/home/runner/work/cccl/cccl/cub/cub/device/device_for.cuh

File members: /home/runner/work/cccl/cccl/cub/cub/device/device_for.cuh

/******************************************************************************
 * Copyright (c) 2024, 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/nvtx.cuh>
#include <cub/device/dispatch/dispatch_for.cuh>
#include <cub/util_namespace.cuh>

#include <thrust/detail/raw_reference_cast.h>
#include <thrust/distance.h>
#include <thrust/iterator/iterator_traits.h>
#include <thrust/system/cuda/detail/core/util.h>
#include <thrust/type_traits/is_contiguous_iterator.h>

#include <cuda/std/type_traits>

CUB_NAMESPACE_BEGIN

namespace detail
{

namespace for_each
{

template <class OffsetT, class OpT, class RandomAccessIteratorT>
struct op_wrapper_t
{
  RandomAccessIteratorT input;
  OpT op;

  _CCCL_DEVICE _CCCL_FORCEINLINE void operator()(OffsetT i)
  {
    // Dereferencing `thrust::device_vector<T>` iterators returns a `thrust::device_reference<T>`
    // instead of `T`. Since user-provided operator expects `T` as an argument, we need to unwrap.
    (void) op(THRUST_NS_QUALIFIER::raw_reference_cast(*(input + i)));
  }
};

template <class OffsetT, class OpT, class T>
struct op_wrapper_vectorized_t
{
  const T* input; // Raw pointer to the input data
  OpT op; // User-provided operator
  OffsetT partially_filled_vector_id; // Index of the vector that doesn't have all elements
  OffsetT num_items; // Total number of non-vectorized items

  // TODO Can be extracted into tuning
  constexpr static int vec_size = 4;

  // Type of the vector that is used to load the input data
  using vector_t = typename CubVector<T, vec_size>::Type;

  _CCCL_DEVICE _CCCL_FORCEINLINE void operator()(OffsetT i)
  {
    // Surrounding `Bulk` call doesn't invoke this operator on invalid indices, so we don't need to
    // check for out-of-bounds access here.
    if (i != partially_filled_vector_id)
    { // Case of fully filled vector
      const vector_t vec = *reinterpret_cast<const vector_t*>(input + vec_size * i);

#pragma unroll
      for (int j = 0; j < vec_size; j++)
      {
        (void) op(*(reinterpret_cast<const T*>(&vec) + j));
      }
    }
    else
    { // Case of partially filled vector
      for (OffsetT j = i * vec_size; j < num_items; j++)
      {
        (void) op(input[j]);
      }
    }
  }
};

} // namespace for_each
} // namespace detail

struct DeviceFor
{
private:
  template <class VectorT, class T>
  CUB_RUNTIME_FUNCTION static bool is_aligned(const T* ptr)
  {
    return (reinterpret_cast<std::size_t>(ptr) & (sizeof(VectorT) - 1)) == 0;
  }

  template <class RandomAccessIteratorT, class OffsetT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t for_each_n(
    RandomAccessIteratorT first,
    OffsetT num_items,
    OpT op,
    cudaStream_t stream,
    ::cuda::std::false_type /* do_not_vectorize */)
  {
    using wrapped_op_t = detail::for_each::op_wrapper_t<OffsetT, OpT, RandomAccessIteratorT>;
    return detail::for_each::dispatch_t<OffsetT, wrapped_op_t>::dispatch(num_items, wrapped_op_t{first, op}, stream);
  }

  template <class RandomAccessIteratorT, class OffsetT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t for_each_n(
    RandomAccessIteratorT first, OffsetT num_items, OpT op, cudaStream_t stream, ::cuda::std::true_type /* vectorize */)
  {
    auto unwrapped_first = THRUST_NS_QUALIFIER::raw_pointer_cast(&*first);
    using wrapped_op_t =
      detail::for_each::op_wrapper_vectorized_t<OffsetT, OpT, detail::value_t<RandomAccessIteratorT>>;

    if (is_aligned<typename wrapped_op_t::vector_t>(unwrapped_first))
    { // Vectorize loads
      const OffsetT num_vec_items = cub::DivideAndRoundUp(num_items, wrapped_op_t::vec_size);

      return detail::for_each::dispatch_t<OffsetT, wrapped_op_t>::dispatch(
        num_vec_items,
        wrapped_op_t{
          unwrapped_first, op, num_items % wrapped_op_t::vec_size ? num_vec_items - 1 : num_vec_items, num_items},
        stream);
    }

    // Fallback to non-vectorized version
    return for_each_n(first, num_items, op, stream, ::cuda::std::false_type{});
  }

public:
  template <class ShapeT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t
  Bulk(void* d_temp_storage, size_t& temp_storage_bytes, ShapeT shape, OpT op, cudaStream_t stream = {})
  {
    static_assert(::cuda::std::is_integral<ShapeT>::value, "ShapeT must be an integral type");

    if (d_temp_storage == nullptr)
    {
      temp_storage_bytes = 1;
      return cudaSuccess;
    }

    return Bulk(shape, op, stream);
  }

  template <class RandomAccessIteratorT, class NumItemsT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t ForEachN(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    RandomAccessIteratorT first,
    NumItemsT num_items,
    OpT op,
    cudaStream_t stream = {})
  {
    if (d_temp_storage == nullptr)
    {
      temp_storage_bytes = 1;
      return cudaSuccess;
    }

