Skip to content

[SYCL] Fix the type trait 'known_identity' #3227

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
Feb 26, 2021
Merged

[SYCL] Fix the type trait 'known_identity' #3227

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
701377b
[SYCL] Fix the type trait 'known_identity'
v-klochkov Feb 18, 2021
9b355ef
Fix LIT test fails + address reviewer's comment
v-klochkov Feb 23, 2021
0b2702b
Merge remote-tracking branch 'intel_llvm/sycl' into public_vklochkov_…
v-klochkov Feb 23, 2021
389fe9b
Remove guarding macro around has_known_identity and known_identity
v-klochkov Feb 23, 2021
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
176 changes: 91 additions & 85 deletions sycl/include/CL/sycl/ONEAPI/reduction.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,7 @@
#include "CL/sycl/ONEAPI/accessor_property_list.hpp"
#include <CL/sycl/ONEAPI/group_algorithm.hpp>
#include <CL/sycl/accessor.hpp>
#include <CL/sycl/atomic.hpp>
#include <CL/sycl/handler.hpp>
#include <CL/sycl/kernel.hpp>

Expand All @@ -22,121 +23,103 @@ namespace ONEAPI {

namespace detail {

using cl::sycl::detail::bool_constant;
using cl::sycl::detail::enable_if_t;
using cl::sycl::detail::is_sgenfloat;
using cl::sycl::detail::is_sgeninteger;
using cl::sycl::detail::queue_impl;
using cl::sycl::detail::remove_AS;

__SYCL_EXPORT size_t reduGetMaxWGSize(shared_ptr_class<queue_impl> Queue,
size_t LocalMemBytesPerWorkItem);
__SYCL_EXPORT size_t reduComputeWGSize(size_t NWorkItems, size_t MaxWGSize,
size_t &NWorkGroups);

using cl::sycl::detail::bool_constant;
using cl::sycl::detail::enable_if_t;
using cl::sycl::detail::is_geninteger16bit;
using cl::sycl::detail::is_geninteger32bit;
using cl::sycl::detail::is_geninteger64bit;
using cl::sycl::detail::is_geninteger8bit;
using cl::sycl::detail::remove_AS;

template <typename T, class BinaryOperation>
using IsReduPlus = detail::bool_constant<
std::is_same<BinaryOperation, ONEAPI::plus<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::plus<void>>::value>;
using IsReduPlus =
bool_constant<std::is_same<BinaryOperation, ONEAPI::plus<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::plus<void>>::value>;

template <typename T, class BinaryOperation>
using IsReduMultiplies = detail::bool_constant<
std::is_same<BinaryOperation, std::multiplies<T>>::value ||
std::is_same<BinaryOperation, std::multiplies<void>>::value>;
using IsReduMultiplies =
bool_constant<std::is_same<BinaryOperation, std::multiplies<T>>::value ||
std::is_same<BinaryOperation, std::multiplies<void>>::value>;

template <typename T, class BinaryOperation>
using IsReduMinimum = detail::bool_constant<
std::is_same<BinaryOperation, ONEAPI::minimum<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::minimum<void>>::value>;
using IsReduMinimum =
bool_constant<std::is_same<BinaryOperation, ONEAPI::minimum<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::minimum<void>>::value>;

template <typename T, class BinaryOperation>
using IsReduMaximum = detail::bool_constant<
std::is_same<BinaryOperation, ONEAPI::maximum<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::maximum<void>>::value>;
using IsReduMaximum =
bool_constant<std::is_same<BinaryOperation, ONEAPI::maximum<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::maximum<void>>::value>;

template <typename T, class BinaryOperation>
using IsReduBitOR = detail::bool_constant<
std::is_same<BinaryOperation, ONEAPI::bit_or<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::bit_or<void>>::value>;
using IsReduBitOR =
bool_constant<std::is_same<BinaryOperation, ONEAPI::bit_or<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::bit_or<void>>::value>;

template <typename T, class BinaryOperation>
using IsReduBitXOR = detail::bool_constant<
std::is_same<BinaryOperation, ONEAPI::bit_xor<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::bit_xor<void>>::value>;
using IsReduBitXOR =
bool_constant<std::is_same<BinaryOperation, ONEAPI::bit_xor<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::bit_xor<void>>::value>;

