/* 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. */

#include "paddle/phi/kernels/scale_kernel.h"

#include "paddle/phi/backends/gpu/gpu_context.h"
#include "paddle/phi/core/kernel_registry.h"
#include "paddle/phi/kernels/funcs/elementwise_base.h"

namespace phi {

template <typename DataT, typename ParamT>
struct ScaleFunctor {
  ParamT bias;
  ParamT scale;
  bool bias_after_scale;

  ScaleFunctor(ParamT scale_data, ParamT bias_data, bool is_bias_after_scale)
      : bias(bias_data),
        scale(scale_data),
        bias_after_scale(is_bias_after_scale) {}

  __device__ __forceinline__ DataT operator()(const DataT x) const {
    if (bias_after_scale) {
      return static_cast<DataT>(scale * static_cast<ParamT>(x) + bias);
    } else {
      return static_cast<DataT>(scale * (static_cast<ParamT>(x) + bias));
    }
  }
};

template <typename T, typename Context>
void ScaleKernel(const Context& dev_ctx,
                 const DenseTensor& x,
                 const Scalar& scale,
                 const Scalar& bias,
                 bool bias_after_scale,
                 DenseTensor* out) {
  using MT = typename MPTypeTrait<T>::Type;
  std::vector<const DenseTensor*> inputs;
  std::vector<DenseTensor*> outputs;
  inputs.emplace_back(&x);
  outputs.emplace_back(out);
  dev_ctx.template Alloc<T>(out);
  if (x.numel() <= 0 || (!x.IsInitialized())) {
    return;
  }
  funcs::ElementwiseKernel<T>(
      dev_ctx,
      inputs,
      &outputs,
      ScaleFunctor<T, MT>(scale.to<MT>(), bias.to<MT>(), bias_after_scale));
}

template <typename T, typename Context>
void DivScaleKernel(const Context& dev_ctx,
                    const DenseTensor& x,
                    const Scalar& scale,
                    DenseTensor* out) {
  using MT = typename MPTypeTrait<T>::Type;
  std::vector<const DenseTensor*> inputs;
  std::vector<DenseTensor*> outputs;
  inputs.emplace_back(&x);
  outputs.emplace_back(out);
  dev_ctx.template Alloc<T>(out);
  if (x.numel() <= 0 || (!x.IsInitialized())) {
    return;
  }
  funcs::ElementwiseKernel<T>(
      dev_ctx,
      inputs,
      &outputs,
      ScaleFunctor<T, MT>(
          static_cast<MT>(1.0) / scale.to<MT>(), static_cast<MT>(0), true));
}

#ifdef _WIN32
INSTANCE_SCALAR_KERNEL(int, GPUContext)
INSTANCE_SCALAR_KERNEL(int64_t, GPUContext)
INSTANCE_SCALAR_KERNEL(float, GPUContext)
INSTANCE_SCALAR_KERNEL(double, GPUContext)
INSTANCE_SCALAR_KERNEL(float16, GPUContext)
INSTANCE_SCALAR_KERNEL(int16_t, GPUContext)
INSTANCE_SCALAR_KERNEL(uint8_t, GPUContext)
INSTANCE_SCALAR_KERNEL(int8_t, GPUContext)
#endif
}  // namespace phi

PD_REGISTER_KERNEL(scale,
                   GPU,
                   ALL_LAYOUT,
                   phi::ScaleKernel,
                   bool,
                   float,
                   double,
                   phi::float16,
                   phi::bfloat16,
                   phi::float8_e4m3fn,
                   phi::float8_e5m2,
                   uint8_t,
                   int8_t,
                   int16_t,
                   int,
                   int64_t,
                   phi::complex64,
                   phi::complex128) {}

PD_REGISTER_KERNEL(div_scale,
                   GPU,
                   ALL_LAYOUT,
                   phi::DivScaleKernel,
                   bool,
                   float,
                   double,
                   phi::float16,
                   phi::bfloat16,
                   phi::float8_e4m3fn,
                   phi::float8_e5m2,
                   uint8_t,
                   int8_t,
                   int16_t,
                   int,
                   int64_t,
                   phi::complex64,
                   phi::complex128) {}