depthwise_conv_3x3.cpp

/* Copyright (c) 2018 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 "operators/math/depthwise_conv_3x3.h"
#include <arm_neon.h>
#include <vector>

namespace paddle_mobile {
namespace operators {
namespace math {
void DepthwiseConv3x3(const Tensor *input, vector<int> strides,
                      vector<int> paddings, const Tensor *filter, Tensor *bias,
                      Tensor *output, bool if_bias) {
#if __ARM_NEON
  const int batch_size = input->dims()[0];

  const int input_height = input->dims()[2];

  const int input_width = input->dims()[3];

  const int output_channels = output->dims()[1];

  const int output_height = output->dims()[2];
  const int output_width = output->dims()[3];
  const int _kernel_size = 3;
  const int stride_height = strides[0];
  const int stride_width = strides[1];
  const int padding_height = paddings[0];
  const int padding_width = paddings[1];
  const float zero = 0;
  const int input_channel_stride = input_height * input_width;
  const int output_channel_stride = output_height * output_width;
  const int filter_channel_stride = 9;

  const float *input_data = input->data<float>();
  const float *filter_data = filter->data<float>();
  if (if_bias) {
    math::expand_bias(*bias, 1, output->dims());
    output->ShareDataWith(*bias);
  }
  float *output_data = output->mutable_data<float>();

  const int input_batch_stride = output_channels * input_channel_stride;
  const int output_batch_stride = output_channels * output_channel_stride;
  const int filter_batch_stride = output_channels * output_channel_stride;
  const float *pos1, *pos2, *pos3, *filter1, *filter2, *filter3, *output_ptr;
  int hstart, wstart, hend, wend;
  float result;
  for (int i = 0; i < batch_size; ++i) {
    for (int c = 0; c < output_channels; ++c) {
      filter1 = filter_data;
      filter2 = filter1 + 3;
      filter3 = filter2 + 3;

      for (int ph = 0; ph < output_height; ph++) {
        for (int pw = 0; pw < output_width; pw++) {
          hstart = ph * stride_height - padding_height;
          wstart = pw * stride_width - padding_width;
          hend = min(hstart + _kernel_size, input_height + padding_height);
          wend = min(wstart + _kernel_size, input_width + padding_width);
          hstart = max(hstart, 0);
          wstart = max(wstart, 0);
          hend = min(hend, input_height);
          wend = min(wend, input_width);
          pos1 = input_data + hstart * input_width + wstart;
          pos2 = input_data + (hstart + 1) * input_width + wstart;
          pos3 = input_data + (hstart + 2) * input_width + wstart;
          output_ptr = output_data + ph * output_width + pw;

          if (hend - hstart != 3 || wend - wstart != 3) {
            result = 0;
            float fake_input[9] = {0};
            if (hstart == 0 && wstart == 0) {
              // 左上角
              for (int j = 0; j < 3; ++j) {
                for (int k = 0; k < 3; ++k) {
                  if (j >= 3 - hend && k >= 3 - wend) {
                    fake_input[3 * j + k] =
                        input_data[(j - (3 - hend)) * input_width + k -
                                   (3 - wend)];
                  }
                }
              }
            } else if (hstart == 0 && wend == input_width) {
              // 右上角
              for (int j = 0; j < 3; ++j) {
                for (int k = 0; k < 3; ++k) {
                  if (j >= 3 - hend && k <= input_width - wstart - 1) {
                    fake_input[3 * j + k] =
                        input_data[(j - (3 - hend)) * input_width + k + wstart];
                  }
                }
              }

            } else if (hend == input_height && wstart == 0) {
              // 左下角

              for (int j = 0; j < 3; ++j) {
                for (int k = 0; k < 3; ++k) {
                  if (j <= input_height - 1 - hstart && k >= 3 - wend) {
                    fake_input[3 * j + k] =
                        input_data[(j + hstart) * input_width + k - (3 - wend)];
                  }
                }
              }
            } else if (hend == input_height && wend == input_width) {
              // 右下角
              for (int j = 0; j < 3; ++j) {
                for (int k = 0; k < 3; ++k) {
                  if (j <= input_height - hstart - 1 &&
                      k <= input_width - wstart - 1) {
                    fake_input[3 * j + k] =
                        input_data[(j + hstart) * input_width + k + wstart];
                  }
                }
              }
            } else if (hstart == 0) {
              // 顶部
              for (int j = 0; j < 3; ++j) {
                for (int k = 0; k < 3; ++k) {
                  if (j >= 3 - hend) {
                    fake_input[3 * j + k] =
                        input_data[(j - (3 - hend)) * input_width + k + wstart];
                  }
                }
              }

