segment_pooling.cu

/* Copyright (c) 2020 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. */

#include <algorithm>

#include "paddle/fluid/platform/device/gpu/gpu_primitives.h"
#include "paddle/phi/backends/gpu/gpu_context.h"
#include "paddle/phi/backends/gpu/gpu_launch_config.h"
#include "paddle/phi/kernels/funcs/gather.cu.h"
#include "paddle/phi/kernels/funcs/math_function.h"
#include "paddle/phi/kernels/funcs/segment_pooling.h"

namespace phi {
namespace funcs {

using Tensor = DenseTensor;

template <typename T, typename Index, int DimTileSize>
__global__ void SegmentSumIdsKernel(const Index* segment_ids,
                                    T* summed_ids,
                                    const Index input_length_size,
                                    const Index total_stripe_count) {
  CUDA_KERNEL_LOOP(stripe_index, total_stripe_count) {
    const Index segment_offset = stripe_index;
    const Index dim_index_base = stripe_index * Index(DimTileSize);
    const Index actual_height =
        min(Index(DimTileSize), input_length_size - dim_index_base);

    Index first_segment_id = segment_ids[dim_index_base];
    Index last_segment_id = -1;
    if (dim_index_base > 0) {
      last_segment_id = segment_ids[dim_index_base - 1];
    }
    T sum = T(0);
    for (Index j = 0; j < actual_height; j++) {
      Index current_segment_id = segment_ids[dim_index_base + j];
      PADDLE_ENFORCE(current_segment_id >= last_segment_id,
                     "the segment ids should be sorted, but got "
                     "segment_ids[%d]:%d > segment_ids[%d]:%d.",
                     dim_index_base + j - 1,
                     dim_index_base + j,
                     last_segment_id,
                     current_segment_id);
      if (current_segment_id > last_segment_id) {
        for (Index interval_id = last_segment_id + 1;
             interval_id < current_segment_id;
             ++interval_id) {
          *(summed_ids + interval_id) = 0;
        }
        if (j > 0) {
          if (last_segment_id == first_segment_id) {
            paddle::platform::CudaAtomicAdd(summed_ids + last_segment_id, sum);
          } else {
            *(summed_ids + last_segment_id) = sum;
          }
          sum = T(0);
        }
      }
      sum += T(1);
      last_segment_id = current_segment_id;
    }
    paddle::platform::CudaAtomicAdd(summed_ids + last_segment_id, sum);
  }
}

template <typename T, typename Index, int DimTileSize>
__global__ void SegmentMeanKernel(const Index* segment_ids,
                                  const T* input,
                                  T* output,
                                  T* summed_ids,
                                  const Index input_length_size,
                                  const Index inner_dim_size,
                                  const Index output_length_size,
                                  const Index total_stripe_count) {
  CUDA_KERNEL_LOOP(stripe_index, total_stripe_count) {
    const Index segment_offset = stripe_index % inner_dim_size;
    const Index dim_index_base =
        stripe_index / inner_dim_size * Index(DimTileSize);
    const Index actual_height =
        min(Index(DimTileSize), input_length_size - dim_index_base);

    Index first_segment_id = segment_ids[dim_index_base];
    Index last_segment_id = -1;
    if (dim_index_base > 0) {
      last_segment_id = segment_ids[dim_index_base - 1];
    }
    T sum = T(0);
    for (Index j = 0; j < actual_height; j++) {
      Index current_segment_id = segment_ids[dim_index_base + j];
      if (current_segment_id > last_segment_id) {
        // reset the interval value which do not have corresponding ids.
        for (Index interval_id = last_segment_id + 1;
             interval_id < current_segment_id;
             ++interval_id) {
          *(output + interval_id * inner_dim_size + segment_offset) = T(0);
        }

        if (j > 0) {
          Index output_index =
              last_segment_id * inner_dim_size + segment_offset;

          if (last_segment_id == first_segment_id) {
            paddle::platform::CudaAtomicAdd(
                output + output_index, sum / *(summed_ids + last_segment_id));
          } else {
            *(output + output_index) = sum / *(summed_ids + last_segment_id);
          }
          sum = T(0);
        }
      }
      sum += input[(dim_index_base + j) * inner_dim_size + segment_offset];
      last_segment_id = current_segment_id;
    }
    Index output_index = last_segment_id * inner_dim_size + segment_offset;
    paddle::platform::CudaAtomicAdd(output + output_index,
                                    sum / *(summed_ids + last_segment_id));
  }
}

template <typename T, typename Index, typename Helper, typename Pool>
__global__ void __launch_bounds__(1024, 1) SegmentOpsKernel(
    const Index* segment_ids, const T* input, T* output, Helper h, Pool pool) {
  CUDA_KERNEL_LOOP(stripe_index, h.total_stripe_count) {
    Index segment_offset, dim_index_base, actual_height;
    Index inner_dim_size = h.inner_dim_size;
    h.calculate(stripe_index, &segment_offset, &dim_index_base, &actual_height);

