gru_cpu_kernel.h

/* Copyright (c) 2016 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 <type_traits>
#include "paddle/fluid/framework/eigen.h"
#include "paddle/fluid/operators/activation_op.h"
#include "paddle/fluid/operators/math/detail/activation_functions.h"
#include "paddle/fluid/operators/math/gru_compute.h"

namespace paddle {
namespace operators {
namespace math {
namespace detail {
using Array1 = Eigen::DSizes<int64_t, 1>;
template <typename T, int MajorType = Eigen::RowMajor,
          typename IndexType = Eigen::DenseIndex>
using EigenVector = framework::EigenVector<T, MajorType, IndexType>;

#ifndef __NVCC__

template <class OpResetOutput, typename T>
void hl_naive_gru_forward_reset_output(
    OpResetOutput op_reset_output, T *gate_value, T *reset_output_value,
    const T *prev_output_value, int frame_size, ActivationType active_gate,
    bool old_version = true, const T *reset_bias = nullptr) {
  T r_value_update_gate;
  T r_value_reset_gate;
  T r_value_reset_output;
  T r_prev_out = 0;
  T r_reset_bias = 0;
  T *update_gate = nullptr;
  T *reset_gate = nullptr;
  if (old_version) {
    update_gate = gate_value;
    reset_gate = gate_value + frame_size;
  } else {
    reset_gate = gate_value;
    update_gate = gate_value + frame_size;
  }
  for (int i = 0; i < frame_size; i++) {
    r_value_update_gate = update_gate[i];
    r_value_reset_gate = reset_gate[i];
    if (!old_version) {
      r_value_reset_output = reset_output_value[i];
      r_reset_bias = reset_bias[i];
    }
    if (prev_output_value) {
      r_prev_out = prev_output_value[i];
    }

    op_reset_output(&r_value_update_gate, &r_value_reset_gate, &r_prev_out,
                    &r_value_reset_output, active_gate, &r_reset_bias,
                    old_version);

    update_gate[i] = r_value_update_gate;
    reset_gate[i] = r_value_reset_gate;
    reset_output_value[i] = r_value_reset_output;
  }
}

template <class OpFinalOutput, typename T>
void hl_naive_gru_forward_final_output(
    OpFinalOutput op_final_output, T *gate_value, const T *prev_output_value,
    T *output_value, int frame_size, ActivationType active_node,
    bool origin_mode, bool old_version = true) {
  T r_value_update_gate;
  T r_value_frame_state;
  T r_prev_out = 0;
  T r_output;
  T *update_gate;
  if (old_version) {
    update_gate = gate_value;
  } else {
    update_gate = gate_value + frame_size;
  }
  T *frame_state = gate_value + frame_size * 2;

  for (int i = 0; i < frame_size; i++) {
    r_value_update_gate = update_gate[i];
    r_value_frame_state = frame_state[i];
    if (prev_output_value) {
      r_prev_out = prev_output_value[i];
    }

    op_final_output(&r_value_update_gate, &r_value_frame_state, &r_prev_out,
                    &r_output, active_node, origin_mode);

    frame_state[i] = r_value_frame_state;
    output_value[i] = r_output;
  }
}

template <class OpResetOutput, typename T>
void hl_avx_gru_forward_reset_output(OpResetOutput op_reset_output,
                                     T *gate_value, T *reset_output_value,
                                     const T *prev_output_value, int frame_size,
                                     ActivationType active_gate,
                                     bool old_version = true,
                                     const T *reset_bias = nullptr) {
#ifdef __AVX__
  __m256 r_value_update_gate, r_value_update_gate_last = _mm256_set1_ps(0.0f);
  __m256 r_value_reset_gate, r_value_reset_gate_last = _mm256_set1_ps(0.0f);
  __m256 r_value_reset_output;
  __m256 r_prev_out = _mm256_set1_ps(0.0f),
         r_prev_out_last = _mm256_set1_ps(0.0f);
  __m256 r_reset_bias = _mm256_set1_ps(0.0f);
  T *update_gate;
  T *reset_gate;
  if (old_version) {
    update_gate = gate_value;
    reset_gate = gate_value + frame_size;
  } else {
    reset_gate = gate_value;
    update_gate = gate_value + frame_size;
  }
  int block = 8;
  const int n = frame_size;
  const int rest = n % block;
  const int end = n - rest;
  int i = 0;