    return ForEachN(first, num_items, op, stream);
  }

  template <class RandomAccessIteratorT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t ForEach(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    RandomAccessIteratorT first,
    RandomAccessIteratorT last,
    OpT op,
    cudaStream_t stream = {})
  {
    if (d_temp_storage == nullptr)
    {
      temp_storage_bytes = 1;
      return cudaSuccess;
    }

    return ForEach(first, last, op, stream);
  }

  template <class RandomAccessIteratorT, class NumItemsT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t ForEachCopyN(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    RandomAccessIteratorT first,
    NumItemsT num_items,
    OpT op,
    cudaStream_t stream = {})
  {
    if (d_temp_storage == nullptr)
    {
      temp_storage_bytes = 1;
      return cudaSuccess;
    }

    return ForEachCopyN(first, num_items, op, stream);
  }

  template <class RandomAccessIteratorT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t ForEachCopy(
    void* d_temp_storage,
    size_t& temp_storage_bytes,
    RandomAccessIteratorT first,
    RandomAccessIteratorT last,
    OpT op,
    cudaStream_t stream = {})
  {
    if (d_temp_storage == nullptr)
    {
      temp_storage_bytes = 1;
      return cudaSuccess;
    }

    return ForEachCopy(first, last, op, stream);
  }

  template <class ShapeT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t Bulk(ShapeT shape, OpT op, cudaStream_t stream = {})
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE("cub::DeviceFor::Bulk");
    static_assert(::cuda::std::is_integral<ShapeT>::value, "ShapeT must be an integral type");
    using offset_t = ShapeT;
    return detail::for_each::dispatch_t<offset_t, OpT>::dispatch(static_cast<offset_t>(shape), op, stream);
  }

private:
  // Internal version without NVTX raNGE
  template <class RandomAccessIteratorT, class NumItemsT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t
  ForEachNNoNVTX(RandomAccessIteratorT first, NumItemsT num_items, OpT op, cudaStream_t stream = {})
  {
    using offset_t            = NumItemsT;
    using use_vectorization_t = ::cuda::std::integral_constant<bool, false>;

    // Disable auto-vectorization for now:
    // constexpr bool use_vectorization =
    //   detail::for_each::can_regain_copy_freedom<detail::value_t<RandomAccessIteratorT>, OpT>::value
    //   && THRUST_NS_QUALIFIER::is_contiguous_iterator<RandomAccessIteratorT>::value;

    return for_each_n<RandomAccessIteratorT, offset_t, OpT>(first, num_items, op, stream, use_vectorization_t{});
  }

public:
  template <class RandomAccessIteratorT, class NumItemsT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t
  ForEachN(RandomAccessIteratorT first, NumItemsT num_items, OpT op, cudaStream_t stream = {})
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE("cub::DeviceFor::ForEachN");
    return ForEachNNoNVTX(first, num_items, op, stream);
  }

  template <class RandomAccessIteratorT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t
  ForEach(RandomAccessIteratorT first, RandomAccessIteratorT last, OpT op, cudaStream_t stream = {})
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE("cub::DeviceFor::ForEach");

    using offset_t = typename THRUST_NS_QUALIFIER::iterator_traits<RandomAccessIteratorT>::difference_type;

    const auto num_items = static_cast<offset_t>(THRUST_NS_QUALIFIER::distance(first, last));

    return ForEachNNoNVTX(first, num_items, op, stream);
  }

private:
  // Internal version without NVTX range
  template <class RandomAccessIteratorT, class NumItemsT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t
  ForEachCopyNNoNVTX(RandomAccessIteratorT first, NumItemsT num_items, OpT op, cudaStream_t stream = {})
  {
    static_assert(THRUST_NS_QUALIFIER::is_contiguous_iterator<RandomAccessIteratorT>::value,
                  "Iterator must be contiguous");

    using offset_t            = NumItemsT;
    using use_vectorization_t = ::cuda::std::integral_constant<bool, true>;

    return for_each_n<RandomAccessIteratorT, offset_t, OpT>(first, num_items, op, stream, use_vectorization_t{});
  }

public:
  template <class RandomAccessIteratorT, class NumItemsT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t
  ForEachCopyN(RandomAccessIteratorT first, NumItemsT num_items, OpT op, cudaStream_t stream = {})
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE("cub::DeviceFor::ForEachCopyN");
    return ForEachCopyNNoNVTX(first, num_items, op, stream);
  }

  template <class RandomAccessIteratorT, class OpT>
  CUB_RUNTIME_FUNCTION static cudaError_t
  ForEachCopy(RandomAccessIteratorT first, RandomAccessIteratorT last, OpT op, cudaStream_t stream = {})
  {
    CUB_DETAIL_NVTX_RANGE_SCOPE("cub::DeviceFor::ForEachCopy");
    static_assert(THRUST_NS_QUALIFIER::is_contiguous_iterator<RandomAccessIteratorT>::value,
                  "Iterator must be contiguous");

    using offset_t = typename THRUST_NS_QUALIFIER::iterator_traits<RandomAccessIteratorT>::difference_type;

    const auto num_items = static_cast<offset_t>(THRUST_NS_QUALIFIER::distance(first, last));

    return ForEachCopyNNoNVTX(first, num_items, op, stream);
  }
};

CUB_NAMESPACE_END