template <typename T, class BinaryOperation>
using IsReduBitAND = detail::bool_constant<
std::is_same<BinaryOperation, ONEAPI::bit_and<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::bit_and<void>>::value>;
using IsReduBitAND =
bool_constant<std::is_same<BinaryOperation, ONEAPI::bit_and<T>>::value ||
std::is_same<BinaryOperation, ONEAPI::bit_and<void>>::value>;

template <typename T, class BinaryOperation>
using IsReduOptForFastAtomicFetch =
detail::bool_constant<(is_geninteger32bit<T>::value ||
is_geninteger64bit<T>::value) &&
(IsReduPlus<T, BinaryOperation>::value ||
IsReduMinimum<T, BinaryOperation>::value ||
IsReduMaximum<T, BinaryOperation>::value ||
IsReduBitOR<T, BinaryOperation>::value ||
IsReduBitXOR<T, BinaryOperation>::value ||
IsReduBitAND<T, BinaryOperation>::value)>;
bool_constant<is_sgeninteger<T>::value &&
sycl::detail::IsValidAtomicType<T>::value &&
(IsReduPlus<T, BinaryOperation>::value ||
IsReduMinimum<T, BinaryOperation>::value ||
IsReduMaximum<T, BinaryOperation>::value ||
IsReduBitOR<T, BinaryOperation>::value ||
IsReduBitXOR<T, BinaryOperation>::value ||
IsReduBitAND<T, BinaryOperation>::value)>;

template <typename T, class BinaryOperation>
using IsReduOptForFastReduce = detail::bool_constant<
(is_geninteger32bit<T>::value || is_geninteger64bit<T>::value ||
std::is_same<T, half>::value || std::is_same<T, float>::value ||
std::is_same<T, double>::value) &&
(IsReduPlus<T, BinaryOperation>::value ||
IsReduMinimum<T, BinaryOperation>::value ||
IsReduMaximum<T, BinaryOperation>::value)>;
using IsReduOptForFastReduce =
bool_constant<((is_sgeninteger<T>::value &&
(sizeof(T) == 32 || sizeof(T) == 64)) ||
is_sgenfloat<T>::value) &&
(IsReduPlus<T, BinaryOperation>::value ||
IsReduMinimum<T, BinaryOperation>::value ||
IsReduMaximum<T, BinaryOperation>::value)>;

// Identity = 0
template <typename T, class BinaryOperation>
using IsZeroIdentityOp = bool_constant<
((is_geninteger8bit<T>::value || is_geninteger16bit<T>::value ||
is_geninteger32bit<T>::value || is_geninteger64bit<T>::value) &&
(IsReduPlus<T, BinaryOperation>::value ||
IsReduBitOR<T, BinaryOperation>::value ||
IsReduBitXOR<T, BinaryOperation>::value)) ||
((std::is_same<T, half>::value || std::is_same<T, float>::value ||
std::is_same<T, double>::value) &&
IsReduPlus<T, BinaryOperation>::value)>;
(is_sgeninteger<T>::value && (IsReduPlus<T, BinaryOperation>::value ||
IsReduBitOR<T, BinaryOperation>::value ||
IsReduBitXOR<T, BinaryOperation>::value)) ||
(is_sgenfloat<T>::value && IsReduPlus<T, BinaryOperation>::value)>;

// Identity = 1
template <typename T, class BinaryOperation>
using IsOneIdentityOp = bool_constant<
(is_geninteger8bit<T>::value || is_geninteger16bit<T>::value ||
is_geninteger32bit<T>::value || is_geninteger64bit<T>::value ||
std::is_same<T, half>::value || std::is_same<T, float>::value ||
std::is_same<T, double>::value) &&
IsReduMultiplies<T, BinaryOperation>::value>;
using IsOneIdentityOp =
bool_constant<(is_sgeninteger<T>::value || is_sgenfloat<T>::value) &&
IsReduMultiplies<T, BinaryOperation>::value>;