    T minmax = pool.initial();
    Index first_segment_id = segment_ids[dim_index_base];
    // -1 is for the start value when interval_id = 0
    Index last_segment_id = -1;
    if (dim_index_base > 0) {
      last_segment_id = segment_ids[dim_index_base - 1];
    }

    for (Index j = 0; j < actual_height; j++) {
      Index current_segment_id = segment_ids[dim_index_base + j];
      // ensure the segment_ids is sorted.
      PADDLE_ENFORCE(current_segment_id >= last_segment_id,
                     "The segment ids should be sorted, but got "
                     "segment_ids[%d]:%d > segment_ids[%d]:%d.",
                     dim_index_base + j - 1,
                     dim_index_base + j,
                     last_segment_id,
                     current_segment_id);

      if (current_segment_id > last_segment_id) {
        // reset the interval value which do not have corresponding ids.
        for (Index interval_id = last_segment_id + 1;
             interval_id < current_segment_id;
             ++interval_id) {
          *(output + interval_id * inner_dim_size + segment_offset) = T(0);
        }
        // don't update result when j=0
        if (j > 0) {
          const Index output_index =
              last_segment_id * inner_dim_size + segment_offset;
          if (last_segment_id == first_segment_id) {
            pool.atomic(output + output_index, minmax);
          } else {
            *(output + output_index) = minmax;
          }
          minmax = pool.initial();
        }
      }
      pool.compute(
          input[(dim_index_base + j) * inner_dim_size + segment_offset],
          &minmax);
      last_segment_id = current_segment_id;
    }
    const Index output_index =
        last_segment_id * inner_dim_size + segment_offset;
    pool.atomic(output + output_index, minmax);
  }
}

template <typename T, typename Index, typename Helper>
__global__ void SegmentIndexGradKernel(const Index* segment_ids,
                                       const T* input,
                                       const T* output,
                                       const T* out_grad,
                                       T* in_grad,
                                       Helper h) {
  CUDA_KERNEL_LOOP(stripe_index, h.total_stripe_count) {
    Index segment_offset, dim_index_base, actual_height;
    h.calculate(stripe_index, &segment_offset, &dim_index_base, &actual_height);

    for (Index j = 0; j < actual_height; j++) {
      Index current_segment_id = segment_ids[dim_index_base + j];
      Index input_index =
          (dim_index_base + j) * h.inner_dim_size + segment_offset;
      Index output_index =
          current_segment_id * h.inner_dim_size + segment_offset;
      if (input[input_index] == output[output_index]) {
        in_grad[input_index] = out_grad[output_index];
      }
    }
  }
}

template <class T>
class MaxPool {
 public:
  DEVICE inline T initial() { return static_cast<T>(-FLT_MAX); }
  DEVICE inline void compute(const T& x, T* y) { *y = *y > x ? *y : x; }
  DEVICE inline T atomic(T* address, const T val) {
    return paddle::platform::CudaAtomicMax(address, val);
  }
};

template <class T>
class MinPool {
 public:
  DEVICE inline T initial() { return static_cast<T>(FLT_MAX); }
  DEVICE inline void compute(const T& x, T* y) { *y = *y < x ? *y : x; }
  DEVICE inline T atomic(T* address, const T val) {
    return paddle::platform::CudaAtomicMin(address, val);
  }
};

template <class T>
class SumPool {
 public:
  DEVICE inline T initial() { return static_cast<T>(0); }
  DEVICE inline void compute(const T& x, T* y) { *y = *y + x; }
  DEVICE inline T atomic(T* address, const T val) {
    return paddle::platform::CudaAtomicAdd(address, val);
  }
};

template <class T>
class ArrangeHelper {
 public:
  const T input_total_size;
  const T input_length_size;
  const T output_length_size;
  T inner_dim_size;
  T total_stripe_count;
  const T DimTileSize = 8;

