aboutsummaryrefslogtreecommitdiff
path: root/src/core/gpu/cl/kernels/ClElementwiseKernel.h
blob: 4ed8ae73ab8291db90b1f8d3a3c499e8c4967b0d (plain)
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
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
/*
 * Copyright (c) 2018-2021 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_CL_ELEMENTWISE_KERNEL_H
#define ARM_COMPUTE_CL_ELEMENTWISE_KERNEL_H

#include "src/core/KernelTypes.h"
#include "src/core/common/Macros.h"
#include "src/core/gpu/cl/ClCompileContext.h"
#include "src/core/gpu/cl/IClKernel.h"

namespace arm_compute
{
namespace opencl
{
namespace kernels
{
/** Interface for an element-wise operation kernel
 *
 * Element-wise operation is computed by:
 * @f[ dst(x,y) = OP(src1(x,y), src2(x,y))@f]
 *
 */
class ClElementwiseKernel : public IClKernel
{
public:
    /** Default constructor */
    ClElementwiseKernel() = default;
    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClElementwiseKernel);

    // Inherited methods overridden:
    void run_op(ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override;

protected:
    /** The name of the operation */
    virtual std::string name() = 0;

    /** Configure kernel for a given list of arguments
     *
     * @param[in] src1 First source tensor info. Data types supported: U8/S8/QASYMM8/QASYMM8_SIGNED/U16/S16/F16/U32/S32/F32.
     * @param[in] src2 Second source tensor info. Data types supported: same as @p src1.
     * @param[in] dst  Destination tensor info. Data types supported: same as @p src1.
     *
     * @return a pair of Status and Window
     */
    virtual std::pair<Status, Window> validate_and_configure_window(ITensorInfo &src1, ITensorInfo &src2, ITensorInfo &dst) = 0;

    /** Generate the build options for the specific kernel
     *
     * @reutrn a CLBuildOptions struct
     */
    virtual CLBuildOptions generate_build_options(const ITensorInfo &src1, const ITensorInfo &src2, const ITensorInfo &dst) = 0;

    /** Generate the identifier for tuning
     *
     * @reutrn a string
     */
    virtual std::string generate_id_for_tuning(const std::string &kernel_name, const ITensorInfo &src1, const ITensorInfo &dst) = 0;

    /** Commmon configure function for element-wise operators with no additional options (e.g., Div, Min, Max, SquaredDiff)
     *
     */
    void configure_common(ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst);
    /** Commmon configure function for element-wise operators with no additional options (e.g., Div, Min, Max, SquaredDiff)
     *
     */
    void configure_common(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst);

    ActivationLayerInfo _act_info{};

private:
    const ITensorInfo *_src1{ nullptr }; /**< Source tensor info 1 */
    const ITensorInfo *_src2{ nullptr }; /**< Source tensor info 2 */
    ITensorInfo       *_dst{ nullptr };  /**< Destination tensor info */
};

class ClLogicalBinaryKernel : public ClElementwiseKernel
{
public:
    /** Default constructor */
    ClLogicalBinaryKernel() = default;
    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClLogicalBinaryKernel);
    /** Function to configure kernel
     *
     * @param[in] compile_context The compile context to be used.
     * @param[in] op              Logical binary operation to be executed.
     * @param[in] src1            First source tensor info. Data types supported: U8.
     * @param[in] src2            Second source tensor info. Data types supported: same as @p src1.
     * @param[in] dst             Destination tensor info. Data types supported: same as @p src1.
     */
    void configure(const ClCompileContext &compile_context, LogicalOperation op, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst);
    /** Static function to check if the given configuration is valid for this kernel
     *
     * @param[in] op   Logical binary operation to be executed.
     * @param[in] src1 First source tensor info. Data types supported: U8.
     * @param[in] src2 Second source tensor info. Data types supported: same as @p src1.
     * @param[in] dst  Destination tensor info. Data types supported: same as @p src1.
     */
    static Status validate(LogicalOperation op, const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst);

private:
    // Inherited methods overridden:
    std::string name() override;
    std::pair<Status, Window> validate_and_configure_window(ITensorInfo &src1, ITensorInfo &src2, ITensorInfo &dst) override;
    CLBuildOptions generate_build_options(const ITensorInfo &src1, const ITensorInfo &src2, const ITensorInfo &dst) override;
    std::string generate_id_for_tuning(const std::string &kernel_name, const ITensorInfo &src1, const ITensorInfo &dst) override;