  if (rest > 0) {
    i = n - block;
    r_value_update_gate_last =
        _mm256_loadu_ps((const float *)(update_gate + i));
    r_value_reset_gate_last = _mm256_loadu_ps((const float *)(reset_gate + i));
    if (prev_output_value) {
      r_prev_out_last = _mm256_loadu_ps((const float *)(prev_output_value + i));
    }
  }

  for (i = 0; i < end; i += block) {
    r_value_update_gate = _mm256_loadu_ps((const float *)(update_gate + i));
    r_value_reset_gate = _mm256_loadu_ps((const float *)(reset_gate + i));
    if (prev_output_value) {
      r_prev_out = _mm256_loadu_ps((const float *)(prev_output_value + i));
    }
    if (!old_version) {
      r_reset_bias = _mm256_loadu_ps((const float *)(reset_bias + i));
      r_value_reset_output =
          _mm256_loadu_ps((const float *)(reset_output_value + i));
    }

    op_reset_output(&r_value_update_gate, &r_value_reset_gate, &r_prev_out,
                    &r_value_reset_output, active_gate, &r_reset_bias,
                    old_version);

    _mm256_storeu_ps(reinterpret_cast<float *>(update_gate + i),
                     r_value_update_gate);
    _mm256_storeu_ps(reinterpret_cast<float *>(reset_gate + i),
                     r_value_reset_gate);
    _mm256_storeu_ps(reinterpret_cast<float *>(reset_output_value + i),
                     r_value_reset_output);
  }

  if (rest > 0) {
    i = n - block;

    op_reset_output(&r_value_update_gate_last, &r_value_reset_gate_last,
                    &r_prev_out_last, &r_value_reset_output, active_gate,
                    &r_reset_bias, old_version);

    _mm256_storeu_ps(reinterpret_cast<float *>(update_gate + i),
                     r_value_update_gate_last);
    _mm256_storeu_ps(reinterpret_cast<float *>(reset_gate + i),
                     r_value_reset_gate_last);
    _mm256_storeu_ps(reinterpret_cast<float *>(reset_output_value + i),
                     r_value_reset_output);
  }
#endif
}

template <class OpFinalOutput, typename T>
void hl_avx_gru_forward_final_output(OpFinalOutput op_final_output,
                                     T *gate_value, const T *prev_output_value,
                                     T *output_value, int frame_size,
                                     ActivationType active_node,
                                     bool origin_mode,
                                     bool old_version = true) {
#ifdef __AVX__
  __m256 r_value_update_gate, r_value_update_gate_last = _mm256_set1_ps(0.0f);
  __m256 r_value_frame_state, r_value_frame_state_last = _mm256_set1_ps(0.0f);
  __m256 r_prev_out = _mm256_set1_ps(0.0f),
         r_prev_out_last = _mm256_set1_ps(0.0f);
  __m256 r_output;
  T *update_gate;
  if (old_version) {
    update_gate = gate_value;
  } else {
    update_gate = gate_value + frame_size;
  }

  T *frame_state = gate_value + frame_size * 2;
  int block = 8;
  const int n = frame_size;
  const int rest = n % block;
  const int end = n - rest;
  int i = 0;

  if (rest > 0) {
    i = n - block;
    r_value_update_gate_last =
        _mm256_loadu_ps((const float *)(update_gate + i));
    r_value_frame_state_last =
        _mm256_loadu_ps((const float *)(frame_state + i));
    if (prev_output_value) {
      r_prev_out_last = _mm256_loadu_ps((const float *)(prev_output_value + i));
    }
  }

  for (i = 0; i < end; i += block) {
    r_value_update_gate = _mm256_loadu_ps((const float *)(update_gate + i));
    r_value_frame_state = _mm256_loadu_ps((const float *)(frame_state + i));
    if (prev_output_value) {
      r_prev_out = _mm256_loadu_ps((const float *)(prev_output_value + i));
    }

    op_final_output(&r_value_update_gate, &r_value_frame_state, &r_prev_out,
                    &r_output, active_node, origin_mode);

