1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
|
/* Copyright 2017 The TensorFlow 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.
==============================================================================*/
#ifndef TENSORFLOW_CONTRIB_LITE_KERNELS_INTERNAL_COMMON_H_
#define TENSORFLOW_CONTRIB_LITE_KERNELS_INTERNAL_COMMON_H_
#ifndef ALLOW_SLOW_GENERIC_DEPTHWISECONV_FALLBACK
#ifdef GEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK
#define ALLOW_SLOW_GENERIC_DEPTHWISECONV_FALLBACK
#endif
#endif
#ifndef USE_NEON
#if defined(__ARM_NEON__) || defined(__ARM_NEON)
#define USE_NEON
#include <arm_neon.h>
#endif
#if defined __GNUC__ && defined __SSE4_1__
#define USE_NEON
#define OPTIMIZED_OPS_H__IGNORE_DEPRECATED_DECLARATIONS
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wdeprecated-declarations"
#pragma GCC diagnostic ignored "-Wattributes"
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wnarrowing"
#pragma GCC diagnostic ignored "-Wsequence-point"
#include "NEON_2_SSE.h"
#pragma GCC diagnostic pop
#endif
#endif
#include "public/gemmlowp.h"
#include "tensorflow/contrib/lite/kernels/internal/types.h"
namespace tflite {
inline void GetActivationMinMax(FusedActivationFunctionType ac,
float* output_activation_min,
float* output_activation_max) {
switch (ac) {
case FusedActivationFunctionType::kNone:
*output_activation_min = std::numeric_limits<float>::lowest();
*output_activation_max = std::numeric_limits<float>::max();
break;
case FusedActivationFunctionType::kRelu:
*output_activation_min = 0.f;
*output_activation_max = std::numeric_limits<float>::max();
break;
case FusedActivationFunctionType::kRelu1:
*output_activation_min = -1.f;
*output_activation_max = 1.f;
break;
case FusedActivationFunctionType::kRelu6:
*output_activation_min = 0.f;
*output_activation_max = 6.f;
break;
}
}
inline float ActivationFunctionWithMinMax(float x, float output_activation_min,
float output_activation_max) {
return std::min(std::max(x, output_activation_min), output_activation_max);
}
// Legacy function, left for compatibility only.
template <FusedActivationFunctionType Ac>
float ActivationFunction(float x) {
float output_activation_min, output_activation_max;
GetActivationMinMax(Ac, &output_activation_min, &output_activation_max);
return ActivationFunctionWithMinMax(x, output_activation_min,
output_activation_max);
}
inline int32 MultiplyByQuantizedMultiplierSmallerThanOneExp(
int32 x, int32 quantized_multiplier, int left_shift) {
using gemmlowp::RoundingDivideByPOT;
using gemmlowp::SaturatingRoundingDoublingHighMul;
return RoundingDivideByPOT(
SaturatingRoundingDoublingHighMul(x, quantized_multiplier), -left_shift);
}
inline int32 MultiplyByQuantizedMultiplierGreaterThanOne(
int32 x, int32 quantized_multiplier, int left_shift) {
using gemmlowp::SaturatingRoundingDoublingHighMul;
return SaturatingRoundingDoublingHighMul(x * (1 << left_shift),
quantized_multiplier);
}
inline int32 MultiplyByQuantizedMultiplier(int32 x, int32 quantized_multiplier,
int shift) {
using gemmlowp::RoundingDivideByPOT;
using gemmlowp::SaturatingRoundingDoublingHighMul;
int left_shift = shift > 0 ? shift : 0;
int right_shift = shift > 0 ? 0 : -shift;
return RoundingDivideByPOT(SaturatingRoundingDoublingHighMul(
x * (1 << left_shift), quantized_multiplier),
right_shift);
}
template <typename T>
int CountLeadingZeros(T integer_input) {
static_assert(std::is_unsigned<T>::value,
"Only unsigned integer types handled.");
const T one_in_leading_positive = static_cast<T>(1)
<< (std::numeric_limits<T>::digits - 1);
int leading_zeros = 0;
while (integer_input < one_in_leading_positive) {
integer_input <<= 1;
++leading_zeros;
}
return leading_zeros;
}
} // namespace tflite
#endif // TENSORFLOW_CONTRIB_LITE_KERNELS_INTERNAL_COMMON_H_
|
