cub/device/device_select.cuh

File members: cub/device/device_select.cuh

/******************************************************************************
 * Copyright (c) 2011, Duane Merrill.  All rights reserved.
 * Copyright (c) 2011-2022, NVIDIA CORPORATION.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions are met:
 *     * Redistributions of source code must retain the above copyright
 *       notice, this list of conditions and the following disclaimer.
 *     * Redistributions in binary form must reproduce the above copyright
 *       notice, this list of conditions and the following disclaimer in the
 *       documentation and/or other materials provided with the distribution.
 *     * Neither the name of the NVIDIA CORPORATION nor the
 *       names of its contributors may be used to endorse or promote products
 *       derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE FOR ANY
 * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 ******************************************************************************/

#pragma once

#include <cub/config.cuh>

#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 <cub/detail/choose_offset.cuh>
#include <cub/detail/nvtx.cuh>
#include <cub/device/dispatch/dispatch_select_if.cuh>
#include <cub/device/dispatch/dispatch_unique_by_key.cuh>
#include <cub/util_deprecated.cuh>

#include <cuda/std/type_traits>

#include <iterator>

#include <stdio.h>

CUB_NAMESPACE_BEGIN

struct DeviceSelect
{
  template <typename InputIteratorT, typename FlagIterator, typename OutputIteratorT, typename NumSelectedIteratorT>
  CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t Flagged(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    InputIteratorT d_in,
    FlagIterator d_flags,
    OutputIteratorT d_out,
    NumSelectedIteratorT d_num_selected_out,
    ::cuda::std::int64_t num_items,
    cudaStream_t stream = 0)
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE_IF(d_temp_storage, "cub::DeviceSelect::Flagged");

    using OffsetT    = ::cuda::std::int64_t; // Signed integer type for global offsets
    using SelectOp   = NullType; // Selection op (not used)
    using EqualityOp = NullType; // Equality operator (not used)

    return DispatchSelectIf<
      InputIteratorT,
      FlagIterator,
      OutputIteratorT,
      NumSelectedIteratorT,
      SelectOp,
      EqualityOp,
      OffsetT,
      false>::Dispatch(d_temp_storage,
                       temp_storage_bytes,
                       d_in,
                       d_flags,
                       d_out,
                       d_num_selected_out,
                       SelectOp(),
                       EqualityOp(),
                       num_items,
                       stream);
  }

#ifndef DOXYGEN_SHOULD_SKIP_THIS // Do not document
  template <typename InputIteratorT, typename FlagIterator, typename OutputIteratorT, typename NumSelectedIteratorT>
  CUB_DETAIL_RUNTIME_DEBUG_SYNC_IS_NOT_SUPPORTED CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t Flagged(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    InputIteratorT d_in,
    FlagIterator d_flags,
    OutputIteratorT d_out,
    NumSelectedIteratorT d_num_selected_out,
    ::cuda::std::int64_t num_items,
    cudaStream_t stream,
    bool debug_synchronous)
  {
    CUB_DETAIL_RUNTIME_DEBUG_SYNC_USAGE_LOG

    return Flagged<InputIteratorT, FlagIterator, OutputIteratorT, NumSelectedIteratorT>(
      d_temp_storage, temp_storage_bytes, d_in, d_flags, d_out, d_num_selected_out, num_items, stream);
  }
#endif // DOXYGEN_SHOULD_SKIP_THIS

  template <typename IteratorT, typename FlagIterator, typename NumSelectedIteratorT>
  CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t Flagged(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    IteratorT d_data,
    FlagIterator d_flags,
    NumSelectedIteratorT d_num_selected_out,
    ::cuda::std::int64_t num_items,
    cudaStream_t stream = 0)
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE_IF(d_temp_storage, "cub::DeviceSelect::Flagged");

    using OffsetT    = ::cuda::std::int64_t; // Signed integer type for global offsets
    using SelectOp   = NullType; // Selection op (not used)
    using EqualityOp = NullType; // Equality operator (not used)

    constexpr bool may_alias = true;

    return DispatchSelectIf<
      IteratorT,
      FlagIterator,
      IteratorT,
      NumSelectedIteratorT,
      SelectOp,
      EqualityOp,
      OffsetT,
      false,
      may_alias>::Dispatch(d_temp_storage,
                           temp_storage_bytes,
                           d_data, // in
                           d_flags,
                           d_data, // out
                           d_num_selected_out,
                           SelectOp(),
                           EqualityOp(),
                           num_items,
                           stream);
  }