    _mm256_storeu_ps(reinterpret_cast<float *>(frame_state + i),
                     r_value_frame_state);
    _mm256_storeu_ps(reinterpret_cast<float *>(output_value + i), r_output);
  }

  if (rest > 0) {
    i = n - block;
    op_final_output(&r_value_update_gate_last, &r_value_frame_state_last,
                    &r_prev_out_last, &r_output, active_node, origin_mode);

    _mm256_storeu_ps(reinterpret_cast<float *>(frame_state + i),
                     r_value_frame_state_last);
    _mm256_storeu_ps(reinterpret_cast<float *>(output_value + i), r_output);
  }

#endif
}

template <typename T>
inline void forward_reset_outputV2(const platform::CPUDeviceContext &context,
                                   GRUMetaValue<T> value, int frame_size) {
  auto &place = *context.eigen_device();
  auto value_reset_gate =
      typename EigenVector<T>::Type(value.gate_value, Array1(frame_size));
  auto value_update_gate = typename EigenVector<T>::Type(
      value.gate_value + frame_size, Array1(frame_size));
  auto value_reset_output = typename EigenVector<T>::Type(
      value.reset_output_value, Array1(frame_size));
  auto value_reset_bias =
      typename EigenVector<T>::ConstType(value.reset_bias, Array1(frame_size));
  SigmoidFunctor<T>()(place, value_reset_gate, value_reset_gate);
  SigmoidFunctor<T>()(place, value_update_gate, value_update_gate);
  value_reset_output.device(place) =
      (value_reset_output + value_reset_bias) * value_reset_gate;
}

template <class OpResetOutput, typename T>
inline void forward_reset_output(
    OpResetOutput op_reset_output, GRUMetaValue<T> value, int frame_size,
    int batch_size, ActivationType active_gate, bool old_version = true,
    const platform::CPUDeviceContext *context = nullptr) {
  for (int b = 0; b < batch_size; b++) {
    if (!old_version) {
      // use eigen
      forward_reset_outputV2(*context, value, frame_size);
    } else {
      if (OpResetOutput::avx && (frame_size & static_cast<int>(8 - 1)) &&
          (sizeof(T) == 4)) {
        hl_avx_gru_forward_reset_output(
            op_reset_output, value.gate_value, value.reset_output_value,
            value.prev_out_value, frame_size, active_gate, old_version,
            value.reset_bias);
      } else {
        hl_naive_gru_forward_reset_output(
            op_reset_output, value.gate_value, value.reset_output_value,
            value.prev_out_value, frame_size, active_gate, old_version,
            value.reset_bias);
      }
    }
    value.gate_value += frame_size * 3;
    value.reset_output_value += frame_size;
    if (value.prev_out_value) {
      value.prev_out_value += frame_size;
    }
  }
}

template <typename T>
inline void forward_final_outputV2(const platform::CPUDeviceContext &context,
                                   GRUMetaValue<T> value, int frame_size) {
  auto &place = *context.eigen_device();
  auto value_update_gate = typename EigenVector<T>::Type(
      value.gate_value + frame_size, Array1(frame_size));
  auto value_frame_state = typename EigenVector<T>::Type(
      value.gate_value + 2 * frame_size, Array1(frame_size));
  auto value_output =
      typename EigenVector<T>::Type(value.output_value, Array1(frame_size));
  TanhFunctor<T>()(place, value_frame_state, value_frame_state);
  value_output.device(place) =
      (static_cast<T>(1.0) - value_update_gate) * value_frame_state;
  if (value.prev_out_value) {
    auto value_prev_out = typename EigenVector<T>::ConstType(
        value.prev_out_value, Array1(frame_size));
    value_output.device(place) =
        value_output + value_update_gate * value_prev_out;
  }
}

