Skip to content

[ESIMD][NFC] Add function extracting L1/L2 hint for esimd/genx intrin #13275

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 4 commits into from
Apr 4, 2024
Merged

[ESIMD][NFC] Add function extracting L1/L2 hint for esimd/genx intrin #13275

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
3352567
[ESIMD][NFC] Add function extracting L1/L2 hint for esimd/genx intrin
v-klochkov Apr 3, 2024
dbd604d
Rename are_both with are_all, and make the 'Val' parameter a template…
v-klochkov Apr 4, 2024
0e58e0d
Fix the obsolete comment section
v-klochkov Apr 4, 2024
71d4b93
Update sycl/include/sycl/ext/intel/esimd/memory.hpp
v-klochkov Apr 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 9 additions & 9 deletions sycl/include/sycl/ext/intel/esimd/common.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -545,7 +545,8 @@ template <cache_hint Hint> class cache_hint_wrap {
}
};

constexpr bool are_both(cache_hint First, cache_hint Second, cache_hint Val) {
template <cache_hint Val>
constexpr bool are_all(cache_hint First, cache_hint Second) {
return First == Val && Second == Val;
}

Expand All @@ -559,9 +560,8 @@ template <typename PropertyListT> constexpr bool has_cache_hints() {
return L1H != cache_hint::none || L2H != cache_hint::none;
}

// Currently, this is just a wrapper around 'check_cache_hint' function.
// It accepts the compile-time properties that may include cache-hints
// to be verified.
// Verifies cache-hint properties from 'PropertyListT`. The parameter 'Action'
// specifies the usage context.
template <cache_action Action, typename PropertyListT>
void check_cache_hints() {
constexpr auto L1H =
Expand All @@ -576,28 +576,28 @@ void check_cache_hints() {
cache_hint::streaming>() &&
L2H.template is_one_of<cache_hint::cached,
cache_hint::uncached>() &&
!are_both(L1H, L2H, cache_hint::uncached),
!are_all<cache_hint::uncached>(L1H, L2H),
"unsupported cache hint");
} else if constexpr (Action == cache_action::load) {
static_assert(
are_both(L1H, L2H, cache_hint::none) ||
are_all<cache_hint::none>(L1H, L2H) ||
(L1H.template is_one_of<cache_hint::uncached, cache_hint::cached,
cache_hint::streaming>() &&
L2H.template is_one_of<cache_hint::uncached,
cache_hint::cached>()) ||
(L1H == cache_hint::read_invalidate && L2H == cache_hint::cached),
"unsupported cache hint");
} else if constexpr (Action == cache_action::store) {
static_assert(are_both(L1H, L2H, cache_hint::none) ||
are_both(L1H, L2H, cache_hint::write_back) ||
static_assert(are_all<cache_hint::none>(L1H, L2H) ||
are_all<cache_hint::write_back>(L1H, L2H) ||
(L1H.template is_one_of<cache_hint::uncached,
cache_hint::write_through,
cache_hint::streaming>() &&
L2H.template is_one_of<cache_hint::uncached,
cache_hint::write_back>()),
"unsupported cache hint");
} else if constexpr (Action == cache_action::atomic) {
static_assert(are_both(L1H, L2H, cache_hint::none) ||
static_assert(are_all<cache_hint::none>(L1H, L2H) ||
(L1H == cache_hint::uncached &&
L2H.template is_one_of<cache_hint::uncached,
cache_hint::write_back>()),
Expand Down
146 changes: 58 additions & 88 deletions sycl/include/sycl/ext/intel/esimd/memory.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,6 +95,20 @@ ESIMD_INLINE simd<T, N> lsc_format_ret(simd<T1, N> Vals) {
}
}

/// Extracts a cache hint with the given 'Level' to pass it to
/// ESIMD/GENX intrinsics. If `PropertyListT` does not have the requested
/// cache-hint, then 'cache_hint::none' is returned.
template <typename PropertyListT, cache_level Level>
constexpr cache_hint getCacheHintForIntrin() {
static_assert(Level == cache_level::L1 || Level == cache_level::L2,
"ESIMD/GENX intrinsics accept only L1/L2 cache hints");
if constexpr (Level == cache_level::L1) {
return getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
} else {
return getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
}
}

