cuda_toolkit/include/cub/agent/agent_segment_fixup.cuh

/******************************************************************************
 * Copyright (c) 2011, Duane Merrill.  All rights reserved.
 * Copyright (c) 2011-2018, 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.
 *
 ******************************************************************************/

/**
 * \file
 * cub::AgentSegmentFixup implements a stateful abstraction of CUDA thread blocks for participating in device-wide reduce-value-by-key.
 */

#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/agent/single_pass_scan_operators.cuh>
#include <cub/block/block_discontinuity.cuh>
#include <cub/block/block_load.cuh>
#include <cub/block/block_scan.cuh>
#include <cub/block/block_store.cuh>
#include <cub/iterator/cache_modified_input_iterator.cuh>
#include <cub/iterator/constant_input_iterator.cuh>

#include <iterator>

CUB_NAMESPACE_BEGIN


/******************************************************************************
 * Tuning policy types
 ******************************************************************************/

/**
 * @brief Parameterizable tuning policy type for AgentSegmentFixup
 *
 * @tparam _BLOCK_THREADS
 *   Threads per thread block
 *
 * @tparam _ITEMS_PER_THREAD
 *   Items per thread (per tile of input)
 *
 * @tparam _LOAD_ALGORITHM
 *   The BlockLoad algorithm to use
 *
 * @tparam _LOAD_MODIFIER
 *   Cache load modifier for reading input elements
 *
 * @tparam _SCAN_ALGORITHM
 *   The BlockScan algorithm to use
 */
template <int _BLOCK_THREADS,
          int _ITEMS_PER_THREAD,
          BlockLoadAlgorithm _LOAD_ALGORITHM,
          CacheLoadModifier _LOAD_MODIFIER,
          BlockScanAlgorithm _SCAN_ALGORITHM>
struct AgentSegmentFixupPolicy
{
  enum
  {
    /// Threads per thread block
    BLOCK_THREADS = _BLOCK_THREADS,

    /// Items per thread (per tile of input)
    ITEMS_PER_THREAD = _ITEMS_PER_THREAD,
  };

  /// The BlockLoad algorithm to use
  static constexpr BlockLoadAlgorithm LOAD_ALGORITHM = _LOAD_ALGORITHM;

  /// Cache load modifier for reading input elements
  static constexpr CacheLoadModifier LOAD_MODIFIER = _LOAD_MODIFIER;

  /// The BlockScan algorithm to use
  static constexpr BlockScanAlgorithm SCAN_ALGORITHM = _SCAN_ALGORITHM;
};

/******************************************************************************
 * Thread block abstractions
 ******************************************************************************/

/**
 * @brief AgentSegmentFixup implements a stateful abstraction of CUDA thread blocks for
 * participating in device-wide reduce-value-by-key
 *
 * @tparam AgentSegmentFixupPolicyT
 *   Parameterized AgentSegmentFixupPolicy tuning policy type
 *
 * @tparam PairsInputIteratorT
 *   Random-access input iterator type for keys
 *
 * @tparam AggregatesOutputIteratorT
 *   Random-access output iterator type for values
 *
 * @tparam EqualityOpT
 *   KeyT equality operator type
 *
 * @tparam ReductionOpT
 *   ValueT reduction operator type
 *
 * @tparam OffsetT
 *   Signed integer type for global offsets
 */
template <typename AgentSegmentFixupPolicyT,
          typename PairsInputIteratorT,
          typename AggregatesOutputIteratorT,
          typename EqualityOpT,
          typename ReductionOpT,
          typename OffsetT>
struct AgentSegmentFixup
{
    //---------------------------------------------------------------------
    // Types and constants
    //---------------------------------------------------------------------

    // Data type of key-value input iterator
    using KeyValuePairT = cub::detail::value_t<PairsInputIteratorT>;

    // Value type
    using ValueT = typename KeyValuePairT::Value;

    // Tile status descriptor interface type
    using ScanTileStateT = ReduceByKeyScanTileState<ValueT, OffsetT>;