template <class OpFinalOutput, typename T>
inline void forward_final_output(
    OpFinalOutput op_final_output, GRUMetaValue<T> value, int frame_size,
    int batch_size, ActivationType active_node, bool origin_mode,
    bool old_version = true,
    const platform::CPUDeviceContext *context = nullptr) {
  for (int b = 0; b < batch_size; b++) {
    if (!old_version) {
      // eigen
      forward_final_outputV2(*context, value, frame_size);
    } else {
      if (OpFinalOutput::avx && (frame_size & static_cast<int>(8 - 1)) &&
          (sizeof(T) == 4)) {
        hl_avx_gru_forward_final_output(op_final_output, value.gate_value,
                                        value.prev_out_value,
                                        value.output_value, frame_size,
                                        active_node, origin_mode, old_version);
      } else {
        hl_naive_gru_forward_final_output(
            op_final_output, value.gate_value, value.prev_out_value,
            value.output_value, frame_size, active_node, origin_mode,
            old_version);
      }
    }
    value.gate_value += frame_size * 3;
    value.output_value += frame_size;
    if (value.prev_out_value) {
      value.prev_out_value += frame_size;
    }
  }
}

template <class OpStateGrad, typename T>
void hl_naive_gru_backward_state_grad(OpStateGrad op_state_grad, T *gate_value,
                                      T *gate_grad, const T *prev_out_value,
                                      T *prev_out_grad, T *output_grad,
                                      int frame_size,
                                      ActivationType active_node,
                                      bool origin_mode) {
  T r_update_gate_value;
  T r_update_gate_grad;
  T r_frame_state_value;
  T r_frame_state_grad;
  T r_out_grad;
  T r_prev_out_value = 0;
  T r_prev_out_grad = 0;
  T *update_gate_value = gate_value;
  T *update_gate_grad = gate_grad;
  T *frame_state_value = gate_value + frame_size * 2;
  T *frame_state_grad = gate_grad + frame_size * 2;

  for (int i = 0; i < frame_size; i++) {
    r_update_gate_value = update_gate_value[i];
    r_frame_state_value = frame_state_value[i];
    r_out_grad = output_grad[i];
    if (prev_out_value) {
      r_prev_out_value = prev_out_value[i];
    }
    if (prev_out_grad) {
      r_prev_out_grad = prev_out_grad[i];
    }

    op_state_grad(&r_update_gate_value, &r_update_gate_grad,
                  &r_frame_state_value, &r_frame_state_grad, &r_prev_out_value,
                  &r_prev_out_grad, &r_out_grad, active_node, origin_mode);

    update_gate_grad[i] = r_update_gate_grad;
    frame_state_grad[i] = r_frame_state_grad;
    if (prev_out_grad) {
      prev_out_grad[i] = r_prev_out_grad;
    }
  }
}

template <class OpResetGrad, typename T>
void hl_naive_gru_backward_reset_grad(OpResetGrad op_reset_grad, T *gate_value,
                                      T *gate_grad, const T *prev_out_value,
                                      T *prev_out_grad, T *reset_output_grad,
                                      int frame_size,
                                      ActivationType active_gate) {
  T r_update_gate_value;
  T r_update_gate_grad;
  T r_reset_gate_value;
  T r_reset_gate_grad;
  T r_reset_output_grad = 0;
  T r_prev_out_value = 0;
  T r_prev_out_grad = 0;
  T *update_gate_value = gate_value;
  T *update_gate_grad = gate_grad;
  T *reset_gate_value = gate_value + frame_size;
  T *reset_gate_grad = gate_grad + frame_size;

  for (int i = 0; i < frame_size; i++) {
    r_update_gate_value = update_gate_value[i];
    r_update_gate_grad = update_gate_grad[i];
    r_reset_gate_value = reset_gate_value[i];

    if (prev_out_value && prev_out_grad) {
      r_reset_output_grad = reset_output_grad[i];
    }
    if (prev_out_value) {
      r_prev_out_value = prev_out_value[i];
    }
    if (prev_out_grad) {
      r_prev_out_grad = prev_out_grad[i];
    }

    op_reset_grad(&r_update_gate_value, &r_update_gate_grad,
                  &r_reset_gate_value, &r_reset_gate_grad, &r_prev_out_value,
                  &r_prev_out_grad, &r_reset_output_grad, active_gate);

    update_gate_grad[i] = r_update_gate_grad;
    reset_gate_grad[i] = r_reset_gate_grad;
    if (prev_out_grad) {
      prev_out_grad[i] = r_prev_out_grad;
    }
  }
}

