of `scatter` API with `offsets` represented as `simd_view` object. More...

template<typename Tx , int N, typename Toffset >
__ESIMD_API std::enable_if_t< std::is_integral_v< Toffset > &&N==1 >	sycl::_V1::ext::intel::esimd::scatter (Tx *p, Toffset offset, simd< Tx, N > vals, simd_mask< N > mask=1)
	A variation of `scatter` API with `offsets` represented as scalar. More...

template<typename Tx , int N, typename Flags = overaligned_tag<detail::OperandSize::OWORD>>
__ESIMD_API std::enable_if_t< is_simd_flag_type_v< Flags >, simd< Tx, N > >	sycl::_V1::ext::intel::esimd::block_load (const Tx *addr, Flags={})
	Loads a contiguous block of memory from the given memory address `addr` and returns the loaded data as a vector. More...

template<typename Tx , int N, typename AccessorTy , typename Flags = vector_aligned_tag, typename = std::enable_if_t< is_simd_flag_type_v<Flags> && sycl::detail::acc_properties::is_accessor_v<AccessorTy> && !sycl::detail::acc_properties::is_local_accessor_v<AccessorTy>>, class T = detail::__raw_t<Tx>>
__ESIMD_API simd< Tx, N >	sycl::_V1::ext::intel::esimd::block_load (AccessorTy acc, uint32_t offset, Flags={})
	Loads a contiguous block of memory from given accessor and offset and returns the loaded data as a vector. More...

template<typename Tx , int N, typename Flags = overaligned_tag<detail::OperandSize::OWORD>>
__ESIMD_API __ESIMD_API std::enable_if_t< is_simd_flag_type_v< Flags > >	sycl::_V1::ext::intel::esimd::block_store (Tx *addr, simd< Tx, N > vals, Flags={})
	Stores elements of the vector `vals` to a contiguous block of memory at the given address `addr`. More...

template<typename Tx , int N, typename AccessorTy , class T = detail::__raw_t<Tx>>
__ESIMD_API std::enable_if_t< sycl::detail::acc_properties::is_accessor_v< AccessorTy > &&!sycl::detail::acc_properties::is_local_accessor_v< AccessorTy > >	sycl::_V1::ext::intel::esimd::block_store (AccessorTy acc, uint32_t offset, simd< Tx, N > vals)
	Stores elements of a vector to a contiguous block of memory represented by an accessor and an offset within this accessor. More...

template<typename T , int N, typename AccessorTy >
__ESIMD_API std::enable_if_t<(sizeof(T)<=4) &&(N==1\|\|N==8\|\|N==16\|\|N==32) &&sycl::detail::acc_properties::is_accessor_v< AccessorTy > &&!sycl::detail::acc_properties::is_local_accessor_v< AccessorTy >, simd< T, N > >	sycl::_V1::ext::intel::esimd::gather (AccessorTy acc, simd< uint32_t, N > offsets, uint32_t glob_offset=0, simd_mask< N > mask=1)

template<typename T , int N, typename AccessorTy >
__ESIMD_API std::enable_if_t<(sizeof(T)<=4) &&(N==1\|\|N==8\|\|N==16\|\|N==32) &&sycl::detail::acc_properties::is_accessor_v< AccessorTy > &&!sycl::detail::acc_properties::is_local_accessor_v< AccessorTy > >	sycl::_V1::ext::intel::esimd::scatter (AccessorTy acc, simd< uint32_t, N > offsets, simd< T, N > vals, uint32_t glob_offset=0, simd_mask< N > mask=1)

template<typename T , typename AccessorTy >
__ESIMD_API T	sycl::_V1::ext::intel::esimd::scalar_load (AccessorTy acc, uint32_t offset)
	Load a scalar value from an accessor. More...

template<typename T , typename AccessorTy >
__ESIMD_API void	sycl::_V1::ext::intel::esimd::scalar_store (AccessorTy acc, uint32_t offset, T val)
	Store a scalar value into an accessor. More...

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset >
__ESIMD_API simd< T, N *get_num_channels_enabled(RGBAMask)>	sycl::_V1::ext::intel::esimd::gather_rgba (const T *p, simd< Toffset, N > offsets, simd_mask< N > mask=1)

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset , typename RegionTy = region1d_t<Toffset, N, 1>>
__ESIMD_API simd< T, N *get_num_channels_enabled(RGBAMask)>	sycl::_V1::ext::intel::esimd::gather_rgba (const T *p, simd_view< Toffset, RegionTy > offsets, simd_mask< N > mask=1)
	A variation of `gather_rgba` API with `offsets` represented as `simd_view` object. More...

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset >
__ESIMD_API std::enable_if_t< std::is_integral_v< Toffset >, simd< T, N *get_num_channels_enabled(RGBAMask)> >	sycl::_V1::ext::intel::esimd::gather_rgba (const T *p, Toffset offset, simd_mask< N > mask=1)
	A variation of `gather_rgba` API with `offsets` represented as scalar. More...

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset >
__ESIMD_API void	sycl::_V1::ext::intel::esimd::scatter_rgba (T p, simd< Toffset, N > offsets, simd< T, N get_num_channels_enabled(RGBAMask)> vals, simd_mask< N > mask=1)

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset , typename RegionTy = region1d_t<Toffset, N, 1>>
__ESIMD_API void	sycl::_V1::ext::intel::esimd::scatter_rgba (T p, simd_view< Toffset, RegionTy > offsets, simd< T, N get_num_channels_enabled(RGBAMask)> vals, simd_mask< N > mask=1)
	A variation of `scatter_rgba` API with `offsets` represented as `simd_view` object. More...

