// Copyright (c) 2022 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 "paddle/phi/backends/gpu/gpu_context.h"
#include "paddle/phi/core/kernel_registry.h"
#include "paddle/phi/kernels/copy_kernel.h"
#include "paddle/phi/kernels/funcs/gather.cu.h"
#include "paddle/phi/kernels/funcs/scatter.cu.h"
#include "paddle/phi/kernels/scatter_grad_kernel.h"

namespace phi {

template <typename T, typename Context>
void ScatterGradKernel(const Context &ctx,
                       const DenseTensor &index,
                       const DenseTensor &updates,
                       const DenseTensor &out_grad,
                       bool overwrite,
                       DenseTensor *x_grad,
                       DenseTensor *updates_grad) {
  auto index_type = index.dtype();
  bool index_type_match =
      index_type == phi::DataType::INT32 || index_type == phi::DataType::INT64;
  PADDLE_ENFORCE_EQ(index_type_match,
                    true,
                    phi::errors::InvalidArgument(
                        "scatter_op index holds the wrong type, it holds [%s],"
                        "but desires to be [%s] or [%s]",
                        index_type,
                        phi::DataType::INT32,
                        phi::DataType::INT64));

  if (x_grad) {
    phi::Copy(ctx, out_grad, ctx.GetPlace(), false, x_grad);
    if (index_type == phi::DataType::INT32) {
      phi::funcs::GPUScatterGradForX<T, int32_t>(ctx, index, x_grad);
    } else {
      phi::funcs::GPUScatterGradForX<T, int64_t>(ctx, index, x_grad);
    }
  }

  if (updates_grad) {
    ctx.template Alloc<T>(updates_grad);
    // Gradient by Gather: dUpdates = dO[Ids]
    if (index_type == phi::DataType::INT32) {
      phi::funcs::GPUGather<T, int32_t>(ctx, out_grad, index, updates_grad);
    } else {
      phi::funcs::GPUGather<T, int64_t>(ctx, out_grad, index, updates_grad);
    }
  }
}

}  // namespace phi

PD_REGISTER_KERNEL(scatter_grad,
                   GPU,
                   ALL_LAYOUT,
                   phi::ScatterGradKernel,
                   float,
                   double,
                   int,
                   int64_t,
                   phi::dtype::float16) {}