xref: /aosp_15_r20/external/ComputeLibrary/src/cpu/operators/CpuDepthwiseConv2dAssemblyDispatch.h (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1 /*
2  * Copyright (c) 2019-2021 Arm Limited.
3  *
4  * SPDX-License-Identifier: MIT
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to
8  * deal in the Software without restriction, including without limitation the
9  * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10  * sell copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in all
14  * copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22  * SOFTWARE.
23  */
24 #ifndef ARM_COMPUTE_CPU_DEPTHWISE_CONV2D_ASSEMBLY_DISPATCH_H
25 #define ARM_COMPUTE_CPU_DEPTHWISE_CONV2D_ASSEMBLY_DISPATCH_H
26 
27 #include "src/core/common/Macros.h"
28 #include "src/cpu/ICpuOperator.h"
29 
30 namespace arm_compute
31 {
32 namespace cpu
33 {
34 /** Depthwise convolution assembly kernel glue */
35 class CpuDepthwiseConv2dAssemblyDispatch : public ICpuOperator
36 {
37 public:
38     CpuDepthwiseConv2dAssemblyDispatch();
39     ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(CpuDepthwiseConv2dAssemblyDispatch);
40     ~CpuDepthwiseConv2dAssemblyDispatch();
41     /** Initialize the function's source, destination, kernels and border_size.
42      *
43      * @note Supports only NHWC format
44      *
45      * @param[in]  src     Source tensor info. Data type supported: QASYMM8/QASYMM8_SIGNED/F16/F32.
46      * @param[in]  weights Weights tensor info. These are 3D tensors with shape [W, H, IFM].
47      *                     Data type supported: same as @p src or QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p src is QASYMM8/QASYMM8_SIGNED.
48      * @param[in]  bias    (Optional) Biases tensor info. A 1D tensor with shape [IFM]. Must be nullptr if not needed.
49      *                     Data type supported: same as @p src or S32 if @p src is quantized.
50      * @param[out] dst     Destination tensor info. Data type supported: same as @p src.
51      * @param[in]  info    Depthwise convolution meta-data.
52      */
53     void configure(const ITensorInfo *src, const ITensorInfo *weights, const ITensorInfo *bias, ITensorInfo *dst, const ConvolutionInfo &info);
54     /** Static function to check if given info will lead to a valid configuration
55      *
56      * Similar to CpuDepthwiseConv2dAssemblyDispatch::configure()
57      *
58      * @return a status
59      */
60     static Status validate(const ITensorInfo *src, const ITensorInfo *weights, const ITensorInfo *bias, const ITensorInfo *dst, const ConvolutionInfo &info);
61     /** Checks if activation is supported by the assembly kernels
62      *
63      * @param[in] activation Activation to check
64      *
65      * @return True if activation is supported else false
66      */
67     static bool is_activation_supported(const ActivationLayerInfo &activation);
68 
69     // Inherited methods overridden:
70     void run(ITensorPack &tensors) override;
71     void prepare(ITensorPack &tensors) override;
72     experimental::MemoryRequirements workspace() const override;
73 
74 private:
75     struct LocalImpl;
76     std::unique_ptr<LocalImpl> _pImpl;
77 };
78 } // namespace cpu
79 } // namespace arm_compute
80 #endif /* ARM_COMPUTE_CPU_DEPTHWISE_CONV2D_ASSEMBLY_DISPATCH_H */
81