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset >
__ESIMD_API std::enable_if_t< std::is_integral_v< Toffset > &&N==1 >	sycl::_V1::ext::intel::esimd::scatter_rgba (T p, Toffset offset, simd< T, N get_num_channels_enabled(RGBAMask)> vals, simd_mask< N > mask=1)
	A variation of `scatter_rgba` API with `offsets` represented as scalar. More...

template<typename T , int N, rgba_channel_mask RGBAMask>
__ESIMD_API std::enable_if_t<(N==8\|\|N==16\|\|N==32) &&sizeof(T)==4 >	sycl::_V1::ext::intel::esimd::scatter_rgba (T p, simd< uint32_t, N > offsets, simd< T, N get_num_channels_enabled(RGBAMask)> vals, simd_mask< N > mask=1)

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename AccessorT , int N, typename T = typename AccessorT::value_type>
__ESIMD_API std::enable_if_t<((N==8\|\|N==16\|\|N==32) &&sizeof(T)==4 &&!std::is_pointer_v< AccessorT > &&!sycl::detail::acc_properties::is_local_accessor_v< AccessorT >), simd< T, N *get_num_channels_enabled(RGBAMask)> >	sycl::_V1::ext::intel::esimd::gather_rgba (AccessorT acc, simd< uint32_t, N > offsets, uint32_t global_offset=0, simd_mask< N > mask=1)
	Gather and transpose pixels from the given memory locations defined by the base specified by `acc`, the global offset `global_offset` and a vector of offsets `offsets`. More...

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename AccessorT , int N, typename T = typename AccessorT::value_type>
__ESIMD_API std::enable_if_t<(N==8\|\|N==16\|\|N==32) &&sizeof(T)==4 &&!std::is_pointer_v< AccessorT > &&!sycl::detail::acc_properties::is_local_accessor_v< AccessorT > >	sycl::_V1::ext::intel::esimd::scatter_rgba (AccessorT acc, simd< uint32_t, N > offsets, simd< T, N *get_num_channels_enabled(RGBAMask)> vals, uint32_t global_offset=0, simd_mask< N > mask=1)
	Gather data from the memory addressed by accessor `acc`, offset common for all loaded elements `global_offset` and per-element offsets `offsets`, and return it as simd vector. More...

template<uint8_t cntl>
__ESIMD_API void	sycl::_V1::ext::intel::esimd::fence ()
	esimd::fence sets the memory read/write order. More...

__ESIMD_API void	sycl::_V1::ext::intel::esimd::fence (fence_mask cntl)

__ESIMD_API void	sycl::_V1::ext::intel::esimd::barrier ()
	Generic work-group barrier. More...

template<typename T , int m, int N, typename AccessorTy , unsigned plane = 0>
__ESIMD_API simd< T, m *N >	sycl::_V1::ext::intel::esimd::media_block_load (AccessorTy acc, unsigned x, unsigned y)
	Media block load. More...

template<typename T , int m, int N, typename AccessorTy , unsigned plane = 0>
__ESIMD_API void	sycl::_V1::ext::intel::esimd::media_block_store (AccessorTy acc, unsigned x, unsigned y, simd< T, m *N > vals)
	Media block store. More...

template<typename Tx , int N, typename AccessorTy , typename Flags = overaligned_tag<detail::OperandSize::OWORD>>
__ESIMD_API std::enable_if_t< sycl::detail::acc_properties::is_local_accessor_v< AccessorTy > &&is_simd_flag_type_v< Flags >, simd< Tx, N > >	sycl::_V1::ext::intel::esimd::block_load (AccessorTy acc, uint32_t offset, Flags={})
	Loads a contiguous block of SLM memory referenced by the given local-accessor `acc` and byte-offset `offset`, then returns the loaded data as a simd object. More...

template<typename Tx , int N, typename AccessorTy , typename Flags = overaligned_tag<detail::OperandSize::OWORD>>
__ESIMD_API std::enable_if_t< sycl::detail::acc_properties::is_local_accessor_v< AccessorTy > &&is_simd_flag_type_v< Flags > >	sycl::_V1::ext::intel::esimd::block_store (AccessorTy acc, uint32_t offset, simd< Tx, N > vals, Flags={})
	Variant of block_store that uses local accessor as a parameter. More...

template<split_barrier_action flag>
__ESIMD_API void	sycl::_V1::ext::intel::experimental::esimd::split_barrier ()
	Generic work-group split barrier. More...

__ESIMD_API void	sycl::_V1::ext::intel::experimental::esimd::split_barrier (split_barrier_action flag)

Detailed Description

ESIMD APIs to access memory via accessors, USM pointers, perform per-element atomic operations.

Enumeration Type Documentation

◆ fence_mask

enum sycl::_V1::ext::intel::esimd::fence_mask : uint8_t

Represetns a bit mask to control behavior of esimd::fence.

Enum elements define semantics of the bits in the mask.

Enumerator
global_coherent_fence	“Commit enable” - wait for fence to complete before continuing.
l3_flush_instructions	Flush the instruction cache.
l3_flush_texture_data	Flush sampler (texture) cache.
l3_flush_constant_data	Flush constant cache.
l3_flush_rw_data	Flush constant cache.
local_barrier	Issue SLM memory barrier only. If not set, the memory barrier is global.
l1_flush_ro_data	Flush L1 read - only data cache.
sw_barrier	Creates a software (compiler) barrier, which does not generate any instruction and only prevents instruction scheduler from reordering instructions across this barrier at compile time.

