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
140
141
142
143
144
|
/*
* Copyright (c) 2019-2022, 2024 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 ACL_SRC_CPU_KERNELS_CPUGEMMLOWPOFFSETCONTRIBUTIONOUTPUTSTAGEKERNEL_H
#define ACL_SRC_CPU_KERNELS_CPUGEMMLOWPOFFSETCONTRIBUTIONOUTPUTSTAGEKERNEL_H
#include "arm_compute/core/KernelDescriptors.h"
#include "src/core/common/Macros.h"
#include "src/cpu/ICpuKernel.h"
namespace arm_compute
{
namespace cpu
{
namespace kernels
{
/** Kernel used to add the offset contribution and perform the output stage after @ref CpuGemmLowpMatrixMultiplyKernel.
*
* The computation is performed in-place
*
* This kernel takes a final int32 accumulator value (the output of @ref CpuGemmLowpMatrixMultiplyKernel),
* and adds to it the offset contribution of matrix A and matrix B in-place.
*
* The output stage can perform either QuantizeDownInt32ToUint8Scale or QuantizeDownInt32ToUint8ScaleByFixedPoint for Uint8.
* The output stage can perform either QuantizeDownInt32ToInt8Scale or QuantizeDownInt32ToInt8ScaleByFixedPoint for Int8.
*
* For QuantizeDownInt32ToUint8Scale/QuantizeDownInt32ToInt8Scale the final result is:
*
* ((mm_result'[i][k] + result_offset) * result_mult_int) >> result_shift
*
* For QuantizeDownInt32ToUint8ScaleByFixedPoint/QuantizeDownInt32ToInt8ScaleByFixedPoint the final result is:
*
* (FixedPointMul(mm_result'[i][k], result_fixedpoint_multiplier) >> result_shift) + result_offset_after_shift
*
* where FixedPointMul(x, y) is the nearest integer to the following
* mathematical expression, evaluated without overflow or intermediate rounding:
*
* (x * y) / 2^31
*
* and mm_result'[i][k] = mm_result[i][k] +
* (vector_sum_col[k] * a_offset) +
* (vector_sum_row[i] * b_offset) +
* (a_offset * b_offset * k)
*/
class CpuGemmLowpOffsetContributionOutputStageKernel : public ICpuKernel<CpuGemmLowpOffsetContributionOutputStageKernel>
{
public:
/** Default constructor */
CpuGemmLowpOffsetContributionOutputStageKernel() = default;
ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuGemmLowpOffsetContributionOutputStageKernel);
/** Initialise the kernel inputs and output.
*
* @param[in] mm_result Input tensor info containing the result of @ref CpuGemmLowpMatrixMultiplyKernel. Data type supported: S32
* @param[in] vector_sum_col Input row-vector tensor info of sums of all the entries in each column of matrix B.
* Can be a 1D or 2D tensor, in case of 2D, y dim is the batch dimension
* Note: vector_sum_col can be a nullptr in case a_offset = 0. Data type supported: same as @p mm_result
* @param[in] vector_sum_row Input row-vector tensor info of sums of all the entries in each row of matrix A.
* Can be a 1D or 2D tensor, in case of 2D, y dim is the batch dimension
* @param[in] bias Biases tensor info. Only shared biases supported and it can be a nullptr if the addition of biases is not required.
* Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as @p mm_result.
* @param[out] dst Output tensor info containing the final quantized result. Data type supported: QASYMM8/QASYMM8_SIGNED
* @param[in] k Number of matrix A columns or Matrix B rows
* @param[in] a_offset Offset to be added to each element of the matrix A.
* @param[in] b_offset Offset to be added to each element of the matrix B.
* @param[in] output_stage GEMMLowp output stage info, providing the type of quantization and the necessary parameters.
*/
void configure(const ITensorInfo *mm_result,
const ITensorInfo *vector_sum_col,
const ITensorInfo *vector_sum_row,
const ITensorInfo *bias,
ITensorInfo *dst,
int32_t k,
int32_t a_offset,
int32_t b_offset,
GEMMLowpOutputStageInfo output_stage);
/** Static function to check if given info will lead to a valid configuration
*
* Similar to CpuGemmLowpOffsetContributionOutputStageKernel::configure()
*
* @return a status
*/
static Status validate(const ITensorInfo *mm_result,
const ITensorInfo *vector_sum_col,
const ITensorInfo *vector_sum_row,
const ITensorInfo *bias,
const ITensorInfo *dst,
int32_t a_offset,
int32_t b_offset,
GEMMLowpOutputStageInfo output_stage);
/** Set the a offset
* Warning: if a_offset is non-zero then vector_sum_col must be set in run_op.
* Run configure or validate again if you aren't sure
*
* @param[in] a_offset Offset to be added to each element of the matrix A.
*/
void set_a_offset(int32_t a_offset);
/** Set the b offset
* Warning: if b_offset is non-zero then vector_sum_col must be set in run_op.
* Run configure or validate again if you aren't sure
*
* @param[in] b_offset Offset to be added to each element of the matrix B.
*/
void set_b_offset(int32_t b_offset);
// Inherited methods overridden:
void run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) override;
const char *name() const override;
private:
/** Function to use for the particular tensors passed to configure() */
int32_t _a_offset{0};
int32_t _b_offset{0};
int32_t _k{0}; // Number of columns of A or rows of B, used in last offset term
bool _is_vector_sum_col_batched{true};
GEMMLowpOutputStageInfo _output_stage{GEMMLowpOutputStageInfo()};
};
} // namespace kernels
} // namespace cpu
} // namespace arm_compute
#endif // ACL_SRC_CPU_KERNELS_CPUGEMMLOWPOFFSETCONTRIBUTIONOUTPUTSTAGEKERNEL_H
|
