reduce_op.cu.h

// Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

#pragma once

#include <algorithm>
#include <cmath>
#include <numeric>
#include <set>
#include <vector>

#ifdef __NVCC__
#include "cub/cub.cuh"
#endif

#ifdef __HIPCC__
#include <hipcub/hipcub.hpp>
namespace cub = hipcub;
#endif

#include "paddle/fluid/framework/array.h"
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/framework/tensor.h"
#include "paddle/fluid/framework/tensor_util.h"
#include "paddle/fluid/operators/amp/fp16_type_traits.h"
#include "paddle/fluid/operators/kernel_primitives/kernel_primitives.h"
#include "paddle/fluid/platform/cuda_device_function.h"
#include "paddle/fluid/platform/fast_divmod.h"

// Reduce split or not, Whether to use ReduceHigherDim
#define REDUCE_SPLIT_BOUNDARY 512
#define REDUCE_VEC_SIZE 4

namespace paddle {
namespace operators {

namespace kps = paddle::operators::kernel_primitives;

namespace details {

static inline int GetLastPow2(int n) {
  n |= (n >> 1);
  n |= (n >> 2);
  n |= (n >> 4);
  n |= (n >> 8);
  n |= (n >> 16);
  return std::max(1, n - (n >> 1));
}

static inline int64_t AlignUp(int64_t a, int64_t b) { return (a + b - 1) / b; }

// get strides of x_dim, reduce_dim and left_dim for reduceLastDim and reduceAny
static inline std::vector<int> GetDimStrides(const std::vector<int>& dims,
                                             const std::vector<int>& idx) {
  int n = static_cast<int>(idx.size());
  if (n == 0) return std::vector<int>();
  std::vector<int> strides(n);
  strides.back() = 1;
  for (int i = n - 2; i >= 0; --i) {
    strides[i] = strides[i + 1] * dims[idx[i + 1]];
  }
  return strides;
}

// get blockDim for reduceLastDim and reduceAny
static inline int GetBlockDim(int block_dim) {
  return block_dim >= kps::details::kReduceMaxThread
             ? kps::details::kReduceMaxThread
             : GetLastPow2(block_dim);
}

// check reduce rand is valid
static inline void CheckReduceRank(int reduce_rank, int rank) {
  if (rank % 2 == 0) {
    PADDLE_ENFORCE_EQ(reduce_rank, rank / 2,
                      platform::errors::InvalidArgument(
                          "ReduceOp: invalid reduce rank. When rank = %d, "
                          "reduce_rank must be %d, but got %d.",
                          rank, rank / 2, reduce_rank));
  } else {
    auto lower_rank = (rank - 1) / 2;
    auto upper_rank = (rank + 1) / 2;
    PADDLE_ENFORCE_EQ(
        reduce_rank == lower_rank || reduce_rank == upper_rank, true,
        platform::errors::InvalidArgument(
            "ReduceOp: invalid reduce rank. When rank = %d, reduce_rank "
            "must be %d or %d, but got %d.",
            rank, lower_rank, upper_rank, reduce_rank));
  }
}

// convert dims from vector to array
template <typename T, size_t ElementCount, typename VectorLikeType>
static inline paddle::framework::Array<T, ElementCount> VectorToArray(
    const VectorLikeType& vec) {
  PADDLE_ENFORCE_LE(vec.size(), ElementCount,
                    platform::errors::InvalidArgument(
                        "Cub reduce Array: size not match. Received "
                        "vec.size() %d > ElementCount %d.",
                        vec.size(), ElementCount));
  size_t n = static_cast<size_t>(vec.size());
  paddle::framework::Array<T, ElementCount> ret;
  for (size_t i = 0; i < n; ++i) {
    ret[i] = vec[i];
  }
  return ret;
}