// Identity = ~0
template <typename T, class BinaryOperation>
using IsOnesIdentityOp = bool_constant<
(is_geninteger8bit<T>::value || is_geninteger16bit<T>::value ||
is_geninteger32bit<T>::value || is_geninteger64bit<T>::value) &&
IsReduBitAND<T, BinaryOperation>::value>;
using IsOnesIdentityOp = bool_constant<is_sgeninteger<T>::value &&
IsReduBitAND<T, BinaryOperation>::value>;

// Identity = <max possible value>
template <typename T, class BinaryOperation>
using IsMinimumIdentityOp = bool_constant<
(is_geninteger8bit<T>::value || is_geninteger16bit<T>::value ||
is_geninteger32bit<T>::value || is_geninteger64bit<T>::value ||
std::is_same<T, half>::value || std::is_same<T, float>::value ||
std::is_same<T, double>::value) &&
IsReduMinimum<T, BinaryOperation>::value>;
using IsMinimumIdentityOp =
bool_constant<(is_sgeninteger<T>::value || is_sgenfloat<T>::value) &&
IsReduMinimum<T, BinaryOperation>::value>;

// Identity = <min possible value>
template <typename T, class BinaryOperation>
using IsMaximumIdentityOp = bool_constant<
(is_geninteger8bit<T>::value || is_geninteger16bit<T>::value ||
is_geninteger32bit<T>::value || is_geninteger64bit<T>::value ||
std::is_same<T, half>::value || std::is_same<T, float>::value ||
std::is_same<T, double>::value) &&
IsReduMaximum<T, BinaryOperation>::value>;
using IsMaximumIdentityOp =
bool_constant<(is_sgeninteger<T>::value || is_sgenfloat<T>::value) &&
IsReduMaximum<T, BinaryOperation>::value>;

