dylanyunlong/project_6_89d52222
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity
Files
7d74ad4e8e0a8b0b72f7c3327413dc9301015978
project_6_89d52222/cccl_upstream/thrust/thrust/detail/unique.inl
EngineX CI 56fd68e7dd [INFRA] Import NVIDIA/CCCL upstream as optimization reference library 
CCCL (CUDA C++ Core Libraries) provides:
- CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk)
- Thrust: high-level parallel algorithms (transform_reduce, sort, scan)
- libcudacxx: CUDA C++ standard library (atomics, barriers, memory)
- cudax: experimental features (memory resources, allocators)
- Tuning policies: per-SM hardware-specific algorithm parameters

Competition optimization vectors mapped to CCCL:
- Output TPS (83% weight): warp_reduce, block_reduce, device_topk
- Input TPS (14% weight): device_scan, block_load, prefetch
- Cache TPS (3% weight): prefix caching strategy patterns
- Memory (0.9 util): pooled/cached/buddy allocators

Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only)
License: Apache-2.0
2026-07-30 09:35:51 +00:00

379 lines
14 KiB
C++
Raw Blame History

// SPDX-FileCopyrightText: Copyright (c) 2008-2013, NVIDIA Corporation. All rights reserved.
// SPDX-License-Identifier: Apache-2.0
#pragma once
#include <thrust/detail/config.h>
#if defined(_CCCL_IMPLICIT_SYSTEM_HEADER_GCC)
# pragma GCC system_header
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_CLANG)
# pragma clang system_header
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
# pragma system_header
#endif // no system header
#include <thrust/detail/temporary_array.h>
#include <thrust/iterator/iterator_traits.h>
#include <thrust/system/detail/generic/select_system.h>
#include <thrust/unique.h>
// Include all active backend system implementations (generic, sequential, host and device)
#include <thrust/system/detail/generic/unique.h>
#include <thrust/system/detail/generic/unique_by_key.h>
#include <thrust/system/detail/sequential/unique.h>
#include <thrust/system/detail/sequential/unique_by_key.h>
#include __THRUST_HOST_SYSTEM_ALGORITH_DETAIL_HEADER_INCLUDE(unique.h)
#include __THRUST_DEVICE_SYSTEM_ALGORITH_DETAIL_HEADER_INCLUDE(unique.h)
#include __THRUST_HOST_SYSTEM_ALGORITH_DETAIL_HEADER_INCLUDE(unique_by_key.h)
#include __THRUST_DEVICE_SYSTEM_ALGORITH_DETAIL_HEADER_INCLUDE(unique_by_key.h)
// Some build systems need a hint to know which files we could include
#if 0
# include <thrust/system/cpp/detail/unique.h>
# include <thrust/system/cpp/detail/unique_by_key.h>
# include <thrust/system/cuda/detail/unique.h>
# include <thrust/system/cuda/detail/unique_by_key.h>
# include <thrust/system/omp/detail/unique.h>
# include <thrust/system/omp/detail/unique_by_key.h>
# include <thrust/system/tbb/detail/unique.h>
# include <thrust/system/tbb/detail/unique_by_key.h>
#endif
THRUST_NAMESPACE_BEGIN
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy, typename ForwardIterator>
_CCCL_HOST_DEVICE ForwardIterator
unique(const thrust::detail::execution_policy_base<DerivedPolicy>& exec, ForwardIterator first, ForwardIterator last)
{
_CCCL_NVTX_RANGE_SCOPE("unique");
using thrust::system::detail::generic::unique;
return unique(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
} // end unique()
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
_CCCL_HOST_DEVICE ForwardIterator unique(
const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
ForwardIterator first,
ForwardIterator last,
BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique");
using thrust::system::detail::generic::unique;
return unique(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, binary_pred);
} // end unique()
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy, typename InputIterator, typename OutputIterator>
_CCCL_HOST_DEVICE OutputIterator unique_copy(
const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
InputIterator first,
InputIterator last,
OutputIterator output)
{
_CCCL_NVTX_RANGE_SCOPE("unique_copy");
using thrust::system::detail::generic::unique_copy;
return unique_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, output);
} // end unique_copy()
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy, typename InputIterator, typename OutputIterator, typename BinaryPredicate>
_CCCL_HOST_DEVICE OutputIterator unique_copy(
const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
InputIterator first,
InputIterator last,
OutputIterator output,
BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique_copy");
using thrust::system::detail::generic::unique_copy;
return unique_copy(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, output, binary_pred);
} // end unique_copy()
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy, typename ForwardIterator1, typename ForwardIterator2>
_CCCL_HOST_DEVICE ::cuda::std::pair<ForwardIterator1, ForwardIterator2> unique_by_key(
const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