}  // namespace details

using Tensor = framework::Tensor;
constexpr int kMaxRank = framework::DDim::kMaxRank;

enum ReduceType {
  kReduceLastDim = 0x01,    // when reduce_dim[0] == x_dim.size() - 1;
  kReduceHigherDim = 0x02,  // ReduceFirstDim or reduceSecondDim
  kReduceAny = 0x03,        // when reduce_dim.size() > 1
};

struct IndexCalculator {
  IndexCalculator(int dim, const std::vector<int>& cal_dims,
                  const std::vector<int>& cal_strides,
                  const std::vector<int>& full_strides)
      : dim(dim) {
    dims = details::VectorToArray<int, kMaxRank>(cal_dims);
    strides = details::VectorToArray<int, kMaxRank>(full_strides);
    std::vector<platform::FastDivMod> cal_divmoders;
    // fast divmod
    for (auto i : cal_strides) {
      cal_divmoders.push_back(platform::FastDivMod(i));
    }
    divmoders =
        details::VectorToArray<platform::FastDivMod, kMaxRank>(cal_divmoders);
  }

  __device__ inline int operator()(int offset) const {
    int index = 0;
#pragma unroll
    for (int i = 0; i < kMaxRank; ++i) {
      if (i == dim) {
        break;
      }
      auto divmod = divmoders[i].Divmod(offset);
      index += (divmod.val[0] * strides[dims[i]]);
      offset = divmod.val[1];
    }
    return index;
  }

  int dim;
  framework::Array<int, kMaxRank> dims;
  framework::Array<int, kMaxRank> strides;
  framework::Array<platform::FastDivMod, kMaxRank> divmoders;
};

// when reduce_type == kReduceLastDim this struct will be used
// for higher performance
struct LastDimIndexCal {
  explicit LastDimIndexCal(int num) : stride(num) {}

  __device__ inline int operator()(int index) const { return index * stride; }
  int stride;
};

// reduce config
template <typename Ty>
struct ReduceConfig {
  ReduceConfig(const std::vector<int>& origin_reduce_dims,
               const std::vector<int>& origin_x_dim)
      : reduce_dims_origin(origin_reduce_dims), x_dim(origin_x_dim) {}

  // get the parameters of reduceKernel
  void Run() {
    // step1: update the reduce_dim left_dim and x_dim
    SetReduceDim();

    // step2: get the strides of dim for reduceAny and reduceLastDim
    SetStrides();

    // step3: get the type of reduce
    SetReduceType();

    // step4: set the block and grid for launch kernel
    SetBlockDim();
  }

  // when should_reduce_again is true, we need malloc temp space for temp data
  void SetOutputData(Ty* y_data, const platform::Place& place,
                     framework::Tensor* tmp) {
    if (should_reduce_again) {
      output_data = tmp->mutable_data<Ty>(
          framework::make_ddim(
              {static_cast<int64_t>(left_num * grid.z * grid.y * sizeof(Ty))}),
          place);
    } else {
      output_data = y_data;
    }
  }

 private:
  // set reduce_dim, left_dim and update x_dim
  // eg: x_dim = [2, 4, 6] origin_reduce_dims = [0, 1]
  //     --SetReduceDim--> x_dim = [8,6], reduce_dim = [0], left_dim = [1]
  void SetReduceDim() {
    std::set<int> reduce_set;
    for (auto e : reduce_dims_origin) {
      auto pos = e >= 0 ? e : e + x_dim.size();
      reduce_set.insert(pos);
    }

    std::vector<int> reduce_dim_temp(reduce_set.begin(), reduce_set.end());
    std::sort(reduce_dim_temp.begin(), reduce_dim_temp.end());

    // update reduce_dim and x_dim
    std::vector<int> x_new_dim;

    reduce_dim.push_back(reduce_dim_temp[0]);
    x_new_dim.push_back(x_dim[0]);

    int idx_reduce = 1;
    int num = 0;

    if (reduce_dim_temp.size() > 1) {
      for (int i = 1; i < x_dim.size(); i++) {
        if ((idx_reduce < reduce_dim_temp.size()) &&
            (i == reduce_dim_temp[idx_reduce])) {
          int result =
              reduce_dim_temp[idx_reduce] - reduce_dim[reduce_dim.size() - 1];
          bool is_equal = ((result - num) == 1);
          if (is_equal) {
            x_new_dim[x_new_dim.size() - 1] *= x_dim[i];
            num++;
          } else {
            reduce_dim.push_back(reduce_dim_temp[idx_reduce] - num);
            x_new_dim.push_back(x_dim[i]);
          }
          idx_reduce++;
        } else {
          x_new_dim.push_back(x_dim[i]);
        }
      }
    } else {
      x_new_dim = x_dim;
    }