template <typename T, class BinaryOperation>
using IsKnownIdentityOp =
Expand Down Expand Up @@ -343,7 +326,7 @@ class reducer<T, BinaryOperation,
/// Atomic ADD operation: *ReduVarPtr += MValue;
template <typename _T = T, class _BinaryOperation = BinaryOperation>
enable_if_t<std::is_same<typename remove_AS<_T>::type, T>::value &&
(is_geninteger32bit<T>::value || is_geninteger64bit<T>::value) &&
IsReduOptForFastAtomicFetch<T, _BinaryOperation>::value &&
IsReduPlus<T, _BinaryOperation>::value>
atomic_combine(_T *ReduVarPtr) const {
atomic<T, access::address_space::global_space>(global_ptr<T>(ReduVarPtr))
Expand All @@ -353,7 +336,7 @@ class reducer<T, BinaryOperation,
/// Atomic BITWISE OR operation: *ReduVarPtr |= MValue;
template <typename _T = T, class _BinaryOperation = BinaryOperation>
enable_if_t<std::is_same<typename remove_AS<_T>::type, T>::value &&
(is_geninteger32bit<T>::value || is_geninteger64bit<T>::value) &&
IsReduOptForFastAtomicFetch<T, _BinaryOperation>::value &&
IsReduBitOR<T, _BinaryOperation>::value>
atomic_combine(_T *ReduVarPtr) const {
atomic<T, access::address_space::global_space>(global_ptr<T>(ReduVarPtr))
Expand All @@ -363,7 +346,7 @@ class reducer<T, BinaryOperation,
/// Atomic BITWISE XOR operation: *ReduVarPtr ^= MValue;
template <typename _T = T, class _BinaryOperation = BinaryOperation>
enable_if_t<std::is_same<typename remove_AS<_T>::type, T>::value &&
(is_geninteger32bit<T>::value || is_geninteger64bit<T>::value) &&
IsReduOptForFastAtomicFetch<T, _BinaryOperation>::value &&
IsReduBitXOR<T, _BinaryOperation>::value>
atomic_combine(_T *ReduVarPtr) const {
atomic<T, access::address_space::global_space>(global_ptr<T>(ReduVarPtr))
Expand All @@ -373,7 +356,7 @@ class reducer<T, BinaryOperation,
/// Atomic BITWISE AND operation: *ReduVarPtr &= MValue;
template <typename _T = T, class _BinaryOperation = BinaryOperation>
enable_if_t<std::is_same<typename remove_AS<_T>::type, T>::value &&
(is_geninteger32bit<T>::value || is_geninteger64bit<T>::value) &&
IsReduOptForFastAtomicFetch<T, _BinaryOperation>::value &&
IsReduBitAND<T, _BinaryOperation>::value>
atomic_combine(_T *ReduVarPtr) const {
atomic<T, access::address_space::global_space>(global_ptr<T>(ReduVarPtr))
Expand All @@ -383,7 +366,7 @@ class reducer<T, BinaryOperation,
/// Atomic MIN operation: *ReduVarPtr = ONEAPI::minimum(*ReduVarPtr, MValue);
template <typename _T = T, class _BinaryOperation = BinaryOperation>
enable_if_t<std::is_same<typename remove_AS<_T>::type, T>::value &&
(is_geninteger32bit<T>::value || is_geninteger64bit<T>::value) &&
IsReduOptForFastAtomicFetch<T, _BinaryOperation>::value &&
IsReduMinimum<T, _BinaryOperation>::value>
atomic_combine(_T *ReduVarPtr) const {
atomic<T, access::address_space::global_space>(global_ptr<T>(ReduVarPtr))
Expand All @@ -393,7 +376,7 @@ class reducer<T, BinaryOperation,
/// Atomic MAX operation: *ReduVarPtr = ONEAPI::maximum(*ReduVarPtr, MValue);
template <typename _T = T, class _BinaryOperation = BinaryOperation>
enable_if_t<std::is_same<typename remove_AS<_T>::type, T>::value &&
(is_geninteger32bit<T>::value || is_geninteger64bit<T>::value) &&
IsReduOptForFastAtomicFetch<T, _BinaryOperation>::value &&
IsReduMaximum<T, _BinaryOperation>::value>
atomic_combine(_T *ReduVarPtr) const {
atomic<T, access::address_space::global_space>(global_ptr<T>(ReduVarPtr))
Expand Down Expand Up @@ -1604,7 +1587,7 @@ reduction(accessor<T, Dims, AccMode, access::target::global_buffer, IsPH> &Acc,
/// The identity value is not passed to this version as it is statically known.
template <typename T, class BinaryOperation, int Dims, access::mode AccMode,
access::placeholder IsPH>
detail::enable_if_t<
std::enable_if_t<
detail::IsKnownIdentityOp<T, BinaryOperation>::value,
detail::reduction_impl<T, BinaryOperation, Dims, false, AccMode, IsPH>>
reduction(accessor<T, Dims, AccMode, access::target::global_buffer, IsPH> &Acc,
Expand Down Expand Up @@ -1632,9 +1615,9 @@ reduction(T *VarPtr, const T &Identity, BinaryOperation BOp) {
/// operation used in the reduction.
/// The identity value is not passed to this version as it is statically known.
template <typename T, class BinaryOperation>
detail::enable_if_t<detail::IsKnownIdentityOp<T, BinaryOperation>::value,
detail::reduction_impl<T, BinaryOperation, 0, true,
access::mode::read_write>>
std::enable_if_t<detail::IsKnownIdentityOp<T, BinaryOperation>::value,
detail::reduction_impl<T, BinaryOperation, 0, true,
access::mode::read_write>>
reduction(T *VarPtr, BinaryOperation) {
return detail::reduction_impl<T, BinaryOperation, 0, true,
access::mode::read_write>(VarPtr);
Expand All @@ -1659,7 +1642,30 @@ template <typename BinaryOperation, typename AccumulatorT>
inline constexpr AccumulatorT known_identity_v =
known_identity<BinaryOperation, AccumulatorT>::value;
#endif

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think it is worth to mark the code that we want to remove when allowed to break API.

