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
133
134
135
136
137
138
139
|
/*
* Copyright (c) 2016-2018 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal in the Software without restriction, including without limitation the
* rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
* sell copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#ifndef __ARM_COMPUTE_NEACCUMULATEKERNEL_H__
#define __ARM_COMPUTE_NEACCUMULATEKERNEL_H__
#include "arm_compute/core/NEON/INESimpleKernel.h"
#include <cstdint>
namespace arm_compute
{
class ITensor;
/** Interface for the accumulate kernel
*
* Accumulation is computed by:
* @f[ accum(x,y) = accum(x,y) + input(x,y) @f]
*/
class NEAccumulateKernel : public INESimpleKernel
{
public:
const char *name() const override
{
return "NEAccumulateKernel";
}
/** Set the input and accumulation tensors
*
* @param[in] input Source tensor. Data type supported: U8.
* @param[out] accum Destination tensor. Data type supported: S16.
*/
void configure(const ITensor *input, ITensor *accum);
// Inherited methods overridden:
void run(const Window &window, const ThreadInfo &info) override;
};
/** Interface for the accumulate weighted kernel
*
* Weighted accumulation is computed:
* @f[ accum(x,y) = (1 - \alpha)*accum(x,y) + \alpha*input(x,y) @f]
*
* Where @f$ 0 \le \alpha \le 1 @f$
* Conceptually, the rounding for this is defined as:
* @f[ output(x,y)= uint8( (1 - \alpha) * float32( int32( output(x,y) ) ) + \alpha * float32( int32( input(x,y) ) ) ) @f]
*/
class NEAccumulateWeightedKernel : public INESimpleKernel
{
public:
const char *name() const override
{
return "NEAccumulateWeightedKernel";
}
/** Default constructor */
NEAccumulateWeightedKernel();
/** Set the input and accumulation tensors, and the scale value
*
* @param[in] input Source tensor. Data type supported: U8.
* @param[in] alpha Scalar value in the range [0.0f, 1.0f]
* @param[in,out] accum Accumulated tensor. Data type supported: U8.
*/
void configure(const ITensor *input, float alpha, ITensor *accum);
// Inherited methods overridden:
void run(const Window &window, const ThreadInfo &info) override;
protected:
float _alpha;
};
#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
/** Interface for the accumulate weighted kernel using F16 */
class NEAccumulateWeightedFP16Kernel : public NEAccumulateWeightedKernel
{
public:
const char *name() const override
{
return "NEAccumulateWeightedFP16Kernel";
}
// Inherited methods overridden:
void run(const Window &window, const ThreadInfo &info) override;
};
#else /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
/** Interface for the accumulate weighted kernel using F16 */
using NEAccumulateWeightedFP16Kernel = NEAccumulateWeightedKernel;
#endif /* __ARM_FEATURE_FP16_VECTOR_ARITHMETIC */
/** Interface for the accumulate squared kernel
*
* The accumulation of squares is computed:
* @f[ accum(x,y) = saturate_{int16} ( (uint16) accum(x,y) + (((uint16)(input(x,y)^2)) >> (shift)) ) @f]
*
* Where @f$ 0 \le shift \le 15 @f$
*/
class NEAccumulateSquaredKernel : public INESimpleKernel
{
public:
const char *name() const override
{
return "NEAccumulateSquaredKernel";
}
/** Default constructor */
NEAccumulateSquaredKernel();
/** Set the input and accumulation tensors and the shift value.
*
* @param[in] input Source tensor. Data type supported: U8.
* @param[in] shift Shift value in the range of [0, 15]
* @param[in,out] accum Accumulated tensor. Data type supported: S16.
*/
void configure(const ITensor *input, uint32_t shift, ITensor *accum);
// Inherited methods overridden:
void run(const Window &window, const ThreadInfo &info) override;
private:
uint32_t _shift;
};
} // namespace arm_compute
#endif /*__ARM_COMPUTE_NEACCUMULATEKERNEL_H__ */
|
