conv_common.cpp 4.2 KB
Newer Older
Z
zhaojiaying01 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13
/* 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. */
朔-望's avatar
朔-望 已提交
14

H
backup  
hjchen2 已提交
15 16
#include "operators/kernel/arm/convolution/conv_common.h"
#include "operators/math/winograd/winograd_transform.h"
朔-望's avatar
朔-望 已提交
17 18

namespace paddle_mobile {
朔-望's avatar
朔-望 已提交
19 20
namespace operators {

H
backup  
hjchen2 已提交
21
void InitBaseConvKernel(ConvParam<CPU> *param) {
22 23
  bool conv3x3 = param->Filter()->dims()[2] == param->Filter()->dims()[3] &&
                 param->Filter()->dims()[2] == 3;
24 25
  bool conv5x5 = param->Filter()->dims()[2] == param->Filter()->dims()[3] &&
                 param->Filter()->dims()[2] == 5;
26 27
  bool depth3x3 = conv3x3 && param->Groups() == param->Input()->dims()[1] &&
                  param->Input()->dims()[1] == param->Output()->dims()[1];
H
backup  
hjchen2 已提交
28

29 30
  bool depth5x5 = conv5x5 && param->Groups() == param->Input()->dims()[1] &&
                  param->Input()->dims()[1] == param->Output()->dims()[1];
31

32
  if (param->Filter()->type() == type_id<int8_t>().hash_code()) {
33
#ifndef __aarch64__
34
    if (depth3x3 && param->Strides()[0] < 3 &&
35
        param->Strides()[0] == param->Strides()[1]) {
H
hjchen2 已提交
36
      param->ExecMode() = ConvParam<CPU>::EXEC_DEPTHWISE3x3_INT8;
37 38 39
    } else if (depth5x5 && param->Strides()[0] < 2 &&
               param->Strides()[0] == param->Strides()[1]) {
      param->ExecMode() = ConvParam<CPU>::EXEC_DEPTHWISE5x5_INT8;
H
hjchen2 已提交
40
    } else {
41
#endif  // __aarch64__
H
hjchen2 已提交
42
      param->ExecMode() = ConvParam<CPU>::EXEC_GEMM_INT8;
43
#ifndef __aarch64__
H
hjchen2 已提交
44
    }
45
#endif  // __aarch64__
H
hjchen2 已提交
46
  } else {
47
    if (depth3x3 && param->Strides()[0] == param->Strides()[1] &&
48 49
        param->Strides()[0] == 1) {
      param->ExecMode() = ConvParam<CPU>::EXEC_DEPTHWISE3x3S1_FLOAT;
50
    } else if (depth3x3 && param->Strides()[0] == param->Strides()[1] &&
51 52
               param->Strides()[0] == 2) {
      param->ExecMode() = ConvParam<CPU>::EXEC_DEPTHWISE3x3S2_FLOAT;
53 54
    } else if (depth5x5 && param->Strides()[0] == param->Strides()[1] &&
               param->Strides()[0] == 1) {
55
      param->ExecMode() = ConvParam<CPU>::EXEC_DEPTHWISE5x5_FLOAT;
H
hjchen2 已提交
56
    } else if (conv3x3 && param->Groups() == 1 &&
57
               param->Strides()[0] == param->Strides()[1] &&
H
hjchen2 已提交
58
               param->Dilations()[0] == param->Dilations()[1] &&
59
               param->Strides()[0] == 1 && param->Dilations()[0] == 1
60
#if 1
S
StarryRain 已提交
61 62
               && (param->Input()->dims()[1] >= 8 &&
                   param->Output()->dims()[1] >= 8)
63 64
#endif
    ) {
H
hjchen2 已提交
65 66
      param->ExecMode() = ConvParam<CPU>::EXEC_WINOGRAD3X3_FLOAT;
      // transform weight
67 68 69
      Variable *transformed_var = param->GetScope()->Var();
      param->transformed_filter_ =
          transformed_var->GetMutable<framework::LoDTensor>();
H
hjchen2 已提交
70 71
      operators::math::winograd_transform_weight<8, 3>(
          *param->Filter(), param->transformed_filter_);
H
update  
hjchen2 已提交
72
    } else if (conv3x3 && param->Groups() == 1 &&
S
StarryRain 已提交
73 74 75 76 77 78 79 80 81
               param->Strides()[0] == param->Strides()[1] &&
               param->Dilations()[0] == param->Dilations()[1] &&
               param->Strides()[0] == 1 && param->Dilations()[0] == 1
#if 1
               && (param->Input()->dims()[2] >= 48 &&
                   param->Output()->dims()[1] <= 24)
#endif
    ) {
      param->ExecMode() = ConvParam<CPU>::EXEC_SLIDINGWINDOW3x3S1_FLOAT;
H
update  
hjchen2 已提交
82
    } else if (conv3x3 && param->Groups() == 1 &&
S
StarryRain 已提交
83 84 85 86 87 88 89 90 91
               param->Strides()[0] == param->Strides()[1] &&
               param->Dilations()[0] == param->Dilations()[1] &&
               param->Strides()[0] == 2 && param->Dilations()[0] == 1
#if 1
               && (param->Input()->dims()[2] >= 48 &&
                   param->Output()->dims()[1] <= 24)
#endif
    ) {
      param->ExecMode() = ConvParam<CPU>::EXEC_SLIDINGWINDOW3x3S2_FLOAT;
H
hjchen2 已提交
92 93 94 95
    } else {
      param->ExecMode() = ConvParam<CPU>::EXEC_GEMM_FLOAT;
    }
  }
朔-望's avatar
朔-望 已提交
96 97
}

朔-望's avatar
朔-望 已提交
98 99
}  // namespace operators
}  // namespace paddle_mobile