} // namespace ONEAPI

// Currently, the type traits defined below correspond to SYCL 1.2.1 ONEAPI
// reduction extension. That may be changed later when SYCL 2020 reductions
// are implemented.
template <typename BinaryOperation, typename AccumulatorT>
struct has_known_identity
: ONEAPI::has_known_identity<BinaryOperation, AccumulatorT> {};

#if __cplusplus >= 201703L
template <typename BinaryOperation, typename AccumulatorT>
inline constexpr bool has_known_identity_v =
has_known_identity<BinaryOperation, AccumulatorT>::value;
#endif

template <typename BinaryOperation, typename AccumulatorT>
struct known_identity : ONEAPI::known_identity<BinaryOperation, AccumulatorT> {
};

#if __cplusplus >= 201703L
template <typename BinaryOperation, typename AccumulatorT>
inline constexpr AccumulatorT known_identity_v =
known_identity<BinaryOperation, AccumulatorT>::value;
#endif

} // namespace sycl
} // __SYCL_INLINE_NAMESPACE(cl)
24 changes: 21 additions & 3 deletions sycl/test/basic_tests/reduction_ctor.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,10 @@

using namespace cl::sycl;

bool toBool(bool V) { return V; }
bool toBool(vec<int, 2> V) { return V.x() && V.y(); }
bool toBool(vec<int, 4> V) { return V.x() && V.y() && V.z() && V.w(); }

template <typename T, typename Reduction>
void test_reducer(Reduction &Redu, T A, T B) {
typename Reduction::reducer_type Reducer;
Expand All @@ -29,7 +33,7 @@ void test_reducer(Reduction &Redu, T Identity, BinaryOperation BOp, T A, T B) {
Reducer.combine(B);

T ExpectedValue = BOp(A, B);
assert(ExpectedValue == Reducer.MValue &&
assert(toBool(ExpectedValue == Reducer.MValue) &&
"Wrong result of binary operation.");
}

Expand All @@ -40,14 +44,17 @@ template <typename SpecializationKernelName, typename T, int Dim,
void testKnown(T Identity, BinaryOperation BOp, T A, T B) {
buffer<T, 1> ReduBuf(1);

static_assert(has_known_identity<BinaryOperation, T>::value);
queue Q;
Q.submit([&](handler &CGH) {
// Reduction needs a global_buffer accessor as a parameter.
// This accessor is not really used in this test.
accessor<T, Dim, access::mode::discard_write, access::target::global_buffer>
ReduAcc(ReduBuf, CGH);
auto Redu = ONEAPI::reduction(ReduAcc, BOp);
assert(Redu.getIdentity() == Identity && "Failed getIdentity() check().");
assert(toBool(Redu.getIdentity() == Identity) &&
toBool(known_identity<BinaryOperation, T>::value == Identity) &&
"Failed getIdentity() check().");
test_reducer(Redu, A, B);
test_reducer(Redu, Identity, BOp, A, B);

Expand All @@ -67,7 +74,8 @@ void testUnknown(T Identity, BinaryOperation BOp, T A, T B) {
accessor<T, Dim, access::mode::discard_write, access::target::global_buffer>
ReduAcc(ReduBuf, CGH);
auto Redu = ONEAPI::reduction(ReduAcc, Identity, BOp);
assert(Redu.getIdentity() == Identity && "Failed getIdentity() check().");
bool IsCorrectVal = toBool(Redu.getIdentity() == Identity);
assert(IsCorrectVal && "Failed getIdentity() check().");
test_reducer(Redu, Identity, BOp, A, B);

// Command group must have at least one task in it. Use an empty one.
Expand Down Expand Up @@ -124,6 +132,16 @@ int main() {
testUnknown<class KernelName_zhF, int, 0>(
0, [](auto a, auto b) { return a | b; }, 1, 8);

int2 IdentityI2 = {0, 0};
int2 AI2 = {1, 2};
int2 BI2 = {7, 13};
testUnknown<class KNI2, int2, 0>(IdentityI2, ONEAPI::plus<int2>(), AI2, BI2);

float4 IdentityF4 = {0, 0, 0, 0};
float4 AF4 = {1, 2, -1, -34};
float4 BF4 = {7, 13, 0, 35};
testUnknown<class KNF4, float4, 0>(IdentityF4, ONEAPI::plus<>(), AF4, BF4);

std::cout << "Test passed\n";
return 0;
}
Loading