    // Constants
    enum
    {
      BLOCK_THREADS    = AgentSegmentFixupPolicyT::BLOCK_THREADS,
      ITEMS_PER_THREAD = AgentSegmentFixupPolicyT::ITEMS_PER_THREAD,
      TILE_ITEMS       = BLOCK_THREADS * ITEMS_PER_THREAD,

      // Whether or not do fixup using RLE + global atomics
      USE_ATOMIC_FIXUP = (std::is_same<ValueT, float>::value ||
                          std::is_same<ValueT, int>::value ||
                          std::is_same<ValueT, unsigned int>::value ||
                          std::is_same<ValueT, unsigned long long>::value),

      // Whether or not the scan operation has a zero-valued identity value
      // (true if we're performing addition on a primitive type)
      HAS_IDENTITY_ZERO = (std::is_same<ReductionOpT, cub::Sum>::value) &&
                          (Traits<ValueT>::PRIMITIVE),
    };

    // Cache-modified Input iterator wrapper type (for applying cache modifier) for keys
    // Wrap the native input pointer with CacheModifiedValuesInputIterator
    // or directly use the supplied input iterator type
    using WrappedPairsInputIteratorT = cub::detail::conditional_t<
      std::is_pointer<PairsInputIteratorT>::value,
      CacheModifiedInputIterator<AgentSegmentFixupPolicyT::LOAD_MODIFIER,
                                 KeyValuePairT,
                                 OffsetT>,
      PairsInputIteratorT>;

    // Cache-modified Input iterator wrapper type (for applying cache modifier) for fixup values
    // Wrap the native input pointer with CacheModifiedValuesInputIterator
    // or directly use the supplied input iterator type
    using WrappedFixupInputIteratorT = cub::detail::conditional_t<
      std::is_pointer<AggregatesOutputIteratorT>::value,
      CacheModifiedInputIterator<AgentSegmentFixupPolicyT::LOAD_MODIFIER,
                                 ValueT,
                                 OffsetT>,
      AggregatesOutputIteratorT>;

    // Reduce-value-by-segment scan operator
    using ReduceBySegmentOpT = ReduceByKeyOp<cub::Sum>;

    // Parameterized BlockLoad type for pairs
    using BlockLoadPairs = BlockLoad<KeyValuePairT,
                                     BLOCK_THREADS,
                                     ITEMS_PER_THREAD,
                                     AgentSegmentFixupPolicyT::LOAD_ALGORITHM>;

    // Parameterized BlockScan type
    using BlockScanT = BlockScan<KeyValuePairT,
                                 BLOCK_THREADS,
                                 AgentSegmentFixupPolicyT::SCAN_ALGORITHM>;

    // Callback type for obtaining tile prefix during block scan
    using TilePrefixCallbackOpT =
      TilePrefixCallbackOp<KeyValuePairT, ReduceBySegmentOpT, ScanTileStateT>;

    // Shared memory type for this thread block
    union _TempStorage
    {
        struct ScanStorage
        {
          // Smem needed for tile scanning
          typename BlockScanT::TempStorage scan;

          // Smem needed for cooperative prefix callback
          typename TilePrefixCallbackOpT::TempStorage prefix;
        } scan_storage;

        // Smem needed for loading keys
        typename BlockLoadPairs::TempStorage load_pairs;
    };

    // Alias wrapper allowing storage to be unioned
    struct TempStorage : Uninitialized<_TempStorage> {};


    //---------------------------------------------------------------------
    // Per-thread fields
    //---------------------------------------------------------------------

    _TempStorage &temp_storage;                   ///< Reference to temp_storage
    WrappedPairsInputIteratorT d_pairs_in;        ///< Input keys
    AggregatesOutputIteratorT d_aggregates_out;   ///< Output value aggregates
    WrappedFixupInputIteratorT d_fixup_in;        ///< Fixup input values
    InequalityWrapper<EqualityOpT> inequality_op; ///< KeyT inequality operator
    ReductionOpT reduction_op;                    ///< Reduction operator
    ReduceBySegmentOpT scan_op;                   ///< Reduce-by-segment scan operator