  ArrangeHelper(T a, T b, T c)
      : input_total_size(a), input_length_size(b), output_length_size(c) {
    T input_outer_dim_num_stripe =
        (input_length_size + DimTileSize - 1) / DimTileSize;
    inner_dim_size = input_total_size / input_length_size;
    total_stripe_count = inner_dim_size * input_outer_dim_num_stripe;
  }

  DEVICE inline void calculate(T stripe_index,
                               T* segment_offset,
                               T* dim_index_base,
                               T* actual_height) {
    *segment_offset = stripe_index % inner_dim_size;
    *dim_index_base = stripe_index / inner_dim_size * DimTileSize;
    *actual_height = min(DimTileSize, input_length_size - *dim_index_base);
  }
};

template <typename T, typename Index>
void SegmentPoolCUDAGradFunctor(const phi::GPUContext& ctx,
                                const DenseTensor& input,
                                const DenseTensor& segment_ids,
                                const DenseTensor& output,
                                const DenseTensor& out_grad,
                                DenseTensor* in_grad,
                                const std::string pooltype = "SUM") {
  auto h = ArrangeHelper<Index>(
      input.numel(), segment_ids.dims()[0], output.dims()[0]);
  auto config =
      phi::backends::gpu::GetGpuLaunchConfig1D(ctx, h.total_stripe_count);
  if (pooltype == "MAX" || pooltype == "MIN") {
    SegmentIndexGradKernel<T, Index, ArrangeHelper<Index>>
        <<<config.block_per_grid.x,
           config.thread_per_block.x,
           0,
           ctx.stream()>>>(segment_ids.data<Index>(),
                           input.data<T>(),
                           output.data<T>(),
                           out_grad.data<T>(),
                           in_grad->data<T>(),
                           h);
  } else {
    PADDLE_THROW(phi::errors::InvalidArgument(
        "Unsupported segment pooling grad operation, Only MAX, MIN "
        "available, but got %s.",
        pooltype));
  }
}

template <typename T>
__global__ void SimpleDiv(T* x, const T* y, const int len, const int dim) {
  for (int i = blockIdx.x; i < len; i += gridDim.x) {
    __shared__ T y_i;
    auto base = i * dim;
    if (threadIdx.x == 0) {
      y_i = y[i];
    }
    __syncthreads();
    for (int j = threadIdx.x; j < dim; j += blockDim.x) {
      x[base + j] /= y_i;
    }
  }
}

template <typename T, typename IndexT>
class SegmentPoolFunctor<phi::GPUContext, T, IndexT> {
 public:
  void operator()(const phi::GPUContext& ctx,
                  const DenseTensor& input,
                  const DenseTensor& segment_ids,
                  DenseTensor* output,
                  DenseTensor* summed_ids = nullptr,
                  const std::string pooltype = "SUM") {
    if (pooltype == "MEAN") {
      // Sum the segment id num first
      T DimTileSize = 8;
      auto input_length_size = segment_ids.numel();
      auto total_stripe_count =
          (input_length_size + DimTileSize - 1) / DimTileSize;
      auto config =
          phi::backends::gpu::GetGpuLaunchConfig1D(ctx, total_stripe_count);
      SegmentSumIdsKernel<T, IndexT, IndexT(8)>
          <<<config.block_per_grid.x,
             config.thread_per_block.x,
             0,
             ctx.stream()>>>(segment_ids.data<IndexT>(),
                             summed_ids->data<T>(),
                             input_length_size,
                             total_stripe_count);
    }