Definition at line 2012 of file memory.hpp.

Function Documentation

◆ barrier()

__ESIMD_API void sycl::_V1::ext::intel::esimd::barrier ( )

Generic work-group barrier.

Performs barrier synchronization for all threads within the same thread group. The barrier instruction causes the executing thread to wait until all threads in the same thread group have executed the barrier instruction. Memory ordering is also guaranteed by this instruction. The behavior is undefined if this instruction is executed in divergent control flow.

Definition at line 2049 of file memory.hpp.

References sycl::_V1::ext::intel::esimd::global_coherent_fence, and sycl::_V1::ext::intel::esimd::local_barrier.

Referenced by sycl::_V1::ext::intel::experimental::esimd::named_barrier_signal().

◆ block_load() [1/3]

template<typename Tx , int N, typename AccessorTy , typename Flags = vector_aligned_tag, typename = std::enable_if_t< is_simd_flag_type_v<Flags> && sycl::detail::acc_properties::is_accessor_v<AccessorTy> && !sycl::detail::acc_properties::is_local_accessor_v<AccessorTy>>, class T = detail::__raw_t<Tx>>

__ESIMD_API simd<Tx, N> sycl::_V1::ext::intel::esimd::block_load	(	AccessorTy	acc,
		uint32_t	offset,
		Flags	= `{}`
	)

Loads a contiguous block of memory from given accessor and offset and returns the loaded data as a vector.

Actual code generated depends on the alignment parameter.

Template Parameters

Tx	Element type.
N	Number of elements to load, `N * sizeof(Tx)` must be 1, 2, 4 or 8 owords long.
AccessorTy	Accessor type (auto-deduced).
Flags	The alignment specifier type tag. Auto-deduced from the `Flags` parameter. If it is less than `16`, then slower unaligned access is generated, otherwise the access is aligned.

Parameters

acc	The accessor.
offset	The offset to load from in bytes.
Flags	Specifies the alignment.

Returns: A vector of loaded elements.

Definition at line 313 of file memory.hpp.

◆ block_load() [2/3]

template<typename Tx , int N, typename AccessorTy , typename Flags = overaligned_tag<detail::OperandSize::OWORD>>

__ESIMD_API std::enable_if_t< sycl::detail::acc_properties::is_local_accessor_v<AccessorTy> && is_simd_flag_type_v<Flags>, simd<Tx, N> > sycl::_V1::ext::intel::esimd::block_load	(	AccessorTy	acc,
		uint32_t	offset,
		Flags	= `{}`
	)

Loads a contiguous block of SLM memory referenced by the given local-accessor acc and byte-offset offset, then returns the loaded data as a simd object.

The generated code depends on the combination {T, N, Flags}. Providing flags specifying the alignment of 16-bytes or more produces more efficient code. If the alignment is smaller than 16-bytes, then less efficient gather is generated. If the loaded vector is too long for 1 flat-load GPU instruction, then a series of flat-loads and/or gathers may be generated.

Template Parameters

Tx	Element type.
N	Number of elements to load.
AccessorTy	Accessor type (auto-deduced).
Flags	The alignment specifier type tag.

Parameters

acc	The local accessor.
offset	The offset to load from in bytes.
Flags	Specifies the alignment.

Returns: A vector of loaded elements.

Definition at line 2165 of file memory.hpp.

◆ block_load() [3/3]

template<typename Tx , int N, typename Flags = overaligned_tag<detail::OperandSize::OWORD>>

__ESIMD_API std::enable_if_t<is_simd_flag_type_v<Flags>, simd<Tx, N> > sycl::_V1::ext::intel::esimd::block_load	(	const Tx *	addr,
		Flags	= `{}`
	)

Loads a contiguous block of memory from the given memory address addr and returns the loaded data as a vector.

The generated code depends on the combination {T, N, Flags}. Providing flags specifying the alignment of 16-bytes or more produces more efficient code. If the alignment is smaller than 16-bytes, then less efficient gather is generated. If the loaded vector is too long for 1 flat-load GPU instruction, then a series of flat-loads and/or gathers may be generated.

Template Parameters

Tx	Element type.
N	Number of elements to load.
Flags	The alignment specifier type tag.

Parameters

addr	The address to load from.
Flags	Specifies the alignment.

Returns: A vector of loaded elements.

Definition at line 283 of file memory.hpp.

◆ block_store() [1/3]

template<typename Tx , int N, typename AccessorTy , class T = detail::__raw_t<Tx>>

__ESIMD_API std::enable_if_t< sycl::detail::acc_properties::is_accessor_v<AccessorTy> && !sycl::detail::acc_properties::is_local_accessor_v<AccessorTy> > sycl::_V1::ext::intel::esimd::block_store	(	AccessorTy	acc,
		uint32_t	offset,
		simd< Tx, N >	vals
	)

Stores elements of a vector to a contiguous block of memory represented by an accessor and an offset within this accessor.

Template Parameters

Tx	Element type.
N	Number of elements to store, `N * sizeof(Tx)` must be 1, 2, 4 or 8 owords long.
AccessorTy	Accessor type (auto-deduced).

Parameters

acc	The accessor to store to.
offset	The offset to store at. It is in bytes and must be a multiple of `16`.
vals	The vector to store.