    // update x_dim
    x_dim = x_new_dim;
    std::vector<int>().swap(x_new_dim);

    std::vector<int> reduce_dim_new;
    int is_reduced = 0;
    for (auto e : reduce_dim) {
      is_reduced |= 1 << e;
    }

    std::vector<int>().swap(reduce_dim);

    for (int i = 0; i < x_dim.size(); i++) {
      if ((i == 0) || (((is_reduced >> i) ^ (is_reduced >> (i - 1))) & 1)) {
        x_new_dim.push_back(x_dim[i]);
        if ((is_reduced >> i) & 1)
          reduce_dim_new.push_back(x_new_dim.size() - 1);
      } else {
        x_new_dim[x_new_dim.size() - 1] *= x_dim[i];
      }
    }

    x_dim = x_new_dim;
    reduce_dim = reduce_dim_new;

    int x_rank = static_cast<int>(x_dim.size());
    std::set<int> left_set;

    for (int i = 0; i < x_rank; ++i) {
      left_set.insert(i);
    }

    for (auto e : reduce_dim) {
      left_set.erase(e);
    }

    left_dim.assign(left_set.begin(), left_set.end());

    // if the last dim gets involved in reduction
    reduce_last_dim = (reduce_dim.back() == x_dim.size() - 1);
  }

  // set x_strides, reduce_strides, left_strides for reduceLastDim and reduceAny
  // eg: x_dim = [8, 6], reduce_dim = [0], left_dim = [1]
  //     --SetStrides--> x_strides= [6,1], reduce_strides = [1],
  //     left_strides = [1]
  void SetStrides() {
    std::vector<int> idx_dim;
    for (int i = 0; i < x_dim.size(); i++) {
      idx_dim.push_back(i);
    }

    x_strides = details::GetDimStrides(x_dim, idx_dim);
    reduce_strides = details::GetDimStrides(x_dim, reduce_dim);
    left_strides = details::GetDimStrides(x_dim, left_dim);
    reduce_num = reduce_strides[0] * x_dim[reduce_dim[0]];

    left_num = 1;
    if (left_dim.size()) {
      left_num = left_strides[0] * x_dim[left_dim[0]];
    }
  }

  // get the reduceType
  // eg: x_dim = [8, 6] reduce_dim = [0] --> ReduceHigherDim -->reduceFirstDim
  //     x_dim = [8, 6] reduce_dim = [1] --> reduceLastDim
  //     x_dim = [8] reduce_dim = [0] --> reduceAll
  //     x_dim = [8, 6, 4, 2] reduce_dim = [0, 2] --> reduceAny
  void SetReduceType() {
    int rank = x_dim.size();
    int reduce_rank = reduce_dim.size();
    bool is_last_dim =
        (rank == 2) && (reduce_rank == 1) && (reduce_dim[0] == 1);
    if (rank == reduce_rank || is_last_dim) {
      reduce_type = static_cast<int>(ReduceType::kReduceLastDim);
    } else if (reduce_rank == 1) {
      // ReduceFirstDim and reduceSecondDim
      reduce_type = static_cast<int>(ReduceType::kReduceHigherDim);
    } else {
      reduce_type = static_cast<int>(ReduceType::kReduceAny);
    }
  }

  void SetBlockDimForReduceAny(dim3* block_dim, dim3* grid_dim) {
    constexpr int min_reduce_num_per_thread = 16;
    constexpr int max_reduce_num_per_thread = 256;
    constexpr int max_num_threads = kps::details::kReduceMaxThread;