    auto h = ArrangeHelper<IndexT>(
        input.numel(), segment_ids.dims()[0], output->dims()[0]);
    auto config =
        phi::backends::gpu::GetGpuLaunchConfig1D(ctx, h.total_stripe_count);
    if (pooltype == "MEAN") {
      SegmentMeanKernel<T, IndexT, IndexT(8)>
          <<<config.block_per_grid.x,
             config.thread_per_block.x,
             0,
             ctx.stream()>>>(segment_ids.data<IndexT>(),
                             input.data<T>(),
                             output->data<T>(),
                             summed_ids->data<T>(),
                             h.input_length_size,
                             h.inner_dim_size,
                             h.output_length_size,
                             h.total_stripe_count);
    } else if (pooltype == "SUM") {
      SumPool<T> pool;
      SegmentOpsKernel<T, IndexT, ArrangeHelper<IndexT>, SumPool<T>>
          <<<config.block_per_grid.x,
             config.thread_per_block.x,
             0,
             ctx.stream()>>>(segment_ids.data<IndexT>(),
                             input.data<T>(),
                             output->data<T>(),
                             h,
                             pool);
    } else if (pooltype == "MAX") {
      MaxPool<T> pool;
      SegmentOpsKernel<T, IndexT, ArrangeHelper<IndexT>, MaxPool<T>>
          <<<config.block_per_grid.x,
             config.thread_per_block.x,
             0,
             ctx.stream()>>>(segment_ids.data<IndexT>(),
                             input.data<T>(),
                             output->data<T>(),
                             h,
                             pool);
    } else if (pooltype == "MIN") {
      MinPool<T> pool;
      SegmentOpsKernel<T, IndexT, ArrangeHelper<IndexT>, MinPool<T>>
          <<<config.block_per_grid.x,
             config.thread_per_block.x,
             0,
             ctx.stream()>>>(segment_ids.data<IndexT>(),
                             input.data<T>(),
                             output->data<T>(),
                             h,
                             pool);
    } else {
      PADDLE_THROW(phi::errors::InvalidArgument(
          "Unsupported segment pooling operation, Only MEAN, SUM, MAX, MIN "
          "available, but got %s.",
          pooltype));
    }
  }
};

template <typename T, typename IndexT>
class SegmentPoolGradFunctor<phi::GPUContext, T, IndexT> {
 public:
  void operator()(const phi::GPUContext& dev_ctx,
                  const DenseTensor& input,
                  const DenseTensor& output,
                  const DenseTensor& out_grad,
                  const DenseTensor& segments,
                  DenseTensor* in_grad,
                  const paddle::optional<DenseTensor>& summed_ids,
                  const std::string pooltype = "SUM") {
    if (pooltype == "MAX" || pooltype == "MIN") {
      SegmentPoolCUDAGradFunctor<T, IndexT>(
          dev_ctx, input, segments, output, out_grad, in_grad, pooltype);
    } else if (pooltype == "MEAN") {
      DenseTensor mean_grad;
      mean_grad.Resize(input.dims());
      dev_ctx.template Alloc<T>(&mean_grad);
      paddle::framework::TensorCopy(
          out_grad, dev_ctx.GetPlace(), dev_ctx, &mean_grad);
      int len = output.dims()[0];
      int dim = output.numel() / len;
      auto config = phi::backends::gpu::GetGpuLaunchConfig1D(dev_ctx, len);
      SimpleDiv<T><<<config.block_per_grid.x,
                     config.thread_per_block.x,
                     0,
                     dev_ctx.stream()>>>(
          mean_grad.data<T>(), summed_ids->data<T>(), len, dim);
      phi::funcs::GPUGather<T, IndexT>(dev_ctx, mean_grad, segments, in_grad);
    } else if (pooltype == "SUM") {
      phi::funcs::GPUGather<T, IndexT>(dev_ctx, out_grad, segments, in_grad);
    } else {
      PADDLE_THROW(phi::errors::InvalidArgument(
          "Unsupported segment pooling operation, Only MEAN, SUM, MAX, MIN "
          "available, but got %s.",
          pooltype));
    }
  }
};

using GPU = phi::GPUContext;
template class SegmentPoolFunctor<GPU, float, int>;
template class SegmentPoolFunctor<GPU, float, int64_t>;
template class SegmentPoolFunctor<GPU, double, int>;
template class SegmentPoolFunctor<GPU, double, int64_t>;
template class SegmentPoolFunctor<GPU, int, int>;
template class SegmentPoolFunctor<GPU, int, int64_t>;
template class SegmentPoolFunctor<GPU, int64_t, int>;
template class SegmentPoolFunctor<GPU, int64_t, int64_t>;

template class SegmentPoolGradFunctor<GPU, float, int>;
template class SegmentPoolGradFunctor<GPU, float, int64_t>;
template class SegmentPoolGradFunctor<GPU, double, int>;
template class SegmentPoolGradFunctor<GPU, double, int64_t>;
template class SegmentPoolGradFunctor<GPU, int, int>;
template class SegmentPoolGradFunctor<GPU, int, int64_t>;
template class SegmentPoolGradFunctor<GPU, int64_t, int>;
template class SegmentPoolGradFunctor<GPU, int64_t, int64_t>;

}  // namespace funcs
}  // namespace phi