Definition at line 386 of file memory.hpp.

References sycl::_V1::ext::intel::esimd::detail::isPowerOf2().

◆ block_store() [2/3]

template<typename Tx , int N, typename AccessorTy , typename Flags = overaligned_tag<detail::OperandSize::OWORD>>

__ESIMD_API std::enable_if_t< sycl::detail::acc_properties::is_local_accessor_v<AccessorTy> && is_simd_flag_type_v<Flags> > sycl::_V1::ext::intel::esimd::block_store	(	AccessorTy	acc,
		uint32_t	offset,
		simd< Tx, N >	vals,
		Flags	= `{}`
	)

Variant of block_store that uses local accessor as a parameter.

Stores elements of the vector vals to a contiguous block of SLM memory represented by the given local accessor and the byte-offset offset. The generated code depends on the combination {T, N, Flags}. Providing flags specifying the alignment of 16-bytes or more produces more efficient code. If the alignment is smaller than 16-bytes, then less efficient scatter is generated. If the stored vector is too long for 1 flat-store GPU instruction, then a series of flat-store and/or scatters may be generated.

Template Parameters

Tx	Element type.
N	Number of elements to store.
AccessorTy	Accessor type (auto-deduced).

Parameters

acc	The local accessor to store to.
offset	The byte-offset to store at.
vals	The vector to store.
Flags	Specifies the alignment.

Definition at line 2192 of file memory.hpp.

◆ block_store() [3/3]

template<typename Tx , int N, typename Flags = overaligned_tag<detail::OperandSize::OWORD>>

__ESIMD_API __ESIMD_API std::enable_if_t<is_simd_flag_type_v<Flags> > sycl::_V1::ext::intel::esimd::block_store	(	Tx *	addr,
		simd< Tx, N >	vals,
		Flags	= `{}`
	)

Stores elements of the vector vals to a contiguous block of memory at the given address addr.

The generated code depends on the combination {T, N, Flags}. Providing flags specifying the alignment of 16-bytes or more produces more efficient code. If the alignment is smaller than 16-bytes, then less efficient scatter is generated. If the stored vector is too long for 1 flat-store GPU instruction, then a series of flat-store and/or scatters may be generated.

Template Parameters

Tx	Element type.
N	Number of elements to store.
Flags	The alignment specifier type tag.

Parameters

addr	The memory address to store at.
vals	The vector to store.
Flags	Specifies the alignment.

Definition at line 362 of file memory.hpp.

◆ fence() [1/2]

template<uint8_t cntl>

__ESIMD_API void sycl::_V1::ext::intel::esimd::fence ( )

esimd::fence sets the memory read/write order.

Template Parameters

cntl	A bitmask composed from `fence_mask` bits.

Definition at line 2036 of file memory.hpp.

◆ fence() [2/2]

__ESIMD_API void sycl::_V1::ext::intel::esimd::fence ( fence_mask cntl )

Definition at line 2039 of file memory.hpp.

◆ gather() [1/4]

template<typename T , int N, typename AccessorTy >

__ESIMD_API std::enable_if_t< (sizeof(T) <= 4) && (N == 1 \|\| N == 8 \|\| N == 16 \|\| N == 32) && sycl::detail::acc_properties::is_accessor_v<AccessorTy> && !sycl::detail::acc_properties::is_local_accessor_v<AccessorTy>, simd<T, N> > sycl::_V1::ext::intel::esimd::gather	(	AccessorTy	acc,
		simd< uint32_t, N >	offsets,
		uint32_t	glob_offset = `0`,
		simd_mask< N >	mask = `1`
	)

Accessor-based gather.

Collects elements located at given offsets in an accessor and returns them as a single simd object. An element can be a 1, 2 or 4-byte value.

Template Parameters

T	Element type; can only be a 1,2,4-byte integer, `sycl::half` or `float`.
N	The number of vector elements. Can be `1`, `8`, `16` or `32`.
AccessorTy	The accessor type.

Parameters

acc	The accessor to gather from.
offsets	Per-element offsets in bytes.
glob_offset	Offset in bytes added to each individual element's offset to compute actual memory access offset for that element.
mask	Memory access mask. Elements with zero corresponding mask's predicate are not accessed, their values in the resulting vector are undefined.

Definition at line 528 of file memory.hpp.

◆ gather() [2/4]

template<typename Tx , int N, typename Toffset >

__ESIMD_API simd<Tx, N> sycl::_V1::ext::intel::esimd::gather	(	const Tx *	p,
		simd< Toffset, N >	offsets,
		simd_mask< N >	mask = `1`
	)

Loads ("gathers") elements from different memory locations and returns a vector of them.

Each memory location is base address plus an offset - a value of the corresponding element in the input offset vector. Access to any element's memory location can be disabled via the input vector of predicates (mask).

Template Parameters

Tx	Element type, must be of size 4 or less.
N	Number of elements to read; can be `1`, `2`, `4`, `8`, `16` or `32`.

Parameters

p	The base address.
offsets	the vector of 32-bit or 64-bit offsets in bytes. For each lane `i`, ((byte*)p + offsets[i]) must be element size aligned.
mask	The access mask, defaults to all 1s.

Returns: A vector of elements read. Elements in masked out lanes are undefined.

Definition at line 125 of file memory.hpp.

References sycl::_V1::ext::intel::esimd::detail::isPowerOf2().