    LogicalOperation _op{ LogicalOperation::Unknown };
};

/** Addition operation */
class ClSaturatedArithmeticKernel : public ClElementwiseKernel
{
public:
    ClSaturatedArithmeticKernel() = default;
    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClSaturatedArithmeticKernel);
    /** Static function to check if given info will lead to a valid configuration of @ref ClSaturatedArithmeticKernel
     *
     * @param[in] compile_context The compile context to be used.
     * @param[in] op              Arithmetic operation to be executed.
     * @param[in] input1          First tensor input info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/S32/F32.
     * @param[in] input2          Second tensor input info. Data types supported: Same as @p input1.
     * @param[in] output          Output tensor info. Data types supported: Same as @p input1.
     * @param[in] policy          Policy to use to handle overflow.
     * @param[in] act_info        (Optional) Activation layer information in case of a fused activation.
     */
    void configure(const ClCompileContext &compile_context, ArithmeticOperation op, ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output, const ConvertPolicy &policy,
                   const ActivationLayerInfo &act_info = ActivationLayerInfo());

    /** Static function to check if given info will lead to a valid configuration of @ref ClSaturatedArithmeticKernel
     *
     * @param[in] op       Arithmetic operation to be executed.
     * @param[in] input1   First tensor input info info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/S32/F32.
     * @param[in] input2   Second tensor input info info. Data types supported: Same as @p input1.
     * @param[in] output   Output tensor info info. Data types supported: Same as @p input1.
     * @param[in] policy   Policy to use to handle overflow.
     * @param[in] act_info (Optional) Activation layer information in case of a fused activation.
     *
     * @return a Status
     */
    static Status validate(ArithmeticOperation op, const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output, const ConvertPolicy &policy,
                           const ActivationLayerInfo &act_info = ActivationLayerInfo());

protected:
    // Inherited methods overridden:
    std::string name() override;
    std::pair<Status, Window> validate_and_configure_window(ITensorInfo &input1, ITensorInfo &input2, ITensorInfo &output) override;
    CLBuildOptions generate_build_options(const ITensorInfo &input1, const ITensorInfo &input2, const ITensorInfo &output) override;
    std::string generate_id_for_tuning(const std::string &kernel_name, const ITensorInfo &input1, const ITensorInfo &output) override;

private:
    ConvertPolicy       _policy{};
    ArithmeticOperation _op{};
};

class ClArithmeticKernel : public ClElementwiseKernel
{
public:
    ClArithmeticKernel() = default;
    ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(ClArithmeticKernel);

    /** Static function to check if given info will lead to a valid configuration of @ref ClArithmeticKernel
     *
     * @param[in] compile_context The compile context to be used.
     * @param[in] op              Arithmetic operation to be executed.
     * @param[in] src1            First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/S32/F32.
     * @param[in] src2            Second source tensor info. Data types supported: same as @p src1.
     * @param[in] dst             Destination tensor info. Data types supported: same as @p src1.
     * @param[in] act_info        (Optional) Activation layer information in case of a fused activation.
     */
    void configure(const ClCompileContext &compile_context, ArithmeticOperation op, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst,
                   const ActivationLayerInfo &act_info = ActivationLayerInfo());

    /** Static function to check if given info will lead to a valid configuration of @ref ClArithmeticKernel
     *
     * @param[in] op       Arithmetic operation to be executed.
     * @param[in] src1     First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/S32/F32.
     * @param[in] src2     Second source tensor info. Data types supported: same as @p src1.
     * @param[in] dst      Destination tensor info. Data types supported: same as @p src1.
     * @param[in] act_info (Optional) Activation layer information in case of a fused activation.
     *
     * @return a Status
     */
    static Status validate(ArithmeticOperation op, const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo());

protected:
    // Inherited methods overridden:
    std::string name() override;
    std::pair<Status, Window> validate_and_configure_window(ITensorInfo &src1, ITensorInfo &src2, ITensorInfo &dst) override;
    CLBuildOptions generate_build_options(const ITensorInfo &src1, const ITensorInfo &src2, const ITensorInfo &dst) override;
    std::string generate_id_for_tuning(const std::string &kernel_name, const ITensorInfo &src1, const ITensorInfo &dst) override;

private:
    ArithmeticOperation _op{};
};
} // namespace kernels
} // namespace opencl
} // namespace arm_compute
#endif /* ARM_COMPUTE_CL_ELEMENTWISE_KERNEL_H */