            } else if (hend == input_height) {
              // 底部
              for (int j = 0; j < 3; ++j) {
                for (int k = 0; k < 3; ++k) {
                  if (j <= input_height - hstart - 1) {
                    fake_input[3 * j + k] =
                        input_data[(j + hstart) * input_width + k + wstart];
                  }
                }
              }

            } else if (wstart == 0) {
              // 左侧
              for (int j = 0; j < 3; ++j) {
                for (int k = 0; k < 3; ++k) {
                  if (k >= 3 - wend) {
                    fake_input[3 * j + k] =
                        input_data[(j + hstart) * input_width +
                                   (k - (3 - wend))];
                  }
                }
              }

            } else if (wend == input_width) {
              // 右侧
              for (int j = 0; j < 3; ++j) {
                for (int k = 0; k < 3; ++k) {
                  if (k <= input_width - wstart - 1) {
                    fake_input[3 * j + k] =
                        input_data[(j + hstart) * input_width + k + wstart];
                  }
                }
              }
            }
            for (int l = 0; l < 9; ++l) {
              result += fake_input[l] * filter1[l];
            }
            if (if_bias) {
              output_data[ph * output_width + pw] += result;
            } else {
              output_data[ph * output_width + pw] = result;
            }

          } else {
#if defined(ARMV17)
            asm volatile(

                "vld1.32  {q1}, [%[pos1]]        \n\t"
                "vld1.32  {q4}, [%[filter1]]        \n\t"
                "vmov.f32 q0,    #0.0              \n\t"

                "vld1.32  {q2}, [%[pos2]]        \n\t"
                "vld1.32  {q5}, [%[filter2]]        \n\t"
                "vmla.f32 q0, q1, q4           \n\t"

                "vld1.32  {q3}, [%[pos3]]        \n\t"
                "vld1.32  {q6}, [%[filter3]]        \n\t"

                "vmla.f32 q0, q2, q5           \n\t"
                "vmla.f32 q0, q3, q6          \n\t"

                "vmov.f32 d1[1],  %[zero]         \n\t"

                "vadd.f32  d4, d0, d1           \n\t"
                "vadd.f32  s10, s8, s9            \n\t"
                "vst1.32 {d5[0]},[%[output_ptr]]    \n\t"
                :
                : [input_data] "r"(input_data), [pos1] "r"(pos1),
                  [pos2] "r"(pos2), [pos3] "r"(pos3), [filter1] "r"(filter1),
                  [filter2] "r"(filter2), [filter3] "r"(filter3),
                  [output_ptr] "r"(output_ptr), [zero] "r"(zero)
                : "memory", "q0", "q1", "q2", "q3", "q4", "q5", "q6");
#else
            const float32x4_t data1 = vld1q_f32(pos1);
            const float32x4_t data2 = vld1q_f32(pos2);
            const float32x4_t data3 = vld1q_f32(pos3);

            const float32x4_t v_filter1 = vld1q_f32(filter1);
            const float32x4_t v_filter2 = vld1q_f32(filter2);
            const float32x4_t v_filter3 = vld1q_f32(filter3);
            float32x4_t mula = vmulq_f32(data1, v_filter1);
            mula = vmlaq_f32(mula, data2, v_filter2);
            mula = vmlaq_f32(mula, data3, v_filter3);
            float32x2_t res = vpadd_f32(
                vget_high_f32(vsetq_lane_f32(0, mula, 3)), vget_low_f32(mula));
            res = vpadd_f32(res, res);
            if (if_bias) {
              output_data[ph * output_width + pw] += vget_lane_f32(res, 0);
            } else {
              output_data[ph * output_width + pw] = vget_lane_f32(res, 0);
            }
#endif
          }
        }
      }
      input_data += input_channel_stride;
      output_data += output_channel_stride;
      filter_data += filter_channel_stride;
    }
    input_data += input_batch_stride;
    output_data += output_batch_stride;
  }
#endif
}

void DepthwiseConv3x3s1p1(const Tensor *input, const Tensor *filter,
                          Tensor *output, Tensor *bias, bool if_bias) {
  const float *input_data = input->data<float>();
  const float *filter_data = filter->data<float>();
  float *output_data = output->data<float>();
  const float *bias_data = bias->data<float>();

  const int h = static_cast<int>(input->dims()[2]);
  const int w = static_cast<int>(input->dims()[3]);
  const int l = h;