#ifndef DOXYGEN_SHOULD_SKIP_THIS // Do not document
  template <typename IteratorT, typename FlagIterator, typename NumSelectedIteratorT>
  CUB_DETAIL_RUNTIME_DEBUG_SYNC_IS_NOT_SUPPORTED CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t Flagged(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    IteratorT d_data,
    FlagIterator d_flags,
    NumSelectedIteratorT d_num_selected_out,
    ::cuda::std::int64_t num_items,
    cudaStream_t stream,
    bool debug_synchronous)
  {
    CUB_DETAIL_RUNTIME_DEBUG_SYNC_USAGE_LOG

    return Flagged<IteratorT, FlagIterator, NumSelectedIteratorT>(
      d_temp_storage, temp_storage_bytes, d_data, d_flags, d_num_selected_out, num_items, stream);
  }
#endif // DOXYGEN_SHOULD_SKIP_THIS

  template <typename InputIteratorT, typename OutputIteratorT, typename NumSelectedIteratorT, typename SelectOp>
  CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t
  If(void* d_temp_storage,
     size_t& temp_storage_bytes,
     InputIteratorT d_in,
     OutputIteratorT d_out,
     NumSelectedIteratorT d_num_selected_out,
     ::cuda::std::int64_t num_items,
     SelectOp select_op,
     cudaStream_t stream = 0)
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE_IF(d_temp_storage, "cub::DeviceSelect::If");

    using OffsetT      = ::cuda::std::int64_t; // Signed integer type for global offsets
    using FlagIterator = NullType*; // FlagT iterator type (not used)
    using EqualityOp   = NullType; // Equality operator (not used)

    return DispatchSelectIf<
      InputIteratorT,
      FlagIterator,
      OutputIteratorT,
      NumSelectedIteratorT,
      SelectOp,
      EqualityOp,
      OffsetT,
      false>::Dispatch(d_temp_storage,
                       temp_storage_bytes,
                       d_in,
                       nullptr,
                       d_out,
                       d_num_selected_out,
                       select_op,
                       EqualityOp(),
                       num_items,
                       stream);
  }

#ifndef DOXYGEN_SHOULD_SKIP_THIS // Do not document
  template <typename InputIteratorT, typename OutputIteratorT, typename NumSelectedIteratorT, typename SelectOp>
  CUB_DETAIL_RUNTIME_DEBUG_SYNC_IS_NOT_SUPPORTED CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t
  If(void* d_temp_storage,
     size_t& temp_storage_bytes,
     InputIteratorT d_in,
     OutputIteratorT d_out,
     NumSelectedIteratorT d_num_selected_out,
     ::cuda::std::int64_t num_items,
     SelectOp select_op,
     cudaStream_t stream,
     bool debug_synchronous)
  {
    CUB_DETAIL_RUNTIME_DEBUG_SYNC_USAGE_LOG

    return If<InputIteratorT, OutputIteratorT, NumSelectedIteratorT, SelectOp>(
      d_temp_storage, temp_storage_bytes, d_in, d_out, d_num_selected_out, num_items, select_op, stream);
  }
#endif // DOXYGEN_SHOULD_SKIP_THIS

  template <typename IteratorT, typename NumSelectedIteratorT, typename SelectOp>
  CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t
  If(void* d_temp_storage,
     size_t& temp_storage_bytes,
     IteratorT d_data,
     NumSelectedIteratorT d_num_selected_out,
     ::cuda::std::int64_t num_items,
     SelectOp select_op,
     cudaStream_t stream = 0)
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE_IF(d_temp_storage, "cub::DeviceSelect::If");

    using OffsetT      = ::cuda::std::int64_t; // Signed integer type for global offsets
    using FlagIterator = NullType*; // FlagT iterator type (not used)
    using EqualityOp   = NullType; // Equality operator (not used)

    constexpr bool may_alias = true;

    return DispatchSelectIf<
      IteratorT,
      FlagIterator,
      IteratorT,
      NumSelectedIteratorT,
      SelectOp,
      EqualityOp,
      OffsetT,
      false,
      may_alias>::Dispatch(d_temp_storage,
                           temp_storage_bytes,
                           d_data, // in
                           nullptr,
                           d_data, // out
                           d_num_selected_out,
                           select_op,
                           EqualityOp(),
                           num_items,
                           stream);
  }