    // set block size.
    // 1. If reduce_last_dim == true, all the threads whose threadIdx.y are same
    //    will process the reduction for one output.
    //    The number of output for one block is blockDim.y;
    // 2. If reduce_last_dim == false, different threadIdx.x will process
    //    different reduction and gets the output separately. If it is
    //    necessary, it should reduce in block y.
    //    The number of output for one block is blockDim.x;
    int block_x, block_y;
    int grid_num, reduce_num_per_thread;
    if (reduce_last_dim) {
      block_x = details::GetBlockDim(reduce_num);
      block_y = details::GetBlockDim(left_num);
      block_dim->x = block_x;
      block_dim->y =
          std::min(block_y, static_cast<int>(max_num_threads / block_dim->x));
      grid_num = details::AlignUp(left_num, block_dim->y);
      reduce_num_per_thread = details::AlignUp(reduce_num, block_dim->x);
    } else {
      block_x = details::GetBlockDim(left_num);
      block_y = details::GetBlockDim(reduce_num);
      block_dim->x = std::min(block_x, 32);
      block_dim->y =
          std::min(block_y, static_cast<int>(max_num_threads / block_dim->x));
      block_dim->x =
          std::min(block_x, static_cast<int>(max_num_threads / block_dim->y));
      grid_num = details::AlignUp(left_num, block_dim->x);
      reduce_num_per_thread = details::AlignUp(reduce_num, block_dim->y);
    }
    int device_id = platform::GetCurrentDeviceId();
    int max_mp = platform::GetCUDAMultiProcessors(device_id);
    int max_threads_per_mp =
        platform::GetCUDAMaxThreadsPerMultiProcessor(device_id);
    int max_threads = max_threads_per_mp * max_mp;
    int num_threads = block_dim->x * block_dim->y;
    int max_num_blocks = max_threads / num_threads;

    // set grid size.
    // Whether to set grid.y larger than 1, there are 3 following rules:
    // 1. The number that each thread process should no less than
    //    min_reduce_num_per_threadbut no more than max_reduce_num_per_thread;
    // 2. It should maximize the utilization of SM.
    // So we choose the minimum between input_split_num_1 and input_split_num_3
    // to make each thread process as mush data as possible. Meanwhile,
    // the number cannot be larger than max_reduce_num_per_thread, so we
    // choose the maximum between the result above and input_split_num_2.
    int input_split_num_1 =
        details::AlignUp(reduce_num_per_thread, min_reduce_num_per_thread);
    int input_split_num_2 =
        details::AlignUp(reduce_num_per_thread, max_reduce_num_per_thread);
    int input_split_num_3 = details::AlignUp(max_num_blocks, grid_num);

    grid_dim->x = grid_num;
    grid_dim->y = std::max(std::min(input_split_num_1, input_split_num_3),
                           input_split_num_2);
    // if grid.y > 1, we need launch reduce kernel again.
    if (grid_dim->y > 1) {
      should_reduce_again = true;
    }
  }

  // set block and grid for launch kernel
  // for ReduceHigherDim: if block is enough -> splite reduce_num
  //                     else init block(32, 1) grid(block_num, 1)
  // for others: block(block_num, 1) , grid(left_num, 1)
  void SetBlockDim() {
    // init
    int block_num = details::GetBlockDim(reduce_num);
    should_reduce_again = false;

    dim3 block_dim(block_num, 1);
    dim3 grid_dim(left_num, 1);
    blocking_size = reduce_num;

    if (reduce_type == ReduceType::kReduceHigherDim) {
      int last_dim_num = x_dim.back();
      // update left_num
      int grid_z = left_num / last_dim_num;
      left_num = last_dim_num;

      block_dim.z = 1;
      grid_dim.z = grid_z;

      int device_id = platform::GetCurrentDeviceId();
      int max_mp = platform::GetCUDAMultiProcessors(device_id);
      int max_threads_per_mp =
          platform::GetCUDAMaxThreadsPerMultiProcessor(device_id);
      int max_threads = max_threads_per_mp * max_mp;