  const int batch_size = static_cast<int>(input->dims()[0]);
  const int c = static_cast<int>(input->dims()[1]);
  const int hxw = h * w;
  float32x4_t vbias = vdupq_n_f32(0.0);
  for (int b = 0; b < batch_size; ++b) {
    const float *filter_data_tmp = filter_data;

    for (int j = 0; j < c; ++j) {
      if (if_bias) {
        vbias = vdupq_n_f32(bias_data[j]);
      }

      int l_mid = l - 2;  // l=1->l_mid=-1,l=2->l_mid=0
      float w00 = filter_data_tmp[0];
      float w01 = filter_data_tmp[1];
      float w02 = filter_data_tmp[2];
      float w10 = filter_data_tmp[3];
      float w11 = filter_data_tmp[4];
      float w12 = filter_data_tmp[5];
      float w20 = filter_data_tmp[6];
      float w21 = filter_data_tmp[7];
      float w22 = filter_data_tmp[8];

      output_data[0] = w11 * input_data[0] + w12 * input_data[1] +
                       w21 * input_data[l] + w22 * input_data[l + 1];
      output_data[l - 1] = w10 * input_data[l - 2] + w11 * input_data[l - 1] +
                           w20 * input_data[2 * l - 2] +
                           w21 * input_data[2 * l - 1];
      output_data[(l - 1) * l] =
          w01 * input_data[(l - 2) * l] + w02 * input_data[(l - 2) * l + 1] +
          w11 * input_data[(l - 1) * l] + w12 * input_data[(l - 1) * l + 1];
      output_data[l * l - 1] = w00 * input_data[(l - 2) * (l + 1)] +
                               w01 * input_data[(l - 2) * (l + 1) + 1] +
                               w10 * input_data[l * l - 2] +
                               w11 * input_data[l * l - 1];
      if (if_bias) {
        output_data[0] += bias_data[j];
        output_data[l - 1] += bias_data[j];
        output_data[(l - 1) * l] += bias_data[j];
        output_data[l * l - 1] += bias_data[j];
      }
      for (int i = 1; i < l - 1; ++i) {
        output_data[i * l] =
            w01 * input_data[i * l - l] + w02 * input_data[i * l - l + 1] +
            w11 * input_data[i * l] + w12 * input_data[i * l + 1] +
            w21 * input_data[i * l + l] + w22 * input_data[i * l + l + 1];
        output_data[i * l + l - 1] = w00 * input_data[i * l + l - 1 - l - 1] +
                                     w01 * input_data[i * l + l - 1 - l] +
                                     w10 * input_data[i * l + l - 1 - 1] +
                                     w11 * input_data[i * l + l - 1] +
                                     w20 * input_data[i * l + l - 1 + l - 1] +
                                     w21 * input_data[i * l + l - 1 + l];
        if (if_bias) {
          output_data[i * l] += bias_data[j];
          output_data[i * l + l - 1] += bias_data[j];
        }
      }

      // top 1 row and bottom 1 row
      const float *input_tmp = input_data;

      float32x4_t in0, in1, in2, in3, in4, in5, in6, in7, tmp0, tmp1, tmp2,
          tmp3, tmp4, tmp5, out0;
      in0 = vld1q_f32(input_tmp);
      in2 = vld1q_f32(input_tmp + l);
      const float *input_tmp_end = input_tmp + (l - 2) * l;
      in4 = vld1q_f32(input_tmp_end);
      in6 = vld1q_f32(input_tmp_end + l);
      int c_mid = l_mid;
      auto output_ptr = output_data + 1;
      for (; c_mid > 3; c_mid -= 4) {
        in1 = vld1q_f32(input_tmp + 4);
        in3 = vld1q_f32(input_tmp + l + 4);

        tmp0 = vextq_f32(in0, in1, 1);
        tmp1 = vextq_f32(in0, in1, 2);

        tmp2 = vextq_f32(in2, in3, 1);
        tmp3 = vextq_f32(in2, in3, 2);

        out0 = vmulq_n_f32(in0, w10);
        out0 = vmlaq_n_f32(out0, tmp0, w11);
        out0 = vmlaq_n_f32(out0, tmp1, w12);
        out0 = vmlaq_n_f32(out0, in2, w20);
        out0 = vmlaq_n_f32(out0, tmp2, w21);
        out0 = vmlaq_n_f32(out0, tmp3, w22);
        out0 = vaddq_f32(out0, vbias);

        vst1q_f32(output_ptr, out0);

        in5 = vld1q_f32(input_tmp_end + 4);
        in7 = vld1q_f32(input_tmp_end + l + 4);