/// USM pointer gather.
/// Supported platforms: DG2, PVC
/// VISA instruction: lsc_load.ugm
Expand Down Expand Up @@ -123,10 +137,8 @@ __ESIMD_API simd<T, N * NElts> gather_impl(const T *p, simd<OffsetT, N> offsets,
check_lsc_vector_size<NElts>();
check_lsc_data_size<T, DS>();
check_cache_hints<cache_action::load, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down Expand Up @@ -167,10 +179,8 @@ __ESIMD_API void scatter_impl(T *p, simd<Toffset, N> offsets,
check_lsc_vector_size<NElts>();
check_lsc_data_size<T, DS>();
check_cache_hints<cache_action::store, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down Expand Up @@ -904,10 +914,8 @@ block_load_impl(const T *p, simd_mask<1> pred, simd<T, NElts> pass_thru) {
using LoadElemT = __ESIMD_DNS::__raw_t<
std::conditional_t<SmallIntFactor == 1, T,
std::conditional_t<Use64BitData, uint64_t, uint32_t>>>;
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();

constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
Expand Down Expand Up @@ -1005,10 +1013,8 @@ __ESIMD_API
using LoadElemT = __ESIMD_DNS::__raw_t<
std::conditional_t<SmallIntFactor == 1, T,
std::conditional_t<Use64BitData, uint64_t, uint32_t>>>;
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size ActualDS =
Expand Down Expand Up @@ -1105,10 +1111,8 @@ __ESIMD_API
using LoadElemT = __ESIMD_DNS::__raw_t<
std::conditional_t<SmallIntFactor == 1, T,
std::conditional_t<Use64BitData, uint64_t, uint32_t>>>;
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size ActualDS =
Expand Down Expand Up @@ -1165,10 +1169,8 @@ block_store_impl(T *p, simd<T, NElts> vals, simd_mask<1> pred) {
using StoreType = __ESIMD_DNS::__raw_t<
std::conditional_t<SmallIntFactor == 1, T,
std::conditional_t<Use64BitData, uint64_t, uint32_t>>>;
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size ActualDS =
Expand Down Expand Up @@ -1230,10 +1232,8 @@ __ESIMD_API
using StoreElemT = __ESIMD_DNS::__raw_t<
std::conditional_t<SmallIntFactor == 1, T,
std::conditional_t<Use64BitData, uint64_t, uint32_t>>>;
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size ActualDS =
Expand Down Expand Up @@ -2586,10 +2586,8 @@ scatter_impl(AccessorTy acc, simd<OffsetT, N> offsets, simd<T, N * NElts> vals,
check_lsc_vector_size<NElts>();
check_lsc_data_size<T, DS>();
check_cache_hints<cache_action::store, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down Expand Up @@ -2686,10 +2684,8 @@ gather_impl(AccessorT acc, simd<OffsetT, N / VS> byte_offsets,
constexpr lsc_vector_size LSCVS = to_lsc_vector_size<VS>();
constexpr auto Transposed = lsc_data_order::nontranspose;
using MsgT = typename lsc_expand_type<T>::type;
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
auto SI = get_surface_index(acc);
simd<uint32_t, N / VS> ByteOffsets32 = convert<uint32_t>(byte_offsets);
simd<MsgT, N> PassThruExpanded = lsc_format_input<MsgT>(pass_thru);
Expand Down Expand Up @@ -2793,10 +2789,8 @@ __ESIMD_API void prefetch_impl(const T *p, simd<Toffset, N> byte_offsets,
check_lsc_vector_size<NElts>();
check_lsc_data_size<T, DS>();
check_cache_hints<cache_action::prefetch, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand All @@ -2817,10 +2811,8 @@ prefetch_impl(const T *p, Toffset offset, simd_mask<1> pred) {
check_lsc_vector_size<NElts>();
check_lsc_data_size<T, DS>();
check_cache_hints<cache_action::prefetch, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = finalize_data_size<T, DS>();
Expand Down Expand Up @@ -2872,10 +2864,8 @@ prefetch_impl(AccessorTy acc, simd<OffsetT, N> byte_offsets,
check_lsc_vector_size<NElts>();
check_lsc_data_size<T, DS>();
check_cache_hints<cache_action::prefetch, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down Expand Up @@ -2919,10 +2909,8 @@ prefetch_impl(AccessorTy acc, OffsetT byte_offset, simd_mask<1> pred) {
check_lsc_vector_size<NElts>();
check_lsc_data_size<T, DS>();
check_cache_hints<cache_action::prefetch, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = finalize_data_size<T, DS>();
Expand Down Expand Up @@ -3058,10 +3046,8 @@ __ESIMD_API simd<T, N> load_2d_impl(const T *Ptr, unsigned SurfaceWidth,
unsigned SurfacePitch, int X, int Y) {