template <class OpStateGrad, typename T>
void hl_avx_gru_backward_state_grad(OpStateGrad op_state_grad, T *gate_value,
                                    T *gate_grad, const T *prev_out_value,
                                    T *prev_out_grad, T *output_grad,
                                    int frame_size, ActivationType active_node,
                                    bool origin_mode) {
#ifdef __AVX__
  __m256 r_update_gate_value;
  __m256 r_update_gate_grad;
  __m256 r_frame_state_value;
  __m256 r_frame_state_grad;
  __m256 r_out_grad;
  __m256 r_prev_out_value = _mm256_set1_ps(0.0f);
  __m256 r_prev_out_grad = _mm256_set1_ps(0.0f);
  __m256 *update_gate_value = reinterpret_cast<__m256 *>(gate_value);
  __m256 *update_gate_grad = reinterpret_cast<__m256 *>(gate_grad);
  __m256 *frame_state_value =
      reinterpret_cast<__m256 *>(gate_value + frame_size * 2);
  __m256 *frame_state_grad =
      reinterpret_cast<__m256 *>(gate_grad + frame_size * 2);

  for (int i = 0; i < frame_size / 8; i++) {
    r_update_gate_value = update_gate_value[i];
    r_frame_state_value = frame_state_value[i];
    r_out_grad = (reinterpret_cast<__m256 *>(output_grad))[i];
    if (prev_out_value) {
      r_prev_out_value = (reinterpret_cast<const __m256 *>(prev_out_value))[i];
    }
    if (prev_out_grad) {
      r_prev_out_grad = (reinterpret_cast<__m256 *>(prev_out_grad))[i];
    }

    op_state_grad(&r_update_gate_value, &r_update_gate_grad,
                  &r_frame_state_value, &r_frame_state_grad, &r_prev_out_value,
                  &r_prev_out_grad, &r_out_grad, active_node, origin_mode);

    update_gate_grad[i] = r_update_gate_grad;
    frame_state_grad[i] = r_frame_state_grad;
    if (prev_out_grad) {
      (reinterpret_cast<__m256 *>(prev_out_grad))[i] = r_prev_out_grad;
    }
  }
#endif
}

template <class OpResetGrad, typename T>
void hl_avx_gru_backward_reset_grad(OpResetGrad op_reset_grad, T *gate_value,
                                    T *gate_grad, const T *prev_out_value,
                                    T *prev_out_grad, T *reset_output_grad,
                                    int frame_size,
                                    ActivationType active_gate) {
#ifdef __AVX__
  __m256 r_update_gate_value;
  __m256 r_update_gate_grad;
  __m256 r_reset_gate_value;
  __m256 r_reset_gate_grad;
  __m256 r_reset_output_grad = _mm256_set1_ps(0.0f);
  __m256 r_prev_out_value = _mm256_set1_ps(0.0f);
  __m256 r_prev_out_grad = _mm256_set1_ps(0.0f);
  __m256 *update_gate_value = reinterpret_cast<__m256 *>(gate_value);
  __m256 *update_gate_grad = reinterpret_cast<__m256 *>(gate_grad);
  __m256 *reset_gate_value =
      reinterpret_cast<__m256 *>(gate_value + frame_size);
  __m256 *reset_gate_grad = reinterpret_cast<__m256 *>(gate_grad + frame_size);

  for (int i = 0; i < frame_size / 8; i++) {
    r_update_gate_value = update_gate_value[i];
    r_update_gate_grad = update_gate_grad[i];
    r_reset_gate_value = reset_gate_value[i];

    if (prev_out_value && prev_out_grad) {
      r_reset_output_grad = (reinterpret_cast<__m256 *>(reset_output_grad))[i];
    }
    if (prev_out_value) {
      r_prev_out_value = (reinterpret_cast<const __m256 *>(prev_out_value))[i];
    }
    if (prev_out_grad) {
      r_prev_out_grad = (reinterpret_cast<__m256 *>(prev_out_grad))[i];
    }

    op_reset_grad(&r_update_gate_value, &r_update_gate_grad,
                  &r_reset_gate_value, &r_reset_gate_grad, &r_prev_out_value,
                  &r_prev_out_grad, &r_reset_output_grad, active_gate);