    //---------------------------------------------------------------------
    // Constructor
    //---------------------------------------------------------------------

    /**
     * @param temp_storage
     *   Reference to temp_storage
     *
     * @param d_pairs_in
     *   Input keys
     *
     * @param d_aggregates_out
     *   Output value aggregates
     *
     * @param equality_op
     *   KeyT equality operator
     *
     * @param reduction_op
     *   ValueT reduction operator
     */
    __device__ __forceinline__ AgentSegmentFixup(TempStorage &temp_storage,
                                                 PairsInputIteratorT d_pairs_in,
                                                 AggregatesOutputIteratorT d_aggregates_out,
                                                 EqualityOpT equality_op,
                                                 ReductionOpT reduction_op)
        : temp_storage(temp_storage.Alias())
        , d_pairs_in(d_pairs_in)
        , d_aggregates_out(d_aggregates_out)
        , d_fixup_in(d_aggregates_out)
        , inequality_op(equality_op)
        , reduction_op(reduction_op)
        , scan_op(reduction_op)
    {}

    //---------------------------------------------------------------------
    // Cooperatively scan a device-wide sequence of tiles with other CTAs
    //---------------------------------------------------------------------


    /**
     * @brief Process input tile. Specialized for atomic-fixup
     *
     * @param num_remaining
     *   Number of global input items remaining (including this tile)
     *
     * @param tile_idx
     *   Tile index
     *
     * @param tile_offset
     *   Tile offset
     *
     * @param tile_state
     *   Global tile state descriptor
     *
     * @param use_atomic_fixup
     *   Marker whether to use atomicAdd (instead of reduce-by-key)
     */
    template <bool IS_LAST_TILE>
    __device__ __forceinline__ void ConsumeTile(OffsetT num_remaining,
                                                int tile_idx,
                                                OffsetT tile_offset,
                                                ScanTileStateT &tile_state,
                                                Int2Type<true> use_atomic_fixup)
    {
        KeyValuePairT   pairs[ITEMS_PER_THREAD];

        // Load pairs
        KeyValuePairT oob_pair;
        oob_pair.key = -1;

        if (IS_LAST_TILE)
            BlockLoadPairs(temp_storage.load_pairs).Load(d_pairs_in + tile_offset, pairs, num_remaining, oob_pair);
        else
            BlockLoadPairs(temp_storage.load_pairs).Load(d_pairs_in + tile_offset, pairs);

        // RLE
        #pragma unroll
        for (int ITEM = 1; ITEM < ITEMS_PER_THREAD; ++ITEM)
        {
            ValueT* d_scatter = d_aggregates_out + pairs[ITEM - 1].key;
            if (pairs[ITEM].key != pairs[ITEM - 1].key)
                atomicAdd(d_scatter, pairs[ITEM - 1].value);
            else
                pairs[ITEM].value = reduction_op(pairs[ITEM - 1].value, pairs[ITEM].value);
        }

        // Flush last item if valid
        ValueT* d_scatter = d_aggregates_out + pairs[ITEMS_PER_THREAD - 1].key;
        if ((!IS_LAST_TILE) || (pairs[ITEMS_PER_THREAD - 1].key >= 0))
            atomicAdd(d_scatter, pairs[ITEMS_PER_THREAD - 1].value);
    }

    /**
     * @brief Process input tile. Specialized for reduce-by-key fixup
     *
     * @param num_remaining
     *   Number of global input items remaining (including this tile)
     *
     * @param tile_idx
     *   Tile index
     *
     * @param tile_offset
     *   Tile offset
     *
     * @param tile_state
     *   Global tile state descriptor
     *
     * @param use_atomic_fixup
     *   Marker whether to use atomicAdd (instead of reduce-by-key)
     */
    template <bool IS_LAST_TILE>
    __device__ __forceinline__ void ConsumeTile(OffsetT num_remaining,
                                                int tile_idx,
                                                OffsetT tile_offset,
                                                ScanTileStateT &tile_state,
                                                Int2Type<false> use_atomic_fixup)
    {
        KeyValuePairT   pairs[ITEMS_PER_THREAD];
        KeyValuePairT   scatter_pairs[ITEMS_PER_THREAD];