        tmp0 = vextq_f32(in4, in5, 1);
        tmp1 = vextq_f32(in4, in5, 2);
        tmp2 = vextq_f32(in6, in7, 1);
        tmp3 = vextq_f32(in6, in7, 2);

        out0 = vmulq_n_f32(in4, w00);
        out0 = vmlaq_n_f32(out0, tmp0, w01);
        out0 = vmlaq_n_f32(out0, tmp1, w02);
        out0 = vmlaq_n_f32(out0, in6, w10);
        out0 = vmlaq_n_f32(out0, tmp2, w11);
        out0 = vmlaq_n_f32(out0, tmp3, w12);
        out0 = vaddq_f32(out0, vbias);

        vst1q_f32(output_ptr + (l - 1) * l, out0);

        // can optimize to each 8 stride.
        input_tmp += 4;
        input_tmp_end += 4;
        output_ptr += 4;
        in0 = in1;
        in2 = in3;
        in4 = in5;
        in6 = in7;
      }

      // top right pad
      float32x4_t pad0 = vdupq_n_f32(input_data[l - 1]);
      float32x4_t pad1 = vdupq_n_f32(input_data[2 * l - 1]);

      tmp0 = vextq_f32(in0, pad0, 1);
      tmp1 = vextq_f32(in0, pad0, 2);
      tmp2 = vextq_f32(in2, pad1, 1);
      tmp3 = vextq_f32(in2, pad1, 2);

      out0 = vmulq_n_f32(in0, w10);
      out0 = vmlaq_n_f32(out0, tmp0, w11);
      out0 = vmlaq_n_f32(out0, tmp1, w12);
      out0 = vmlaq_n_f32(out0, in2, w20);
      out0 = vmlaq_n_f32(out0, tmp2, w21);
      out0 = vmlaq_n_f32(out0, tmp3, w22);
      out0 = vaddq_f32(out0, vbias);

      for (int i = 0; i < c_mid; ++i) {
        if (i == 0) {
          vst1q_lane_f32(output_ptr + i, out0, 0);
        }
        if (i == 1) {
          vst1q_lane_f32(output_ptr + i, out0, 1);
        }
        if (i == 2) {
          vst1q_lane_f32(output_ptr + i, out0, 2);
        }
      }

      // bottom right pad
      float32x4_t pad2 = vdupq_n_f32(input_data[l * l - 1 - l]);
      float32x4_t pad3 = vdupq_n_f32(input_data[l * l - 1]);

      tmp0 = vextq_f32(in4, pad2, 1);
      tmp1 = vextq_f32(in4, pad2, 2);
      tmp2 = vextq_f32(in6, pad3, 1);
      tmp3 = vextq_f32(in6, pad3, 2);

      out0 = vmulq_n_f32(in4, w00);
      out0 = vmlaq_n_f32(out0, tmp0, w01);
      out0 = vmlaq_n_f32(out0, tmp1, w02);
      out0 = vmlaq_n_f32(out0, in6, w10);
      out0 = vmlaq_n_f32(out0, tmp2, w11);
      out0 = vmlaq_n_f32(out0, tmp3, w12);
      out0 = vaddq_f32(out0, vbias);

      for (int i = 0; i < c_mid; ++i) {
        if (i == 0) {
          vst1q_lane_f32(output_ptr + (l - 1) * l + i, out0, 0);
        }
        if (i == 1) {
          vst1q_lane_f32(output_ptr + (l - 1) * l + i, out0, 1);
        }
        if (i == 2) {
          vst1q_lane_f32(output_ptr + (l - 1) * l + i, out0, 2);
        }
      }
      // mid

      for (int i = 0; i < l - 2; ++i) {
        auto output_ptr = output_data + (i + 1) * l + 1;
        input_tmp = input_data + i * l;
        auto in0_tmp = vld1q_f32(input_tmp);
        auto in2_tmp = vld1q_f32(input_tmp + l);
        auto in4_tmp = vld1q_f32(input_tmp + l + l);
        c_mid = l_mid;
        for (; c_mid > 3; c_mid -= 4) {
          auto in1_tmp = vld1q_f32(input_tmp + 4);
          auto in3_tmp = vld1q_f32(input_tmp + l + 4);
          auto in5_tmp = vld1q_f32(input_tmp + l + l + 4);

          tmp0 = vextq_f32(in0_tmp, in1_tmp, 1);
          tmp1 = vextq_f32(in0_tmp, in1_tmp, 2);
          tmp2 = vextq_f32(in2_tmp, in3_tmp, 1);
          tmp3 = vextq_f32(in2_tmp, in3_tmp, 2);
          tmp4 = vextq_f32(in4_tmp, in5_tmp, 1);
          tmp5 = vextq_f32(in4_tmp, in5_tmp, 2);