    update_gate_grad[i] = r_update_gate_grad;
    reset_gate_grad[i] = r_reset_gate_grad;
    if (prev_out_grad) {
      (reinterpret_cast<__m256 *>(prev_out_grad))[i] = r_prev_out_grad;
    }
  }
#endif
}

template <class OpGruGrad, typename T>
inline void hl_naive_gru_backward(OpGruGrad op_gru_grad, T *gate_value,
                                  T *gate_grad, const T *prev_out_value,
                                  T *prev_out_grad, T *reset_output_value,
                                  T *reset_output_grad, T *output_grad,
                                  int frame_size, ActivationType active_node,
                                  ActivationType active_gate) {
  T r_value_reset_gate;
  T r_grad_reset_gate;
  T r_value_update_gate;
  T r_grad_update_gate;
  T r_value_frame_state;
  T r_grad_frame_state;
  T r_value_prev_out = 0;
  T r_grad_prev_out = 0;
  T r_grad_output;
  T r_value_reset_output;
  T r_grad_reset_output = 0;
  T *reset_gate_value = gate_value;
  T *reset_gate_grad = gate_grad;
  T *update_gate_value = gate_value + frame_size;
  T *update_gate_grad = gate_grad + frame_size;
  T *frame_state_value = gate_value + 2 * frame_size;
  T *frame_state_grad = gate_grad + 2 * frame_size;

  for (int i = 0; i < frame_size; ++i) {
    r_value_reset_gate = reset_gate_value[i];
    r_grad_reset_gate = reset_gate_grad[i];
    r_value_update_gate = update_gate_value[i];
    r_grad_update_gate = update_gate_grad[i];
    r_value_frame_state = frame_state_value[i];
    r_grad_frame_state = frame_state_grad[i];
    if (prev_out_value) {
      r_value_prev_out = prev_out_value[i];
    }
    if (prev_out_grad) {
      r_grad_prev_out = prev_out_grad[i];
    }
    r_grad_output = output_grad[i];
    r_value_reset_output = reset_output_value[i];
    if (prev_out_value && prev_out_grad) {
      r_grad_reset_output = reset_output_grad[i];
    }

    op_gru_grad(&r_value_reset_gate, &r_grad_reset_gate, &r_value_update_gate,
                &r_grad_update_gate, &r_value_frame_state, &r_grad_frame_state,
                &r_value_prev_out, &r_grad_prev_out, &r_grad_output,
                &r_value_reset_output, &r_grad_reset_output, active_node,
                active_gate);

    reset_gate_grad[i] = r_grad_reset_gate;
    update_gate_grad[i] = r_grad_update_gate;
    frame_state_grad[i] = r_grad_frame_state;
    if (prev_out_grad) {
      prev_out_grad[i] = r_grad_prev_out;
    }
    if (prev_out_value && prev_out_grad) {
      reset_output_grad[i] = r_grad_reset_output;
    }
  }
}

template <class OpGruGrad, typename T>
inline void hl_avx_gru_backward(OpGruGrad op_gru_grad, T *gate_value,
                                T *gate_grad, const T *prev_out_value,
                                T *prev_out_grad, T *reset_output_value,
                                T *reset_output_grad, T *output_grad,
                                int frame_size, ActivationType active_node,
                                ActivationType active_gate) {
#ifdef __AVX__
  __m256 r_value_reset_gate;
  __m256 r_grad_reset_gate;
  __m256 r_value_update_gate;
  __m256 r_grad_update_gate;
  __m256 r_value_frame_state;
  __m256 r_grad_frame_state;
  __m256 r_value_prev_out = _mm256_set1_ps(0.0f);
  __m256 r_grad_prev_out = _mm256_set1_ps(0.0f);
  __m256 r_grad_output;
  __m256 r_value_reset_output;
  __m256 r_grad_reset_output = _mm256_set1_ps(0.0f);
  __m256 *reset_gate_value = reinterpret_cast<__m256 *>(gate_value);
  __m256 *reset_gate_grad = reinterpret_cast<__m256 *>(gate_grad);
  __m256 *update_gate_value =
      reinterpret_cast<__m256 *>(gate_value + frame_size);
  __m256 *update_gate_grad = reinterpret_cast<__m256 *>(gate_grad + frame_size);
  __m256 *frame_state_value =
      reinterpret_cast<__m256 *>(gate_value + 2 * frame_size);
  __m256 *frame_state_grad =
      reinterpret_cast<__m256 *>(gate_grad + 2 * frame_size);