◆ gather() [3/4]

template<typename Tx , int N, typename Toffset , typename RegionTy = region1d_t<Toffset, N, 1>>

__ESIMD_API simd<Tx, N> sycl::_V1::ext::intel::esimd::gather	(	const Tx *	p,
		simd_view< Toffset, RegionTy >	offsets,
		simd_mask< N >	mask = `1`
	)

A variation of gather API with offsets represented as simd_view object.

Template Parameters

Tx	Element type, must be of size 4 or less.
N	Number of elements to read; can be `1`, `2`, `4`, `8`, `16` or `32`.

Parameters

p	The base address.
offsets	the simd_view of 32-bit or 64-bit offsets in bytes. For each lane `i`, ((byte*)p + offsets[i]) must be element size aligned.
mask	The access mask, defaults to all 1s.

Returns: A vector of elements read. Elements in masked out lanes are undefined.

Definition at line 165 of file memory.hpp.

◆ gather() [4/4]

template<typename Tx , int N, typename Toffset >

__ESIMD_API std::enable_if_t<std::is_integral_v<Toffset>, simd<Tx, N> > sycl::_V1::ext::intel::esimd::gather	(	const Tx *	p,
		Toffset	offset,
		simd_mask< N >	mask = `1`
	)

A variation of gather API with offsets represented as scalar.

Template Parameters

Tx	Element type, must be of size 4 or less.
N	Number of elements to read; can be `1`, `2`, `4`, `8`, `16` or `32`.

Parameters

p	The base address.
offset	the scalar 32-bit or 64-bit offset in bytes. ((byte*)p + offset) must be element size aligned.
mask	The access mask, defaults to all 1s.

Returns: A vector of elements read. Elements in masked out lanes are undefined.

Definition at line 185 of file memory.hpp.

◆ gather_rgba() [1/4]

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename AccessorT , int N, typename T = typename AccessorT::value_type>

__ESIMD_API std::enable_if_t< ((N == 8 \|\| N == 16 \|\| N == 32) && sizeof(T) == 4 && !std::is_pointer_v<AccessorT> && !sycl::detail::acc_properties::is_local_accessor_v<AccessorT>), simd<T, N * get_num_channels_enabled(RGBAMask)> > sycl::_V1::ext::intel::esimd::gather_rgba	(	AccessorT	acc,
		simd< uint32_t, N >	offsets,
		uint32_t	global_offset = `0`,
		simd_mask< N >	mask = `1`
	)

Gather and transpose pixels from the given memory locations defined by the base specified by acc, the global offset global_offset and a vector of offsets offsets.

Up to 4 32-bit data elements may be accessed at each address depending on the channel mask RGBAMask. Each pixel's address must be 4-byte aligned. For usage examples, see usm_gather_rgba above, the only difference would be the usage of an accessor instead of a usm pointer.

Template Parameters

RGBAMask	A pixel's channel mask.
AccessorT	The accessor type for the memory to be loaded/gathered. The returned vector elements must match the accessor data type. The loaded elements must be 4 bytes in size.
N	Number of pixels to access (matches the size of the `offsets` vector). Must be 8, 16 or 32.

Parameters

acc	The accessor representing memory address of the access.
offsets	Byte offsets of the pixels relative to the base pointer.
global_offset	Byte offset of the pixels relative to the base pointer.
mask	Memory access mask. Pixels with zero corresponding mask's predicate are not accessed. Their values in the resulting vector are undefined.

Returns: Read data - up to N*4 values of type Tx.

Definition at line 884 of file memory.hpp.

References simd_mask< _Tp, _Abi >::data(), and sycl::_V1::ext::intel::esimd::get_surface_index().

◆ gather_rgba() [2/4]

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset >

__ESIMD_API simd<T, N * get_num_channels_enabled(RGBAMask)> sycl::_V1::ext::intel::esimd::gather_rgba	(	const T *	p,
		simd< Toffset, N >	offsets,
		simd_mask< N >	mask = `1`
	)

Gather and transpose pixels from given memory locations defined by the base pointer p and offsets. Up to 4 32-bit data elements may be accessed at each address depending on the channel mask Mask template parameter. Each pixel's address must be 4 byte aligned. As an example, let's assume we want to read n pixels at address addr, skipping G and B channels. Each channel is a 32-bit float and the pixel data at given address in memory is:

R1 G1 B1 A1 R2 G2 B2 A2 ... Rn Gn Bn An

Then this can be achieved by using

simd<uint32_t, n> byte_offsets(0, 4*4 /* byte size of a single pixel */);

auto x = gather_rgba<float, n, rgba_channel_mask::AR>(addr, byte_offsets);

Returned x will contain 2*n float elements:

R1 R2 ... Rn A1 A2 ... An

Template Parameters

T	Element type of the returned vector. Must be 4 bytes in size.
N	Number of pixels to access (matches the size of the `offsets` vector). Must be 8, 16 or 32.
Mask	A pixel's channel mask.

Parameters

p	The USM base pointer representing memory address of the access.
offsets	vector of byte offsets of the pixels relative to the base pointer.
mask	Memory access mask. Pixels with zero corresponding mask's predicate are not accessed. Their values in the resulting vector are undefined.

Returns: Read data - up to N*4 values of type Tx.

Definition at line 690 of file memory.hpp.

References simd_mask< _Tp, _Abi >::data().