ForwardIterator1 keys_first,
ForwardIterator1 keys_last,
ForwardIterator2 values_first)
{
_CCCL_NVTX_RANGE_SCOPE("unique_by_key");
using thrust::system::detail::generic::unique_by_key;
return unique_by_key(
thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first);
} // end unique_by_key()
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy, typename ForwardIterator1, typename ForwardIterator2, typename BinaryPredicate>
_CCCL_HOST_DEVICE ::cuda::std::pair<ForwardIterator1, ForwardIterator2> unique_by_key(
const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
ForwardIterator1 keys_first,
ForwardIterator1 keys_last,
ForwardIterator2 values_first,
BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique_by_key");
using thrust::system::detail::generic::unique_by_key;
return unique_by_key(
thrust::detail::derived_cast(thrust::detail::strip_const(exec)), keys_first, keys_last, values_first, binary_pred);
} // end unique_by_key()
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy,
typename InputIterator1,
typename InputIterator2,
typename OutputIterator1,
typename OutputIterator2>
_CCCL_HOST_DEVICE ::cuda::std::pair<OutputIterator1, OutputIterator2> unique_by_key_copy(
const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
InputIterator1 keys_first,
InputIterator1 keys_last,
InputIterator2 values_first,
OutputIterator1 keys_output,
OutputIterator2 values_output)
{
_CCCL_NVTX_RANGE_SCOPE("unique_by_key_copy");
using thrust::system::detail::generic::unique_by_key_copy;
return unique_by_key_copy(
thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
keys_first,
keys_last,
values_first,
keys_output,
values_output);
} // end unique_by_key_copy()
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy,
typename InputIterator1,
typename InputIterator2,
typename OutputIterator1,
typename OutputIterator2,
typename BinaryPredicate>
_CCCL_HOST_DEVICE ::cuda::std::pair<OutputIterator1, OutputIterator2> unique_by_key_copy(
const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
InputIterator1 keys_first,
InputIterator1 keys_last,
InputIterator2 values_first,
OutputIterator1 keys_output,
OutputIterator2 values_output,
BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique_by_key_copy");
using thrust::system::detail::generic::unique_by_key_copy;
return unique_by_key_copy(
thrust::detail::derived_cast(thrust::detail::strip_const(exec)),
keys_first,
keys_last,
values_first,
keys_output,
values_output,
binary_pred);
} // end unique_by_key_copy()
template <typename ForwardIterator>
ForwardIterator unique(ForwardIterator first, ForwardIterator last)
{
_CCCL_NVTX_RANGE_SCOPE("unique");
using thrust::system::detail::generic::select_system;
using System = typename thrust::iterator_system<ForwardIterator>::type;
System system;
return thrust::unique(select_system(system), first, last);
} // end unique()
template <typename ForwardIterator, typename BinaryPredicate>
ForwardIterator unique(ForwardIterator first, ForwardIterator last, BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique");
using thrust::system::detail::generic::select_system;
using System = typename thrust::iterator_system<ForwardIterator>::type;
System system;
return thrust::unique(select_system(system), first, last, binary_pred);
} // end unique()
template <typename InputIterator, typename OutputIterator>
OutputIterator unique_copy(InputIterator first, InputIterator last, OutputIterator output)
{
_CCCL_NVTX_RANGE_SCOPE("unique_copy");
using thrust::system::detail::generic::select_system;
using System1 = typename thrust::iterator_system<InputIterator>::type;
using System2 = typename thrust::iterator_system<OutputIterator>::type;
System1 system1;
System2 system2;
return thrust::unique_copy(select_system(system1, system2), first, last, output);
} // end unique_copy()
template <typename InputIterator, typename OutputIterator, typename BinaryPredicate>
OutputIterator unique_copy(InputIterator first, InputIterator last, OutputIterator output, BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique_copy");
using thrust::system::detail::generic::select_system;
using System1 = typename thrust::iterator_system<InputIterator>::type;
using System2 = typename thrust::iterator_system<OutputIterator>::type;
System1 system1;
System2 system2;
return thrust::unique_copy(select_system(system1, system2), first, last, output, binary_pred);
} // end unique_copy()
template <typename ForwardIterator1, typename ForwardIterator2>
::cuda::std::pair<ForwardIterator1, ForwardIterator2>
unique_by_key(ForwardIterator1 keys_first, ForwardIterator1 keys_last, ForwardIterator2 values_first)
{
_CCCL_NVTX_RANGE_SCOPE("unique_by_key");
using thrust::system::detail::generic::select_system;
using System1 = typename thrust::iterator_system<ForwardIterator1>::type;
using System2 = typename thrust::iterator_system<ForwardIterator2>::type;
System1 system1;
System2 system2;