  for (int i = 0; i < frame_size / 8; ++i) {
    r_value_reset_gate = reset_gate_value[i];
    r_grad_reset_gate = reset_gate_grad[i];
    r_value_update_gate = update_gate_value[i];
    r_grad_update_gate = update_gate_grad[i];
    r_value_frame_state = frame_state_value[i];
    r_grad_frame_state = frame_state_grad[i];
    if (prev_out_value) {
      r_value_prev_out = (reinterpret_cast<const __m256 *>(prev_out_value))[i];
    }
    if (prev_out_grad) {
      r_grad_prev_out = (reinterpret_cast<__m256 *>(prev_out_grad))[i];
    }
    r_grad_output = (reinterpret_cast<__m256 *>(output_grad))[i];
    r_value_reset_output = (reinterpret_cast<__m256 *>(reset_output_value))[i];
    if (prev_out_value && prev_out_grad) {
      r_grad_reset_output = (reinterpret_cast<__m256 *>(reset_output_grad))[i];
    }

    op_gru_grad(&r_value_reset_gate, &r_grad_reset_gate, &r_value_update_gate,
                &r_grad_update_gate, &r_value_frame_state, &r_grad_frame_state,
                &r_value_prev_out, &r_grad_prev_out, &r_grad_output,
                &r_value_reset_output, &r_grad_reset_output, active_node,
                active_gate);

    reset_gate_grad[i] = r_grad_reset_gate;
    update_gate_grad[i] = r_grad_update_gate;
    frame_state_grad[i] = r_grad_frame_state;
    if (prev_out_grad) {
      (reinterpret_cast<__m256 *>(prev_out_grad))[i] = r_grad_prev_out;
    }
    if (prev_out_value && prev_out_grad) {
      (reinterpret_cast<__m256 *>(reset_output_grad))[i] = r_grad_reset_output;
    }
  }
#endif
}

template <class OpStateGrad, typename T>
inline void backward_state_grad(OpStateGrad op_state_grad,
                                GRUMetaValue<T> value, GRUMetaGrad<T> grad,
                                int frame_size, int batch_size,
                                ActivationType active_node, bool origin_mode) {
  for (int b = 0; b < batch_size; b++) {
    if (OpStateGrad::avx && !(frame_size & (8 - 1)) && (sizeof(T) == 4)) {
      hl_avx_gru_backward_state_grad(op_state_grad, value.gate_value,
                                     grad.gate_grad, value.prev_out_value,
                                     grad.prev_out_grad, grad.output_grad,
                                     frame_size, active_node, origin_mode);
    } else {
      hl_naive_gru_backward_state_grad(op_state_grad, value.gate_value,
                                       grad.gate_grad, value.prev_out_value,
                                       grad.prev_out_grad, grad.output_grad,
                                       frame_size, active_node, origin_mode);
    }

    value.gate_value += frame_size * 3;
    if (value.prev_out_value) {
      value.prev_out_value += frame_size;
    }

    grad.gate_grad += frame_size * 3;
    grad.output_grad += frame_size;
    if (grad.prev_out_grad) {
      grad.prev_out_grad += frame_size;
    }
  }
}

template <class OpResetGrad, typename T>
inline void backward_reset_grad(OpResetGrad op_reset_grad,
                                GRUMetaValue<T> value, GRUMetaGrad<T> grad,
                                int frame_size, int batch_size,
                                ActivationType active_gate) {
  for (int b = 0; b < batch_size; b++) {
    if (OpResetGrad::avx && !(frame_size & (8 - 1)) && (sizeof(T) == 4)) {
      hl_avx_gru_backward_reset_grad(
          op_reset_grad, value.gate_value, grad.gate_grad, value.prev_out_value,
          grad.prev_out_grad, grad.reset_output_grad, frame_size, active_gate);
    } else {
      hl_naive_gru_backward_reset_grad(
          op_reset_grad, value.gate_value, grad.gate_grad, value.prev_out_value,
          grad.prev_out_grad, grad.reset_output_grad, frame_size, active_gate);
    }