◆ gather_rgba() [3/4]

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset , typename RegionTy = region1d_t<Toffset, N, 1>>

__ESIMD_API simd<T, N * get_num_channels_enabled(RGBAMask)> sycl::_V1::ext::intel::esimd::gather_rgba	(	const T *	p,
		simd_view< Toffset, RegionTy >	offsets,
		simd_mask< N >	mask = `1`
	)

A variation of gather_rgba API with offsets represented as simd_view object.

Template Parameters

T	Element type of the returned vector. Must be 4 bytes in size.
N	Number of pixels to access (matches the size of the `offsets` vector). Must be 8, 16 or 32.
Mask	A pixel's channel mask.

Parameters

p	The USM base pointer representing memory address of the access.
offsets	simd_view of byte offsets of the pixels relative to the base pointer.
mask	Memory access mask. Pixels with zero corresponding mask's predicate are not accessed. Their values in the resulting vector are undefined.

Returns: Read data - up to N*4 values of type Tx.

Definition at line 720 of file memory.hpp.

◆ gather_rgba() [4/4]

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset >

__ESIMD_API std::enable_if_t<std::is_integral_v<Toffset>, simd<T, N * get_num_channels_enabled(RGBAMask)> > sycl::_V1::ext::intel::esimd::gather_rgba	(	const T *	p,
		Toffset	offset,
		simd_mask< N >	mask = `1`
	)

A variation of gather_rgba API with offsets represented as scalar.

Template Parameters

T	Element type of the returned vector. Must be 4 bytes in size.
N	Number of pixels to access (matches the size of the `offsets` vector). Must be 8, 16 or 32.
Mask	A pixel's channel mask.

Parameters

p	The USM base pointer representing memory address of the access.
offset	scalar byte offsets of the pixels relative to the base pointer.
mask	Memory access mask. Pixels with zero corresponding mask's predicate are not accessed. Their values in the resulting vector are undefined.

Returns: Read data - up to N*4 values of type Tx.

Definition at line 744 of file memory.hpp.

◆ get_surface_index()

template<typename AccessorTy >

__ESIMD_API SurfaceIndex sycl::_V1::ext::intel::esimd::get_surface_index ( AccessorTy acc )

Get surface index corresponding to a SYCL accessor.

Parameters

acc	a SYCL buffer or image accessor.

Returns: the index of the corresponding surface (aka "binding table index").

Definition at line 49 of file memory.hpp.

References sycl::_V1::ext::intel::esimd::detail::SLM_BTI.

◆ media_block_load()

template<typename T , int m, int N, typename AccessorTy , unsigned plane = 0>

__ESIMD_API simd<T, m * N> sycl::_V1::ext::intel::esimd::media_block_load	(	AccessorTy	acc,
		unsigned	x,
		unsigned	y
	)

Media block load.

Template Parameters

T	is the element data type.
m	is the height of the 2D block.
N	is the width of the 2D block.
AccessorTy	is type of the SYCL accessor.
plane	is planar surface index.

Parameters

acc	is the SYCL accessor.
x	is X-coordinate of the left upper rectangle corner in BYTES.
y	is Y-coordinate of the left upper rectangle corner in ROWS.

Returns: the linearized 2D block data read from surface.

Definition at line 2071 of file memory.hpp.

References sycl::_V1::ext::intel::esimd::get_surface_index().

◆ media_block_store()

template<typename T , int m, int N, typename AccessorTy , unsigned plane = 0>

__ESIMD_API void sycl::_V1::ext::intel::esimd::media_block_store	(	AccessorTy	acc,
		unsigned	x,
		unsigned	y,
		simd< T, m *N >	vals
	)

Media block store.

Template Parameters

T	is the element data type.
m	is the height of the 2D block.
N	is the width of the 2D block.
is	AccessorTy type of the SYCL accessor.
plane	is planar surface index.

Parameters

acc	is the SYCL accessor.
x	is X-coordinate of the left upper rectangle corner in BYTES.
y	is Y-coordinate of the left upper rectangle corner in ROWS.
vals	is the linearized 2D block data to be written to surface.

Definition at line 2112 of file memory.hpp.

References sycl::_V1::ext::intel::esimd::get_surface_index().

◆ scalar_load()

template<typename T , typename AccessorTy >

__ESIMD_API T sycl::_V1::ext::intel::esimd::scalar_load	(	AccessorTy	acc,
		uint32_t	offset
	)

Load a scalar value from an accessor.

Template Parameters

T	Type of the value.
AccessorTy	Type of the accessor.

Parameters

acc	Accessor to load from.
offset	Offset in bytes.

Returns: The loaded value.

Definition at line 623 of file memory.hpp.

◆ scalar_store()

template<typename T , typename AccessorTy >

__ESIMD_API void sycl::_V1::ext::intel::esimd::scalar_store	(	AccessorTy	acc,
		uint32_t	offset,
		T	val
	)

Store a scalar value into an accessor.

Template Parameters

T	Type of the value.
AccessorTy	Type of the accessor.

Parameters

acc	Accessor to store to.
offset	Offset in bytes.
val	The stored value.

Definition at line 643 of file memory.hpp.

◆ scatter() [1/4]

template<typename T , int N, typename AccessorTy >

__ESIMD_API std::enable_if_t< (sizeof(T) <= 4) && (N == 1 \|\| N == 8 \|\| N == 16 \|\| N == 32) && sycl::detail::acc_properties::is_accessor_v<AccessorTy> && !sycl::detail::acc_properties::is_local_accessor_v<AccessorTy> > sycl::_V1::ext::intel::esimd::scatter	(	AccessorTy	acc,
		simd< uint32_t, N >	offsets,
		simd< T, N >	vals,
		uint32_t	glob_offset = `0`,
		simd_mask< N >	mask = `1`
	)

Accessor-based scatter.