#ifndef DOXYGEN_SHOULD_SKIP_THIS // Do not document
  template <typename IteratorT, typename NumSelectedIteratorT, typename SelectOp>
  CUB_DETAIL_RUNTIME_DEBUG_SYNC_IS_NOT_SUPPORTED CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t
  If(void* d_temp_storage,
     size_t& temp_storage_bytes,
     IteratorT d_data,
     NumSelectedIteratorT d_num_selected_out,
     ::cuda::std::int64_t num_items,
     SelectOp select_op,
     cudaStream_t stream,
     bool debug_synchronous)
  {
    CUB_DETAIL_RUNTIME_DEBUG_SYNC_USAGE_LOG

    return If<IteratorT, NumSelectedIteratorT, SelectOp>(
      d_temp_storage, temp_storage_bytes, d_data, d_num_selected_out, num_items, select_op, stream);
  }
#endif // DOXYGEN_SHOULD_SKIP_THIS

  template <typename InputIteratorT,
            typename FlagIterator,
            typename OutputIteratorT,
            typename NumSelectedIteratorT,
            typename SelectOp>
  CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t FlaggedIf(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    InputIteratorT d_in,
    FlagIterator d_flags,
    OutputIteratorT d_out,
    NumSelectedIteratorT d_num_selected_out,
    ::cuda::std::int64_t num_items,
    SelectOp select_op,
    cudaStream_t stream = 0)
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE_IF(d_temp_storage, "cub::DeviceSelect::FlaggedIf");

    using OffsetT    = ::cuda::std::int64_t; // Signed integer type for global offsets
    using EqualityOp = NullType; // Equality operator (not used)

    return DispatchSelectIf<
      InputIteratorT,
      FlagIterator,
      OutputIteratorT,
      NumSelectedIteratorT,
      SelectOp,
      EqualityOp,
      OffsetT,
      false>::Dispatch(d_temp_storage,
                       temp_storage_bytes,
                       d_in,
                       d_flags,
                       d_out,
                       d_num_selected_out,
                       select_op,
                       EqualityOp(),
                       num_items,
                       stream);
  }

  template <typename IteratorT, typename FlagIterator, typename NumSelectedIteratorT, typename SelectOp>
  CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t FlaggedIf(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    IteratorT d_data,
    FlagIterator d_flags,
    NumSelectedIteratorT d_num_selected_out,
    ::cuda::std::int64_t num_items,
    SelectOp select_op,
    cudaStream_t stream = 0)
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE_IF(d_temp_storage, "cub::DeviceSelect::FlaggedIf");

    using OffsetT    = ::cuda::std::int64_t; // Signed integer type for global offsets
    using EqualityOp = NullType; // Equality operator (not used)

    constexpr bool may_alias = true;

    return DispatchSelectIf<
      IteratorT,
      FlagIterator,
      IteratorT,
      NumSelectedIteratorT,
      SelectOp,
      EqualityOp,
      OffsetT,
      false,
      may_alias>::Dispatch(d_temp_storage,
                           temp_storage_bytes,
                           d_data, // in
                           d_flags,
                           d_data, // out
                           d_num_selected_out,
                           select_op,
                           EqualityOp(),
                           num_items,
                           stream);
  }

  template <typename InputIteratorT, typename OutputIteratorT, typename NumSelectedIteratorT>
  CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t Unique(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    InputIteratorT d_in,
    OutputIteratorT d_out,
    NumSelectedIteratorT d_num_selected_out,
    ::cuda::std::int64_t num_items,
    cudaStream_t stream = 0)
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE_IF(d_temp_storage, "cub::DeviceSelect::Unique");

    using OffsetT      = ::cuda::std::int64_t;
    using FlagIterator = NullType*; // FlagT iterator type (not used)
    using SelectOp     = NullType; // Selection op (not used)
    using EqualityOp   = ::cuda::std::equal_to<>; // Default == operator

    return DispatchSelectIf<
      InputIteratorT,
      FlagIterator,
      OutputIteratorT,
      NumSelectedIteratorT,
      SelectOp,
      EqualityOp,
      OffsetT,
      false>::Dispatch(d_temp_storage,
                       temp_storage_bytes,
                       d_in,
                       nullptr,
                       d_out,
                       d_num_selected_out,
                       SelectOp(),
                       EqualityOp(),
                       num_items,
                       stream);
  }