          out0 = vmulq_n_f32(in0_tmp, w00);
          out0 = vmlaq_n_f32(out0, tmp0, w01);
          out0 = vmlaq_n_f32(out0, tmp1, w02);
          out0 = vmlaq_n_f32(out0, in2_tmp, w10);
          out0 = vmlaq_n_f32(out0, tmp2, w11);
          out0 = vmlaq_n_f32(out0, tmp3, w12);
          out0 = vmlaq_n_f32(out0, in4_tmp, w20);
          out0 = vmlaq_n_f32(out0, tmp4, w21);
          out0 = vmlaq_n_f32(out0, tmp5, w22);
          out0 = vaddq_f32(out0, vbias);

          vst1q_f32(output_ptr, out0);

          output_ptr += 4;
          input_tmp += 4;
          in0_tmp = in1_tmp;
          in2_tmp = in3_tmp;
          in4_tmp = in5_tmp;
        }

        float32x4_t pad0 = vdupq_n_f32(input_data[i * l + l - 1]);
        float32x4_t pad1 = vdupq_n_f32(input_data[i * l + l - 1 + l]);
        float32x4_t pad2 = vdupq_n_f32(input_data[i * l + l - 1 + l + l]);

        tmp0 = vextq_f32(in0_tmp, pad0, 1);
        tmp1 = vextq_f32(in0_tmp, pad0, 2);
        tmp2 = vextq_f32(in2_tmp, pad1, 1);
        tmp3 = vextq_f32(in2_tmp, pad1, 2);
        tmp4 = vextq_f32(in4_tmp, pad2, 1);
        tmp5 = vextq_f32(in4_tmp, pad2, 2);

        out0 = vmulq_n_f32(in0_tmp, w00);
        out0 = vmlaq_n_f32(out0, tmp0, w01);
        out0 = vmlaq_n_f32(out0, tmp1, w02);
        out0 = vmlaq_n_f32(out0, in2_tmp, w10);
        out0 = vmlaq_n_f32(out0, tmp2, w11);
        out0 = vmlaq_n_f32(out0, tmp3, w12);
        out0 = vmlaq_n_f32(out0, in4_tmp, w20);
        out0 = vmlaq_n_f32(out0, tmp4, w21);
        out0 = vmlaq_n_f32(out0, tmp5, w22);
        out0 = vaddq_f32(out0, vbias);

        for (int i = 0; i < c_mid; ++i) {
          if (i == 0) {
            vst1q_lane_f32(output_ptr + i, out0, 0);
          }
          if (i == 1) {
            vst1q_lane_f32(output_ptr + i, out0, 1);
          }
          if (i == 2) {
            vst1q_lane_f32(output_ptr + i, out0, 2);
          }
        }
      }
      output_data += hxw;
      input_data += hxw;
      filter_data_tmp += 9;
    }
  }
}

void DepthwiseConvAddBNRelu3x3s1p1(const Tensor *input, Tensor filter,
                                   Tensor *output, Tensor *bias, bool if_bias,
                                   Tensor *new_scale, Tensor *new_bias,
                                   bool if_bn, bool if_relu) {
  const float *input_data = input->data<float>();
  const float *filter_data = filter.data<float>();
  float *output_data = output->data<float>();
  const float *bias_data = bias->data<float>();
  const float *newscale_data = new_scale->data<float>();
  const float *newbias_data = new_bias->data<float>();

  const int h = static_cast<int>(input->dims()[2]);
  const int w = static_cast<int>(input->dims()[3]);
  const int l = h;

  const int batch_size = static_cast<int>(input->dims()[0]);
  const int c = static_cast<int>(input->dims()[1]);
  const int hxw = h * w;
  float32x4_t vbias = vdupq_n_f32(0.0);
  float32x4_t vnewbias = vdupq_n_f32(0.0);
  float32x4_t vnewscale = vdupq_n_f32(1.0);
  float32x4_t vzero = vdupq_n_f32(0);

  for (int b = 0; b < batch_size; ++b) {
    const float *filter_data_tmp = filter_data;