return thrust::unique_by_key(select_system(system1, system2), keys_first, keys_last, values_first);
} // end unique_by_key()
template <typename ForwardIterator1, typename ForwardIterator2, typename BinaryPredicate>
::cuda::std::pair<ForwardIterator1, ForwardIterator2> unique_by_key(
ForwardIterator1 keys_first, ForwardIterator1 keys_last, ForwardIterator2 values_first, BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique_by_key");
using thrust::system::detail::generic::select_system;
using System1 = typename thrust::iterator_system<ForwardIterator1>::type;
using System2 = typename thrust::iterator_system<ForwardIterator2>::type;
System1 system1;
System2 system2;
return thrust::unique_by_key(select_system(system1, system2), keys_first, keys_last, values_first, binary_pred);
} // end unique_by_key()
template <typename InputIterator1, typename InputIterator2, typename OutputIterator1, typename OutputIterator2>
::cuda::std::pair<OutputIterator1, OutputIterator2> unique_by_key_copy(
InputIterator1 keys_first,
InputIterator1 keys_last,
InputIterator2 values_first,
OutputIterator1 keys_output,
OutputIterator2 values_output)
{
_CCCL_NVTX_RANGE_SCOPE("unique_by_key_copy");
using thrust::system::detail::generic::select_system;
using System1 = typename thrust::iterator_system<InputIterator1>::type;
using System2 = typename thrust::iterator_system<InputIterator2>::type;
using System3 = typename thrust::iterator_system<OutputIterator1>::type;
using System4 = typename thrust::iterator_system<OutputIterator2>::type;
System1 system1;
System2 system2;
System3 system3;
System4 system4;
return thrust::unique_by_key_copy(
select_system(system1, system2, system3, system4), keys_first, keys_last, values_first, keys_output, values_output);
} // end unique_by_key_copy()
template <typename InputIterator1,
typename InputIterator2,
typename OutputIterator1,
typename OutputIterator2,
typename BinaryPredicate>
::cuda::std::pair<OutputIterator1, OutputIterator2> unique_by_key_copy(
InputIterator1 keys_first,
InputIterator1 keys_last,
InputIterator2 values_first,
OutputIterator1 keys_output,
OutputIterator2 values_output,
BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique_by_key_copy");
using thrust::system::detail::generic::select_system;
using System1 = typename thrust::iterator_system<InputIterator1>::type;
using System2 = typename thrust::iterator_system<InputIterator2>::type;
using System3 = typename thrust::iterator_system<OutputIterator1>::type;
using System4 = typename thrust::iterator_system<OutputIterator2>::type;
System1 system1;
System2 system2;
System3 system3;
System4 system4;
return thrust::unique_by_key_copy(
select_system(system1, system2, system3, system4),
keys_first,
keys_last,
values_first,
keys_output,
values_output,
binary_pred);
} // end unique_by_key_copy()
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy, typename ForwardIterator, typename BinaryPredicate>
_CCCL_HOST_DEVICE thrust::detail::it_difference_t<ForwardIterator> unique_count(
const thrust::detail::execution_policy_base<DerivedPolicy>& exec,
ForwardIterator first,
ForwardIterator last,
BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique_count");
using thrust::system::detail::generic::unique_count;
return unique_count(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last, binary_pred);
} // end unique_count()
_CCCL_EXEC_CHECK_DISABLE
template <typename DerivedPolicy, typename ForwardIterator>
_CCCL_HOST_DEVICE thrust::detail::it_difference_t<ForwardIterator> unique_count(
const thrust::detail::execution_policy_base<DerivedPolicy>& exec, ForwardIterator first, ForwardIterator last)
{
_CCCL_NVTX_RANGE_SCOPE("unique_count");
using thrust::system::detail::generic::unique_count;
return unique_count(thrust::detail::derived_cast(thrust::detail::strip_const(exec)), first, last);
} // end unique_count()
_CCCL_EXEC_CHECK_DISABLE
template <typename ForwardIterator, typename BinaryPredicate>
_CCCL_HOST_DEVICE thrust::detail::it_difference_t<ForwardIterator>
unique_count(ForwardIterator first, ForwardIterator last, BinaryPredicate binary_pred)
{
_CCCL_NVTX_RANGE_SCOPE("unique_count");
using thrust::system::detail::generic::select_system;
using System = typename thrust::iterator_system<ForwardIterator>::type;
System system;
return thrust::unique_count(select_system(system), first, last, binary_pred);
} // end unique_count()
_CCCL_EXEC_CHECK_DISABLE
template <typename ForwardIterator>
_CCCL_HOST_DEVICE thrust::detail::it_difference_t<ForwardIterator>
unique_count(ForwardIterator first, ForwardIterator last)
{
_CCCL_NVTX_RANGE_SCOPE("unique_count");
using thrust::system::detail::generic::select_system;
using System = typename thrust::iterator_system<ForwardIterator>::type;
System system;
return thrust::unique_count(select_system(system), first, last);
} // end unique_count()
THRUST_NAMESPACE_END
Reference in New Issue View Git Blame Copy Permalink