check_cache_hints<cache_action::load, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
using RawT = __raw_t<T>;
check_lsc_block_2d_restrictions<RawT, BlockWidth, BlockHeight, NBlocks,
Transposed, Transformed, block_2d_op::load>();
Expand Down Expand Up @@ -3172,10 +3158,8 @@ __ESIMD_API void prefetch_2d_impl(const T *Ptr, unsigned SurfaceWidth,
check_cache_hints<cache_action::prefetch, PropertyListT>();
check_lsc_block_2d_restrictions<RawT, BlockWidth, BlockHeight, NBlocks, false,
false, block_2d_op::prefetch>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr lsc_data_size DS =
finalize_data_size<RawT, lsc_data_size::default_size>();
uintptr_t Addr = reinterpret_cast<uintptr_t>(Ptr);
Expand Down Expand Up @@ -3220,10 +3204,8 @@ __ESIMD_API void store_2d_impl(T *Ptr, unsigned SurfaceWidth,
using RawT = __raw_t<T>;
__ESIMD_DNS::check_cache_hints<__ESIMD_DNS::cache_action::store,
PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
check_lsc_block_2d_restrictions<RawT, BlockWidth, BlockHeight, 1, false,
false, block_2d_op::store>();
constexpr lsc_data_size DS =
Expand Down Expand Up @@ -6164,10 +6146,8 @@ atomic_update_impl(T *p, simd<Toffset, N> offsets, simd_mask<N> pred) {
check_atomic<Op, T, N, 0, /*IsLSC*/ true>();
check_lsc_data_size<T, DS>();
check_cache_hints<cache_action::atomic, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down Expand Up @@ -6208,10 +6188,8 @@ atomic_update_impl(T *p, simd<Toffset, N> offsets, simd<T, N> src0,
check_lsc_data_size<T, DS>();
check_atomic<Op, T, N, 1, /*IsLSC*/ true>();
check_cache_hints<cache_action::atomic, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down Expand Up @@ -6254,10 +6232,8 @@ atomic_update_impl(T *p, simd<Toffset, N> offsets, simd<T, N> src0,
check_lsc_data_size<T, DS>();
check_atomic<Op, T, N, 2, /*IsLSC*/ true>();
check_cache_hints<cache_action::atomic, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down Expand Up @@ -6310,10 +6286,8 @@ __ESIMD_API
check_lsc_data_size<T, DS>();
check_atomic<Op, T, N, 0, /*IsLSC*/ true>();
check_cache_hints<cache_action::atomic, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down Expand Up @@ -6365,10 +6339,8 @@ __ESIMD_API
check_lsc_data_size<T, DS>();
check_atomic<Op, T, N, 1, /*IsLSC*/ true>();
check_cache_hints<cache_action::atomic, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down Expand Up @@ -6422,10 +6394,8 @@ __ESIMD_API
check_lsc_data_size<T, DS>();
check_atomic<Op, T, N, 2, /*IsLSC*/ true>();
check_cache_hints<cache_action::atomic, PropertyListT>();
constexpr cache_hint L1H =
getPropertyValue<PropertyListT, cache_hint_L1_key>(cache_hint::none);
constexpr cache_hint L2H =
getPropertyValue<PropertyListT, cache_hint_L2_key>(cache_hint::none);
constexpr auto L1H = getCacheHintForIntrin<PropertyListT, cache_level::L1>();
constexpr auto L2H = getCacheHintForIntrin<PropertyListT, cache_level::L2>();
constexpr uint16_t AddressScale = 1;
constexpr int ImmOffset = 0;
constexpr lsc_data_size EDS = expand_data_size(finalize_data_size<T, DS>());
Expand Down