    for (int j = 0; j < c; ++j) {
      if (if_bias) {
        vbias = vdupq_n_f32(bias_data[j]);
      }
      if (if_bn) {
        vnewbias = vdupq_n_f32(newbias_data[j]);
        vnewscale = vdupq_n_f32(newscale_data[j]);
      }
      int l_mid = l - 2;  // l=1->l_mid=-1,l=2->l_mid=0
      float w00 = filter_data_tmp[0];
      float w01 = filter_data_tmp[1];
      float w02 = filter_data_tmp[2];
      float w10 = filter_data_tmp[3];
      float w11 = filter_data_tmp[4];
      float w12 = filter_data_tmp[5];
      float w20 = filter_data_tmp[6];
      float w21 = filter_data_tmp[7];
      float w22 = filter_data_tmp[8];

      output_data[0] =
          (w11 * input_data[0] + w12 * input_data[1] + w21 * input_data[l] +
           w22 * input_data[l + 1] + bias_data[j]) *
              newscale_data[j] +
          newbias_data[j];
      output_data[l - 1] = (w10 * input_data[l - 2] + w11 * input_data[l - 1] +
                            w20 * input_data[2 * l - 2] +
                            w21 * input_data[2 * l - 1] + bias_data[j]) *
                               newscale_data[j] +
                           newbias_data[j];

      output_data[(l - 1) * l] =
          (w01 * input_data[(l - 2) * l] + w02 * input_data[(l - 2) * l + 1] +
           w11 * input_data[(l - 1) * l] + w12 * input_data[(l - 1) * l + 1] +
           bias_data[j]) *
              newscale_data[j] +
          newbias_data[j];
      output_data[l * l - 1] = (w00 * input_data[(l - 2) * (l + 1)] +
                                w01 * input_data[(l - 2) * (l + 1) + 1] +
                                w10 * input_data[l * l - 2] +
                                w11 * input_data[l * l - 1] + bias_data[j]) *
                                   newscale_data[j] +
                               newbias_data[j];
      if (if_relu) {
        output_data[0] = output_data[0] < 0 ? 0 : output_data[0];
        output_data[l - 1] = output_data[l - 1] < 0 ? 0 : output_data[l - 1];
        output_data[(l - 1) * l] =
            output_data[(l - 1) * l] < 0 ? 0 : output_data[(l - 1) * l];
        output_data[l * l - 1] =
            output_data[l * l - 1] < 0 ? 0 : output_data[l * l - 1];
      }
      for (int i = 1; i < l - 1; ++i) {
        output_data[i * l] =
            (w01 * input_data[i * l - l] + w02 * input_data[i * l - l + 1] +
             w11 * input_data[i * l] + w12 * input_data[i * l + 1] +
             w21 * input_data[i * l + l] + w22 * input_data[i * l + l + 1] +
             bias_data[j]) *
                newscale_data[j] +
            newbias_data[j];
        output_data[i * l + l - 1] =
            (w00 * input_data[i * l + l - 1 - l - 1] +
             w01 * input_data[i * l + l - 1 - l] +
             w10 * input_data[i * l + l - 1 - 1] +
             w11 * input_data[i * l + l - 1] +
             w20 * input_data[i * l + l - 1 + l - 1] +
             w21 * input_data[i * l + l - 1 + l] + bias_data[j]) *
                newscale_data[j] +
            newbias_data[j];
        if (if_relu) {
          output_data[i * l] = output_data[i * l] < 0 ? 0 : output_data[i * l];
          output_data[i * l + l - 1] =
              output_data[i * l + l - 1] < 0 ? 0 : output_data[i * l + l - 1];
        }
      }

      // top 1 row and bottom 1 row
      const float *input_tmp = input_data;

      float32x4_t in0, in1, in2, in3, in4, in5, in6, in7, tmp0, tmp1, tmp2,
          tmp3, tmp4, tmp5, out0;
      in0 = vld1q_f32(input_tmp);
      in2 = vld1q_f32(input_tmp + l);
      const float *input_tmp_end = input_tmp + (l - 2) * l;
      in4 = vld1q_f32(input_tmp_end);
      in6 = vld1q_f32(input_tmp_end + l);
      int c_mid = l_mid;
      auto output_ptr = output_data + 1;
      for (; c_mid > 3; c_mid -= 4) {
        in1 = vld1q_f32(input_tmp + 4);
        in3 = vld1q_f32(input_tmp + l + 4);

        tmp0 = vextq_f32(in0, in1, 1);
        tmp1 = vextq_f32(in0, in1, 2);

        tmp2 = vextq_f32(in2, in3, 1);
        tmp3 = vextq_f32(in2, in3, 2);