Writes elements of a simd object into an accessor at given offsets. An element can be a 1, 2 or 4-byte value.

Template Parameters

T	Element type; can only be a 1,2,4-byte integer, `sycl::half` or `float`.
N	The number of vector elements. Can be `1`, `8`, `16` or `32`.
AccessorTy	The accessor type.

Parameters

acc	The accessor to scatter to.
offsets	Per-element offsets in bytes.
vals	Values to write.
glob_offset	Offset in bytes added to each individual element's offset to compute actual memory access offset for that element.
mask	Memory access mask. Elements with zero corresponding mask's predicate are not accessed.

Definition at line 581 of file memory.hpp.

◆ scatter() [2/4]

template<typename Tx , int N, typename Toffset >

__ESIMD_API void sycl::_V1::ext::intel::esimd::scatter	(	Tx *	p,
		simd< Toffset, N >	offsets,
		simd< Tx, N >	vals,
		simd_mask< N >	mask = `1`
	)

Writes ("scatters") elements of the input vector to different memory locations.

Each memory location is base address plus an offset - a value of the corresponding element in the input offset vector. Access to any element's memory location can be disabled via the input mask.

Template Parameters

Tx	Element type, must be of size 4 or less.
N	Number of elements to write; can be `1`, `2`, `4`, `8`, `16` or `32`.

Parameters

p	The base address.
offsets	A vector of 32-bit or 64-bit offsets in bytes. For each lane `i`, ((byte*)p + offsets[i]) must be element size aligned.
vals	The vector to scatter.
mask	The access mask, defaults to all 1s.

Definition at line 203 of file memory.hpp.

References sycl::_V1::ext::intel::esimd::detail::isPowerOf2().

◆ scatter() [3/4]

template<typename Tx , int N, typename Toffset , typename RegionTy = region1d_t<Toffset, N, 1>>

__ESIMD_API void sycl::_V1::ext::intel::esimd::scatter	(	Tx *	p,
		simd_view< Toffset, RegionTy >	offsets,
		simd< Tx, N >	vals,
		simd_mask< N >	mask = `1`
	)

A variation of scatter API with offsets represented as simd_view object.

Template Parameters

Tx	Element type, must be of size 4 or less.
N	Number of elements to write; can be `1`, `2`, `4`, `8`, `16` or `32`.

Parameters

p	The base address.
offsets	A simd_view of 32-bit or 64-bit offsets in bytes. For each lane `i`, ((byte*)p + offsets[i]) must be element size aligned.
vals	The vector to scatter.
mask	The access mask, defaults to all 1s.

Definition at line 243 of file memory.hpp.

◆ scatter() [4/4]

template<typename Tx , int N, typename Toffset >

__ESIMD_API std::enable_if_t<std::is_integral_v<Toffset> && N == 1> sycl::_V1::ext::intel::esimd::scatter	(	Tx *	p,
		Toffset	offset,
		simd< Tx, N >	vals,
		simd_mask< N >	mask = `1`
	)

A variation of scatter API with offsets represented as scalar.

Template Parameters

Tx	Element type, must be of size 4 or less.
N	Number of elements to write; can be `1`, `2`, `4`, `8`, `16` or `32`.

Parameters

p	The base address.
offset	the scalar 32-bit or 64-bit offset in bytes. ((byte*)p + offset) must be element size aligned.
vals	The vector to scatter.
mask	The access mask, defaults to all 1s.

Definition at line 261 of file memory.hpp.

◆ scatter_rgba() [1/5]

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename AccessorT , int N, typename T = typename AccessorT::value_type>

__ESIMD_API std::enable_if_t< (N == 8 \|\| N == 16 \|\| N == 32) && sizeof(T) == 4 && !std::is_pointer_v<AccessorT> && !sycl::detail::acc_properties::is_local_accessor_v<AccessorT> > sycl::_V1::ext::intel::esimd::scatter_rgba	(	AccessorT	acc,
		simd< uint32_t, N >	offsets,
		simd< T, N *get_num_channels_enabled(RGBAMask)>	vals,
		uint32_t	global_offset = `0`,
		simd_mask< N >	mask = `1`
	)

Gather data from the memory addressed by accessor acc, offset common for all loaded elements global_offset and per-element offsets offsets, and return it as simd vector.

See usm_gather_rgba for information about the operation semantics and parameter restrictions/interdependencies.

Template Parameters

RGBAMask	Pixel's channel mask.
AccessorT	The accessor type for the memory to be stored/scattered. The returned vector elements must match the accessor data type. The loaded elements must be 4 bytes in size.
N	The number of elements to access.

Parameters

offsets	Byte offsets of each element.
vals	values to be written.
global_offset	Byte offset of the pixels relative to the base pointer.
mask	Operation mask. All-1 by default.

Definition at line 941 of file memory.hpp.

References simd_mask< _Tp, _Abi >::data(), and sycl::_V1::ext::intel::esimd::get_surface_index().