      // init
      int num_block = (max_threads / left_num);

      if (num_block > 1 && reduce_num >= REDUCE_SPLIT_BOUNDARY) {
        blocking_size = details::GetLastPow2(reduce_num / num_block);

        if (blocking_size <= 1) {
          blocking_size = details::GetLastPow2(sqrt(reduce_num));
        } else if (blocking_size * 2 < reduce_num) {
          blocking_size *= 2;
        }

        should_reduce_again = true;

        block_dim.x = details::GetBlockDim(left_num);
        block_dim.y = 1;
        grid_dim.x = (left_num + block_dim.x - 1) / block_dim.x;
        grid_dim.y = (reduce_num + blocking_size - 1) / blocking_size;

      } else {
        block_dim.x = details::GetBlockDim(left_num);
        block_dim.y = 1;
        blocking_size = reduce_num;
        grid_dim.x = (left_num + block_dim.x - 1) / block_dim.x;
        grid_dim.y = 1;
      }
    } else {
      SetBlockDimForReduceAny(&block_dim, &grid_dim);
    }

    block = block_dim;
    grid = grid_dim;
  }

 public:
  std::vector<int> reduce_dims_origin;
  std::vector<int> reduce_dim;
  std::vector<int> x_dim;
  std::vector<int> left_dim;
  std::vector<int> x_strides;
  std::vector<int> left_strides;
  std::vector<int> reduce_strides;

  int reduce_type;
  int reduce_num;
  int left_num;
  int blocking_size;
  bool should_reduce_again;
  bool reduce_last_dim;

  Ty* output_data;

  dim3 block;
  dim3 grid;
};
/* size : how many colonms left have to be reduced
 * loop : how many rows data have to be reduced
 * block_size: max rows this block to reduce
 */
template <typename Tx, typename Ty, typename MPType, typename ReduceOp,
          typename TransformOp, bool IsBoundary = false>
__device__ void HigherDimDealSegment(const Tx* x, Ty* y, ReduceOp reducer,
                                     TransformOp transformer, MPType init,
                                     int reduce_num, int left_num,
                                     int block_size) {
  const int NY = 1;
  int idx = blockIdx.x * blockDim.x;
  int idy = blockIdx.y * block_size;
  // block_offset of rows
  Tx reduce_input[NY];
  MPType reduce_compute[NY];
  MPType result = init;
  // the offset of this block
  int block_offset = idy * left_num + idx + blockIdx.z * reduce_num * left_num;
  const Tx* input = x + block_offset;
  int store_offset =
      blockIdx.y * left_num + blockIdx.z * gridDim.y * left_num + idx;
  // how many columns left
  int size = left_num - idx;
  // how many rows have to be reduced
  int loop = reduce_num - idy;
  loop = loop > block_size ? block_size : loop;

  for (int loop_index = 0; loop_index < loop; loop_index += NY) {
    kps::ReadData<Tx, Tx, 1, NY, 1, IsBoundary>(
        &reduce_input[0], input + loop_index * left_num, size, NY, 1, left_num);
    kps::ElementwiseUnary<Tx, MPType, REDUCE_VEC_SIZE, 1, 1, TransformOp>(
        &reduce_compute[0], &reduce_input[0], transformer);
    kps::Reduce<MPType, NY, 1, 1, ReduceOp,
                kps::details::ReduceMode::kLocalMode>(
        &result, &reduce_compute[0], reducer, false);
  }

  Ty temp_data = static_cast<Ty>(result);
  kps::WriteData<Ty, 1, 1, 1, IsBoundary>(y + store_offset, &temp_data, size);
}