        out0 = vmulq_n_f32(in0, w10);
        out0 = vmlaq_n_f32(out0, tmp0, w11);
        out0 = vmlaq_n_f32(out0, tmp1, w12);
        out0 = vmlaq_n_f32(out0, in2, w20);
        out0 = vmlaq_n_f32(out0, tmp2, w21);
        out0 = vmlaq_n_f32(out0, tmp3, w22);
        out0 = vaddq_f32(out0, vbias);
        out0 = vmlaq_f32(vnewbias, vnewscale, out0);
        if (if_relu) {
          out0 = vmaxq_f32(out0, vzero);
        }
        vst1q_f32(output_ptr, out0);

        in5 = vld1q_f32(input_tmp_end + 4);
        in7 = vld1q_f32(input_tmp_end + l + 4);

        tmp0 = vextq_f32(in4, in5, 1);
        tmp1 = vextq_f32(in4, in5, 2);
        tmp2 = vextq_f32(in6, in7, 1);
        tmp3 = vextq_f32(in6, in7, 2);

        out0 = vmulq_n_f32(in4, w00);
        out0 = vmlaq_n_f32(out0, tmp0, w01);
        out0 = vmlaq_n_f32(out0, tmp1, w02);
        out0 = vmlaq_n_f32(out0, in6, w10);
        out0 = vmlaq_n_f32(out0, tmp2, w11);
        out0 = vmlaq_n_f32(out0, tmp3, w12);
        out0 = vaddq_f32(out0, vbias);
        out0 = vmlaq_f32(vnewbias, vnewscale, out0);
        if (if_relu) {
          out0 = vmaxq_f32(out0, vzero);
        }
        vst1q_f32(output_ptr + (l - 1) * l, out0);

        // can optimize to each 8 stride.
        input_tmp += 4;
        input_tmp_end += 4;
        output_ptr += 4;
        in0 = in1;
        in2 = in3;
        in4 = in5;
        in6 = in7;
      }

      // top right pad
      float32x4_t pad0 = vdupq_n_f32(input_data[l - 1]);
      float32x4_t pad1 = vdupq_n_f32(input_data[2 * l - 1]);

      tmp0 = vextq_f32(in0, pad0, 1);
      tmp1 = vextq_f32(in0, pad0, 2);
      tmp2 = vextq_f32(in2, pad1, 1);
      tmp3 = vextq_f32(in2, pad1, 2);

      out0 = vmulq_n_f32(in0, w10);
      out0 = vmlaq_n_f32(out0, tmp0, w11);
      out0 = vmlaq_n_f32(out0, tmp1, w12);
      out0 = vmlaq_n_f32(out0, in2, w20);
      out0 = vmlaq_n_f32(out0, tmp2, w21);
      out0 = vmlaq_n_f32(out0, tmp3, w22);
      out0 = vaddq_f32(out0, vbias);
      out0 = vmlaq_f32(vnewbias, vnewscale, out0);
      if (if_relu) {
        out0 = vmaxq_f32(out0, vzero);
      }
      for (int i = 0; i < c_mid; ++i) {
        if (i == 0) {
          vst1q_lane_f32(output_ptr + i, out0, 0);
        }
        if (i == 1) {
          vst1q_lane_f32(output_ptr + i, out0, 1);
        }
        if (i == 2) {
          vst1q_lane_f32(output_ptr + i, out0, 2);
        }
      }

      // bottom right pad
      float32x4_t pad2 = vdupq_n_f32(input_data[l * l - 1 - l]);
      float32x4_t pad3 = vdupq_n_f32(input_data[l * l - 1]);

      tmp0 = vextq_f32(in4, pad2, 1);
      tmp1 = vextq_f32(in4, pad2, 2);
      tmp2 = vextq_f32(in6, pad3, 1);
      tmp3 = vextq_f32(in6, pad3, 2);

      out0 = vmulq_n_f32(in4, w00);
      out0 = vmlaq_n_f32(out0, tmp0, w01);
      out0 = vmlaq_n_f32(out0, tmp1, w02);
      out0 = vmlaq_n_f32(out0, in6, w10);
      out0 = vmlaq_n_f32(out0, tmp2, w11);
      out0 = vmlaq_n_f32(out0, tmp3, w12);
      out0 = vaddq_f32(out0, vbias);
      out0 = vmlaq_f32(vnewbias, vnewscale, out0);
      if (if_relu) {
        out0 = vmaxq_f32(out0, vzero);
      }
      for (int i = 0; i < c_mid; ++i) {
        if (i == 0) {
          vst1q_lane_f32(output_ptr + (l - 1) * l + i, out0, 0);
        }
        if (i == 1) {
          vst1q_lane_f32(output_ptr + (l - 1) * l + i, out0, 1);
        }
        if (i == 2) {
          vst1q_lane_f32(output_ptr + (l - 1) * l + i, out0, 2);
        }
      }
      // mid