◆ scatter_rgba() [2/5]

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset >

__ESIMD_API void sycl::_V1::ext::intel::esimd::scatter_rgba	(	T *	p,
		simd< Toffset, N >	offsets,
		simd< T, N *get_num_channels_enabled(RGBAMask)>	vals,
		simd_mask< N >	mask = `1`
	)

Transpose and scatter pixels to given memory locations defined by the base pointer p and offsets. Up to 4 32-bit data elements may be accessed at each address depending on the channel mask RGBAMask. Each pixel's address must be 4 byte aligned. This is basically an inverse operation for gather_rgba. Unlike gather_rgba, this function imposes restrictions on possible Mask template argument values. It can only be one of the following: ABGR, BGR, GR, R.

Template Parameters

T	Element type of the returned vector. Must be 4 bytes in size.
N	Number of pixels to access (matches the size of the `offsets` vector). Must be 8, 16 or 32.
RGBAMask	A pixel's channel mask.

Parameters

p	The USM base pointer representing memory address of the access.
vals	values to be written.
offsets	vector of byte offsets of the pixels relative to the base pointer.
mask	Memory access mask. Pixels with zero corresponding mask's predicate are not accessed. Their values in the resulting vector are undefined.

Definition at line 781 of file memory.hpp.

References simd_mask< _Tp, _Abi >::data().

◆ scatter_rgba() [3/5]

template<typename T , int N, rgba_channel_mask RGBAMask>

__ESIMD_API std:: enable_if_t<(N == 8 \|\| N == 16 \|\| N == 32) && sizeof(T) == 4> sycl::_V1::ext::intel::esimd::scatter_rgba	(	T *	p,
		simd< uint32_t, N >	offsets,
		simd< T, N *get_num_channels_enabled(RGBAMask)>	vals,
		simd_mask< N >	mask = `1`
	)

Definition at line 847 of file memory.hpp.

◆ scatter_rgba() [4/5]

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset , typename RegionTy = region1d_t<Toffset, N, 1>>

__ESIMD_API void sycl::_V1::ext::intel::esimd::scatter_rgba	(	T *	p,
		simd_view< Toffset, RegionTy >	offsets,
		simd< T, N *get_num_channels_enabled(RGBAMask)>	vals,
		simd_mask< N >	mask = `1`
	)

A variation of scatter_rgba API with offsets represented as simd_view object.

Template Parameters

T	Element type of the returned vector. Must be 4 bytes in size.
N	Number of pixels to access (matches the size of the `offsets` vector). Must be 8, 16 or 32.
RGBAMask	A pixel's channel mask.

Parameters

p	The USM base pointer representing memory address of the access.
vals	values to be written.
offsets	simd_view of byte offsets of the pixels relative to the base pointer.
mask	Memory access mask. Pixels with zero corresponding mask's predicate are not accessed. Their values in the resulting vector are undefined.

Definition at line 814 of file memory.hpp.

◆ scatter_rgba() [5/5]

template<rgba_channel_mask RGBAMask = rgba_channel_mask::ABGR, typename T , int N, typename Toffset >

__ESIMD_API std::enable_if_t<std::is_integral_v<Toffset> && N == 1> sycl::_V1::ext::intel::esimd::scatter_rgba	(	T *	p,
		Toffset	offset,
		simd< T, N *get_num_channels_enabled(RGBAMask)>	vals,
		simd_mask< N >	mask = `1`
	)

A variation of scatter_rgba API with offsets represented as scalar.

Template Parameters

T	Element type of the returned vector. Must be 4 bytes in size.
N	Number of pixels to access (matches the size of the `offsets` vector). Must be 8, 16 or 32.
RGBAMask	A pixel's channel mask.

Parameters

p	The USM base pointer representing memory address of the access.
vals	values to be written.
offset	scalar byte offset of the pixels relative to the base pointer.
mask	Memory access mask. Pixels with zero corresponding mask's predicate are not accessed. Their values in the resulting vector are undefined.

Definition at line 838 of file memory.hpp.

◆ split_barrier() [1/2]

template<split_barrier_action flag>

__ESIMD_API void sycl::_V1::ext::intel::experimental::esimd::split_barrier ( )

Generic work-group split barrier.

Template Parameters

flag	- split barrier action.

Definition at line 28 of file memory.hpp.

◆ split_barrier() [2/2]

__ESIMD_API void sycl::_V1::ext::intel::experimental::esimd::split_barrier ( split_barrier_action flag )

Definition at line 33 of file memory.hpp.

Modules

Namespaces

Enumerations

Functions

Detailed Description

Enumeration Type Documentation

◆ fence_mask

Function Documentation

◆ barrier()

◆ block_load() [1/3]

◆ block_load() [2/3]

◆ block_load() [3/3]

◆ block_store() [1/3]

◆ block_store() [2/3]

◆ block_store() [3/3]

◆ fence() [1/2]

◆ fence() [2/2]

◆ gather() [1/4]

◆ gather() [2/4]

◆ gather() [3/4]

◆ gather() [4/4]

◆ gather_rgba() [1/4]

◆ gather_rgba() [2/4]

◆ gather_rgba() [3/4]

◆ gather_rgba() [4/4]

◆ get_surface_index()

◆ media_block_load()

◆ media_block_store()

◆ scalar_load()

◆ scalar_store()

◆ scatter() [1/4]

◆ scatter() [2/4]

◆ scatter() [3/4]

◆ scatter() [4/4]

◆ scatter_rgba() [1/5]

◆ scatter_rgba() [2/5]

◆ scatter_rgba() [3/5]

◆ scatter_rgba() [4/5]

◆ scatter_rgba() [5/5]

◆ split_barrier() [1/2]

◆ split_barrier() [2/2]