// when reduce_dim.size() == 1 and reduce_dim[0] == x_dim.size() - 1, or
// when reduce_dim.size() != 1 and reduce_dim.size() != x_dim.size(), this
// function will be used
template <typename Tx, typename Ty, typename MPType, typename ReduceOp,
          typename TransformOp, typename Calculator>
__global__ void ReduceAnyKernel(const Tx* x, Ty* y, ReduceOp reducer,
                                TransformOp transformer, MPType init,
                                int reduce_num, int left_num,
                                bool reduce_last_dim,
                                const Calculator reduce_index_calculator,
                                const Calculator left_index_calculator) {
  int input_idx, left_idx, stride;
  int block_size = 0;
  bool need_store = true;
  int tid = 0;
  // the last dim gets involved in reduction
  if (reduce_last_dim) {
    input_idx = blockIdx.y * blockDim.x;
    left_idx = blockIdx.x * blockDim.y + threadIdx.y;
    stride = gridDim.y * blockDim.x;
    block_size = blockDim.x;
    need_store = (threadIdx.x == 0) && (left_idx < left_num);
    tid = threadIdx.x;
  } else {
    input_idx = blockIdx.y * blockDim.y;
    left_idx = blockIdx.x * blockDim.x + threadIdx.x;
    stride = gridDim.y * blockDim.y;
    block_size = blockDim.y;
    need_store = (threadIdx.y == 0) && (left_idx < left_num);
    tid = threadIdx.y;
  }
  int store_offset = blockIdx.y * left_num + left_idx;
  // calculate the offset, means the addr where each thread really start.
  int input_offset = left_index_calculator(left_idx);
  const Tx* input = x + input_offset;
  MPType reduce_var = init;
  Ty store_data;

  // 1. reduce for each thread
  if (left_idx < left_num) {
    // load REDUCE_VEC_SIZE data once, and then compute
    Tx input_reg[REDUCE_VEC_SIZE];
    MPType input_compute[REDUCE_VEC_SIZE];
    int bound = reduce_num - (REDUCE_VEC_SIZE - 1) * stride;
    for (; input_idx + block_size < bound;
         input_idx += REDUCE_VEC_SIZE * stride) {
      kps::ReadDataReduce<Tx, 1, REDUCE_VEC_SIZE, 1, 1, Calculator>(
          &input_reg[0], input, input_idx, reduce_index_calculator, 1,
          reduce_num, 1, stride, reduce_last_dim);
      kps::ElementwiseUnary<Tx, MPType, REDUCE_VEC_SIZE, 1, 1, TransformOp>(
          &input_compute[0], &input_reg[0], transformer);
      kps::Reduce<MPType, REDUCE_VEC_SIZE, 1, 1, ReduceOp,
                  kps::details::ReduceMode::kLocalMode>(
          &reduce_var, &input_compute[0], reducer, reduce_last_dim);
    }

    kps::Init<MPType, REDUCE_VEC_SIZE>(&input_compute[0], init);
    kps::ReadDataReduce<Tx, 1, REDUCE_VEC_SIZE, 1, 1, Calculator, true>(
        &input_reg[0], input, input_idx, reduce_index_calculator, 1, reduce_num,
        1, stride, reduce_last_dim);
    input_idx += tid;
#pragma unroll
    for (int i = 0; i < REDUCE_VEC_SIZE; ++i) {
      if (input_idx >= reduce_num) {
        break;
      }
      input_compute[i] = static_cast<MPType>(transformer(input_reg[i]));
      input_idx += stride;
    }
    kps::Reduce<MPType, REDUCE_VEC_SIZE, 1, 1, ReduceOp,
                kps::details::ReduceMode::kLocalMode>(
        &reduce_var, &input_compute[0], reducer, reduce_last_dim);
  }

  kps::Reduce<MPType, 1, 1, 1, ReduceOp, kps::details::kGlobalMode>(
      &reduce_var, &reduce_var, reducer, reduce_last_dim);
  if (need_store) {
    y[store_offset] = static_cast<Ty>(reduce_var);
  }
}

template <typename Tx, typename Ty, typename MPType, typename ReduceOp,
          typename TransformOp>
__global__ void ReduceHigherDimKernel(const Tx* x, Ty* y, ReduceOp reducer,
                                      TransformOp transformer, MPType init,
                                      int reduce_num, int left_num,
                                      int blocking_size) {
  // when reduce_dim.size() == 1 and reduce_dim[0] != x_dim.size() - 1, this
  // function will be used
  // eg: x_dim = {nz, ny, nx}, nx != 1, axis can be 0 or 1
  //     if axis = 1 then grid.z = nz, grid.y = ny / block_size, grid.x = nx /
  //     32
  //     else grid.z = 1, grid.y = ny / block_size, grid.x = nx /32
  int idx = blockIdx.x * blockDim.x;
  int size = left_num - idx;
  if (size >= blockDim.x) {  // complete segment
    HigherDimDealSegment<Tx, Ty, MPType, ReduceOp, TransformOp>(
        x, y, reducer, transformer, init, reduce_num, left_num, blocking_size);
  } else {
    HigherDimDealSegment<Tx, Ty, MPType, ReduceOp, TransformOp, true>(
        x, y, reducer, transformer, init, reduce_num, left_num, blocking_size);
  }
}