      for (int i = 0; i < l - 2; ++i) {
        auto output_ptr = output_data + (i + 1) * l + 1;
        input_tmp = input_data + i * l;
        auto in0_tmp = vld1q_f32(input_tmp);
        auto in2_tmp = vld1q_f32(input_tmp + l);
        auto in4_tmp = vld1q_f32(input_tmp + l + l);
        c_mid = l_mid;
        for (; c_mid > 3; c_mid -= 4) {
          auto in1_tmp = vld1q_f32(input_tmp + 4);
          auto in3_tmp = vld1q_f32(input_tmp + l + 4);
          auto in5_tmp = vld1q_f32(input_tmp + l + l + 4);

          tmp0 = vextq_f32(in0_tmp, in1_tmp, 1);
          tmp1 = vextq_f32(in0_tmp, in1_tmp, 2);
          tmp2 = vextq_f32(in2_tmp, in3_tmp, 1);
          tmp3 = vextq_f32(in2_tmp, in3_tmp, 2);
          tmp4 = vextq_f32(in4_tmp, in5_tmp, 1);
          tmp5 = vextq_f32(in4_tmp, in5_tmp, 2);

          out0 = vmulq_n_f32(in0_tmp, w00);
          out0 = vmlaq_n_f32(out0, tmp0, w01);
          out0 = vmlaq_n_f32(out0, tmp1, w02);
          out0 = vmlaq_n_f32(out0, in2_tmp, w10);
          out0 = vmlaq_n_f32(out0, tmp2, w11);
          out0 = vmlaq_n_f32(out0, tmp3, w12);
          out0 = vmlaq_n_f32(out0, in4_tmp, w20);
          out0 = vmlaq_n_f32(out0, tmp4, w21);
          out0 = vmlaq_n_f32(out0, tmp5, w22);
          out0 = vaddq_f32(out0, vbias);
          out0 = vmlaq_f32(vnewbias, vnewscale, out0);
          if (if_relu) {
            out0 = vmaxq_f32(out0, vzero);
          }
          vst1q_f32(output_ptr, out0);

          output_ptr += 4;
          input_tmp += 4;
          in0_tmp = in1_tmp;
          in2_tmp = in3_tmp;
          in4_tmp = in5_tmp;
        }

        float32x4_t pad0 = vdupq_n_f32(input_data[i * l + l - 1]);
        float32x4_t pad1 = vdupq_n_f32(input_data[i * l + l - 1 + l]);
        float32x4_t pad2 = vdupq_n_f32(input_data[i * l + l - 1 + l + l]);

        tmp0 = vextq_f32(in0_tmp, pad0, 1);
        tmp1 = vextq_f32(in0_tmp, pad0, 2);
        tmp2 = vextq_f32(in2_tmp, pad1, 1);
        tmp3 = vextq_f32(in2_tmp, pad1, 2);
        tmp4 = vextq_f32(in4_tmp, pad2, 1);
        tmp5 = vextq_f32(in4_tmp, pad2, 2);

        out0 = vmulq_n_f32(in0_tmp, w00);
        out0 = vmlaq_n_f32(out0, tmp0, w01);
        out0 = vmlaq_n_f32(out0, tmp1, w02);
        out0 = vmlaq_n_f32(out0, in2_tmp, w10);
        out0 = vmlaq_n_f32(out0, tmp2, w11);
        out0 = vmlaq_n_f32(out0, tmp3, w12);
        out0 = vmlaq_n_f32(out0, in4_tmp, w20);
        out0 = vmlaq_n_f32(out0, tmp4, w21);
        out0 = vmlaq_n_f32(out0, tmp5, w22);
        out0 = vaddq_f32(out0, vbias);
        out0 = vmlaq_f32(vnewbias, vnewscale, out0);
        if (if_relu) {
          out0 = vmaxq_f32(out0, vzero);
        }
        for (int i = 0; i < c_mid; ++i) {
          if (i == 0) {
            vst1q_lane_f32(output_ptr + i, out0, 0);
          }
          if (i == 1) {
            vst1q_lane_f32(output_ptr + i, out0, 1);
          }
          if (i == 2) {
            vst1q_lane_f32(output_ptr + i, out0, 2);
          }
        }
      }
      output_data += hxw;
      input_data += hxw;
      filter_data_tmp += 9;
    }
  }
}
}  // namespace math
}  // namespace operators
}  // namespace paddle_mobile