        // Load pairs
        KeyValuePairT oob_pair;
        oob_pair.key = -1;

        if (IS_LAST_TILE)
            BlockLoadPairs(temp_storage.load_pairs).Load(d_pairs_in + tile_offset, pairs, num_remaining, oob_pair);
        else
            BlockLoadPairs(temp_storage.load_pairs).Load(d_pairs_in + tile_offset, pairs);

        CTA_SYNC();

        KeyValuePairT tile_aggregate;
        if (tile_idx == 0)
        {
            // Exclusive scan of values and segment_flags
            BlockScanT(temp_storage.scan_storage.scan).ExclusiveScan(pairs, scatter_pairs, scan_op, tile_aggregate);

            // Update tile status if this is not the last tile
            if (threadIdx.x == 0)
            {
                // Set first segment id to not trigger a flush (invalid from exclusive scan)
                scatter_pairs[0].key = pairs[0].key;

                if (!IS_LAST_TILE)
                    tile_state.SetInclusive(0, tile_aggregate);

            }
        }
        else
        {
            // Exclusive scan of values and segment_flags
            TilePrefixCallbackOpT prefix_op(tile_state, temp_storage.scan_storage.prefix, scan_op, tile_idx);
            BlockScanT(temp_storage.scan_storage.scan).ExclusiveScan(pairs, scatter_pairs, scan_op, prefix_op);
            tile_aggregate = prefix_op.GetBlockAggregate();
        }

        // Scatter updated values
        #pragma unroll
        for (int ITEM = 0; ITEM < ITEMS_PER_THREAD; ++ITEM)
        {
            if (scatter_pairs[ITEM].key != pairs[ITEM].key)
            {
                // Update the value at the key location
                ValueT value    = d_fixup_in[scatter_pairs[ITEM].key];
                value           = reduction_op(value, scatter_pairs[ITEM].value);

                d_aggregates_out[scatter_pairs[ITEM].key] = value;
            }
        }

        // Finalize the last item
        if (IS_LAST_TILE)
        {
            // Last thread will output final count and last item, if necessary
            if (threadIdx.x == BLOCK_THREADS - 1)
            {
                // If the last tile is a whole tile, the inclusive prefix contains accumulated value reduction for the last segment
                if (num_remaining == TILE_ITEMS)
                {
                    // Update the value at the key location
                    OffsetT last_key = pairs[ITEMS_PER_THREAD - 1].key;
                    d_aggregates_out[last_key] = reduction_op(tile_aggregate.value, d_fixup_in[last_key]);
                }
            }
        }
    }

    /**
     * @brief Scan tiles of items as part of a dynamic chained scan
     *
     * @param num_items
     *   Total number of input items
     *
     * @param num_tiles
     *   Total number of input tiles
     *
     * @param tile_state
     *   Global tile state descriptor
     */
    __device__ __forceinline__ void ConsumeRange(OffsetT num_items,
                                                 int num_tiles,
                                                 ScanTileStateT &tile_state)
    {
        // Blocks are launched in increasing order, so just assign one tile per block
        int tile_idx          = (blockIdx.x * gridDim.y) + blockIdx.y; // Current tile index
        OffsetT tile_offset   = tile_idx * TILE_ITEMS;   // Global offset for the current tile
        OffsetT num_remaining = num_items - tile_offset; // Remaining items (including this tile)

        if (num_remaining > TILE_ITEMS)
        {
            // Not the last tile (full)
            ConsumeTile<false>(num_remaining, tile_idx, tile_offset, tile_state, Int2Type<USE_ATOMIC_FIXUP>());
        }
        else if (num_remaining > 0)
        {
            // The last tile (possibly partially-full)
            ConsumeTile<true>(num_remaining, tile_idx, tile_offset, tile_state, Int2Type<USE_ATOMIC_FIXUP>());
        }
    }

};


CUB_NAMESPACE_END