template <typename Tx, typename Ty, typename MPType, typename ReduceOp>
static void LaunchReduceKernel(const Tx* x_data, Ty* y_data,
                               const ReduceOp& reducer, MPType init,
                               gpuStream_t stream, ReduceConfig<Ty> config) {
  using TransformOp = typename ReduceOp::Transformer;

  if (config.reduce_type == kReduceLastDim) {
    int stride_reduce = 1;
    int stride_left = config.reduce_num;
    // for higher performance
    auto reduce_index_calculator = LastDimIndexCal(stride_reduce);
    auto left_index_calculator = LastDimIndexCal(stride_left);

    ReduceAnyKernel<Tx, Ty, MPType, ReduceOp, TransformOp,
                    LastDimIndexCal><<<config.grid, config.block, 0, stream>>>(
        x_data, config.output_data, reducer, TransformOp(config.reduce_num),
        init, config.reduce_num, config.left_num, config.reduce_last_dim,
        reduce_index_calculator, left_index_calculator);

  } else {
    int reduce_rank = config.reduce_strides.size();
    int left_rank = config.left_strides.size();
    auto reduce_index_calculator =
        IndexCalculator(reduce_rank, config.reduce_dim, config.reduce_strides,
                        config.x_strides);
    auto left_index_calculator = IndexCalculator(
        left_rank, config.left_dim, config.left_strides, config.x_strides);
    ReduceAnyKernel<Tx, Ty, MPType, ReduceOp, TransformOp,
                    IndexCalculator><<<config.grid, config.block, 0, stream>>>(
        x_data, config.output_data, reducer, TransformOp(config.reduce_num),
        init, config.reduce_num, config.left_num, config.reduce_last_dim,
        reduce_index_calculator, left_index_calculator);
  }

  if (config.should_reduce_again) {
    dim3 block;
    dim3 grid;
    if (config.reduce_last_dim) {
      block = dim3(32, 1, 1);
      grid = dim3(details::AlignUp(config.left_num, 32), 1, 1);
    } else {
      block = dim3(config.block.x, 1, 1);
      grid = dim3(config.grid.x, 1, config.grid.z);
    }

    ReduceHigherDimKernel<
        Ty, Ty, MPType, ReduceOp,
        kps::details::IdentityFunctor<Ty, MPType>><<<grid, block, 0, stream>>>(
        config.output_data, y_data, reducer,
        kps::details::IdentityFunctor<Ty, MPType>(config.grid.y), init,
        config.grid.y, config.left_num, config.grid.y);
  }
}

template <typename Tx, typename Ty,
          template <typename, typename> class ReduceOp>
void TensorReduceFunctorImpl(const framework::Tensor& x, framework::Tensor* y,
                             std::vector<int> origin_reduce_dims,
                             gpuStream_t stream) {
  auto x_dim = framework::vectorize<int>(x.dims());
  auto config = ReduceConfig<Ty>(origin_reduce_dims, x_dim);
  config.Run();  // get the parameters of LaunchReduceKernel
  int numel = x.numel();
  // after config.run()
  // SetOutputData for ReduceHigherDim when should_reduce_again is true,
  // temp_output should be stored temp_data in output_data space or stored in
  // y_data;
  framework::Tensor tmp;
  auto x_data = x.data<Tx>();
  auto y_data = y->mutable_data<Ty>(x.place());

  if (config.reduce_num == 1) {
    auto out_dims = y->dims();
    framework::TensorCopy(x, y->place(), y);
    y->Resize(out_dims);
    return;
  }