    value.gate_value += frame_size * 3;
    if (value.prev_out_value) {
      value.prev_out_value += frame_size;
    }

    grad.gate_grad += frame_size * 3;
    grad.reset_output_grad += frame_size;
    if (grad.prev_out_grad) {
      grad.prev_out_grad += frame_size;
    }
  }
}

template <typename T>
inline void gru_backward(const platform::CPUDeviceContext &context,
                         GRUMetaValue<T> value, GRUMetaGrad<T> grad,
                         int frame_size) {
  auto &place = *context.eigen_device();

  auto value_reset_gate =
      typename EigenVector<T>::Type(value.gate_value, Array1(frame_size));
  auto grad_reset_gate =
      typename EigenVector<T>::Type(grad.gate_grad, Array1(frame_size));
  auto value_update_gate = typename EigenVector<T>::Type(
      value.gate_value + frame_size, Array1(frame_size));
  auto grad_update_gate = typename EigenVector<T>::Type(
      grad.gate_grad + frame_size, Array1(frame_size));
  auto value_frame_state = typename EigenVector<T>::Type(
      value.gate_value + frame_size * 2, Array1(frame_size));
  auto grad_frame_state = typename EigenVector<T>::Type(
      grad.gate_grad + frame_size * 2, Array1(frame_size));

  auto grad_output =
      typename EigenVector<T>::Type(grad.output_grad, Array1(frame_size));
  auto value_reset_output = typename EigenVector<T>::Type(
      value.reset_output_value, Array1(frame_size));
  auto grad_reset_output =
      typename EigenVector<T>::Type(grad.reset_output_grad, Array1(frame_size));

  if (value.prev_out_value) {
    auto value_prev_out = typename EigenVector<T>::ConstType(
        value.prev_out_value, Array1(frame_size));
    SigmoidGradFunctor<T>()(place, 1 /*useless*/, value_update_gate,
                            (value_prev_out - value_frame_state) * grad_output,
                            grad_update_gate);
  } else {
    SigmoidGradFunctor<T>()(
        place, 1 /*useless*/, value_update_gate,
        static_cast<T>(-1) * value_frame_state * grad_output, grad_update_gate);
  }
  if (grad.prev_out_grad) {
    auto grad_prev_out =
        typename EigenVector<T>::Type(grad.prev_out_grad, Array1(frame_size));
    grad_prev_out.device(place) =
        grad_prev_out + grad_output * value_update_gate;
  }
  TanhGradFunctor<T>()(place, 1 /*useless*/, value_frame_state,
                       grad_output * (static_cast<T>(1.0) - value_update_gate),
                       grad_frame_state);
  SigmoidGradFunctor<T>()(
      place, 1 /*useless*/, value_reset_gate,
      value_reset_output / value_reset_gate * grad_frame_state,
      grad_reset_gate);
  if (value.prev_out_value && grad.prev_out_grad) {
    grad_reset_output.device(place) = value_reset_gate * grad_frame_state;
  }
}

template <class OpGruGrad, typename T>
inline void cpu_gru_backward(const platform::CPUDeviceContext &context,
                             OpGruGrad op_gru_grad, GRUMetaValue<T> value,
                             GRUMetaGrad<T> grad, int frame_size,
                             int batch_size, ActivationType active_node,
                             ActivationType active_gate) {
  for (int b = 0; b < batch_size; ++b) {
    // eigen
    gru_backward(context, value, grad, frame_size);

    value.gate_value += frame_size * 3;
    value.reset_output_value += frame_size;
    if (value.prev_out_value) {
      value.prev_out_value += frame_size;
    }

    grad.gate_grad += frame_size * 3;
    grad.output_grad += frame_size;
    grad.reset_output_grad += frame_size;
    if (grad.prev_out_grad) {
      grad.prev_out_grad += frame_size;
    }
  }
}

#endif

}  // namespace detail
}  // namespace math
}  // namespace operators
}  // namespace paddle