#ifndef DOXYGEN_SHOULD_SKIP_THIS // Do not document
  template <typename InputIteratorT, typename OutputIteratorT, typename NumSelectedIteratorT>
  CUB_DETAIL_RUNTIME_DEBUG_SYNC_IS_NOT_SUPPORTED CUB_RUNTIME_FUNCTION _CCCL_FORCEINLINE static cudaError_t Unique(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    InputIteratorT d_in,
    OutputIteratorT d_out,
    NumSelectedIteratorT d_num_selected_out,
    ::cuda::std::int64_t num_items,
    cudaStream_t stream,
    bool debug_synchronous)
  {
    CUB_DETAIL_RUNTIME_DEBUG_SYNC_USAGE_LOG

    return Unique<InputIteratorT, OutputIteratorT, NumSelectedIteratorT>(
      d_temp_storage, temp_storage_bytes, d_in, d_out, d_num_selected_out, num_items, stream);
  }
#endif // DOXYGEN_SHOULD_SKIP_THIS

  template <typename KeyInputIteratorT,
            typename ValueInputIteratorT,
            typename KeyOutputIteratorT,
            typename ValueOutputIteratorT,
            typename NumSelectedIteratorT,
            typename NumItemsT,
            typename EqualityOpT>
  CUB_RUNTIME_FUNCTION __forceinline__ static //
    typename ::cuda::std::enable_if< //
      !::cuda::std::is_convertible<EqualityOpT, cudaStream_t>::value, //
      cudaError_t>::type
    UniqueByKey(
      void* d_temp_storage,
      size_t& temp_storage_bytes,
      KeyInputIteratorT d_keys_in,
      ValueInputIteratorT d_values_in,
      KeyOutputIteratorT d_keys_out,
      ValueOutputIteratorT d_values_out,
      NumSelectedIteratorT d_num_selected_out,
      NumItemsT num_items,
      EqualityOpT equality_op,
      cudaStream_t stream = 0)
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE_IF(d_temp_storage, "cub::DeviceSelect::UniqueByKey");

    using OffsetT = detail::choose_offset_t<NumItemsT>;

    return DispatchUniqueByKey<
      KeyInputIteratorT,
      ValueInputIteratorT,
      KeyOutputIteratorT,
      ValueOutputIteratorT,
      NumSelectedIteratorT,
      EqualityOpT,
      OffsetT>::Dispatch(d_temp_storage,
                         temp_storage_bytes,
                         d_keys_in,
                         d_values_in,
                         d_keys_out,
                         d_values_out,
                         d_num_selected_out,
                         equality_op,
                         static_cast<OffsetT>(num_items),
                         stream);
  }

  template <typename KeyInputIteratorT,
            typename ValueInputIteratorT,
            typename KeyOutputIteratorT,
            typename ValueOutputIteratorT,
            typename NumSelectedIteratorT,
            typename NumItemsT>
  CUB_RUNTIME_FUNCTION __forceinline__ static cudaError_t UniqueByKey(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    KeyInputIteratorT d_keys_in,
    ValueInputIteratorT d_values_in,
    KeyOutputIteratorT d_keys_out,
    ValueOutputIteratorT d_values_out,
    NumSelectedIteratorT d_num_selected_out,
    NumItemsT num_items,
    cudaStream_t stream = 0)
  {
    return UniqueByKey(
      d_temp_storage,
      temp_storage_bytes,
      d_keys_in,
      d_values_in,
      d_keys_out,
      d_values_out,
      d_num_selected_out,
      num_items,
      ::cuda::std::equal_to<>{},
      stream);
  }

#ifndef DOXYGEN_SHOULD_SKIP_THIS // Do not document
  template <typename KeyInputIteratorT,
            typename ValueInputIteratorT,
            typename KeyOutputIteratorT,
            typename ValueOutputIteratorT,
            typename NumSelectedIteratorT,
            typename NumItemsT>
  CUB_DETAIL_RUNTIME_DEBUG_SYNC_IS_NOT_SUPPORTED CUB_RUNTIME_FUNCTION __forceinline__ static cudaError_t UniqueByKey(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    KeyInputIteratorT d_keys_in,
    ValueInputIteratorT d_values_in,
    KeyOutputIteratorT d_keys_out,
    ValueOutputIteratorT d_values_out,
    NumSelectedIteratorT d_num_selected_out,
    NumItemsT num_items,
    cudaStream_t stream,
    bool debug_synchronous)
  {
    CUB_DETAIL_RUNTIME_DEBUG_SYNC_USAGE_LOG

    return UniqueByKey<KeyInputIteratorT,
                       ValueInputIteratorT,
                       KeyOutputIteratorT,
                       ValueOutputIteratorT,
                       NumSelectedIteratorT,
                       NumItemsT>(
      d_temp_storage,
      temp_storage_bytes,
      d_keys_in,
      d_values_in,
      d_keys_out,
      d_values_out,
      d_num_selected_out,
      num_items,
      stream);
  }
#endif // DOXYGEN_SHOULD_SKIP_THIS
};

CUB_NAMESPACE_END