  config.SetOutputData(y_data, x.place(), &tmp);
  bool use_cub_reduce = (config.reduce_num == numel) &&
                        (!std::is_same<Tx, paddle::platform::float16>::value);
  if (use_cub_reduce) {
    // launch CUB::Reduce
    using TransformOp = typename ReduceOp<Tx, Ty>::Transformer;
    auto reducer = ReduceOp<Tx, Ty>();
    cub::TransformInputIterator<Ty, TransformOp, const Tx*> trans_x(
        x_data, TransformOp(config.reduce_num));
    size_t temp_storage_bytes = 0;
    cub::DeviceReduce::Reduce(nullptr, temp_storage_bytes, trans_x, y_data,
                              config.reduce_num, reducer, reducer.initial(),
                              stream);
    framework::Tensor tmp;
    auto* temp_storage = tmp.mutable_data<uint8_t>(
        framework::make_ddim({static_cast<int64_t>(temp_storage_bytes)}),
        x.place());
    cub::DeviceReduce::Reduce(temp_storage, temp_storage_bytes, trans_x, y_data,
                              config.reduce_num, reducer, reducer.initial(),
                              stream);

    return;
  }

  using MPType = typename details::MPTypeTrait<Ty>::Type;
  auto reducer = ReduceOp<Tx, MPType>();
  // launch ReduceHigherDimKernel
  // when reduce_dim.size() == 1 and reduce_dim[0] != x_dim.size() - 1, this
  // function will be used
  // eg: x_dim = {nz, ny, nx}, nx != 1, axis can be 0 or 1
  //     if axis = 1 then grid.z = nz, grid.y = ny / block_size, grid.x = nx /
  //     32
  //     else grid.z = 1, grid.y = ny / block_size, grid.x = nx /32
  if (config.reduce_type == ReduceType::kReduceHigherDim) {
    using TransformOp = typename ReduceOp<Tx, MPType>::Transformer;

    ReduceHigherDimKernel<
        Tx, Ty, MPType, ReduceOp<Tx, MPType>,
        TransformOp><<<config.grid, config.block, 0, stream>>>(
        x_data, config.output_data, reducer, TransformOp(config.reduce_num),
        reducer.initial(), config.reduce_num, config.left_num,
        config.blocking_size);

    if (config.should_reduce_again) {
      dim3 block = dim3(config.block.x, 1, 1);
      dim3 grid = dim3(config.grid.x, 1, config.grid.z);
      ReduceHigherDimKernel<Ty, Ty, MPType, ReduceOp<Tx, MPType>,
                            kps::details::IdentityFunctor<
                                Ty, MPType>><<<grid, block, 0, stream>>>(
          config.output_data, y_data, reducer,
          kps::details::IdentityFunctor<Ty, MPType>(config.grid.y),
          reducer.initial(), config.grid.y, config.left_num, config.grid.y);
    }
    return;
  }

  // when reduce_dim.size() == 1 and reduce_dim[0] == x_dim.size() - 1, or
  // when reduce_dim.size() != 1 and reduce_dim.size() != x_dim.size(), this
  // function will be used
  LaunchReduceKernel<Tx, Ty, MPType, ReduceOp<Tx, MPType>>(
      x_data, y_data, reducer, reducer.initial(), stream, config);
}

template <typename Tx, template <typename, typename> class ReduceOp>
struct TensorReduceFunc {
  const framework::Tensor& x;
  framework::Tensor* y;
  std::vector<int> origin_reduce_dims;
  gpuStream_t stream;
  TensorReduceFunc(const framework::Tensor& x, framework::Tensor* y,
                   std::vector<int> origin_reduce_dims, gpuStream_t stream)
      : x(x), y(y), origin_reduce_dims(origin_reduce_dims), stream(stream) {}

  template <typename Ty>
  void apply() const {
    TensorReduceFunctorImpl<Tx, Ty, ReduceOp>(x, y, origin_reduce_dims, stream);
  }
};

}  // namespace operators
}  // namespace paddle