1 /* 2 * Copyright (c) 2018-2020 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_CLDEPTHWISECONVOLUTIONNCHWKERNEL3x3_H 25 #define ARM_COMPUTE_CLDEPTHWISECONVOLUTIONNCHWKERNEL3x3_H 26 27 #include "src/core/CL/kernels/ICLDepthwiseConvolutionLayer3x3Kernel.h" 28 29 namespace arm_compute 30 { 31 class ICLTensor; 32 33 /** Interface for the kernel to run a 3x3 depthwise convolution on a tensor when the data layout is NCHW. 34 */ 35 class CLDepthwiseConvolutionLayer3x3NCHWKernel : public ICLDepthwiseConvolutionLayer3x3Kernel 36 { 37 public: 38 /** Default constructor */ 39 CLDepthwiseConvolutionLayer3x3NCHWKernel(); 40 /** Initialize the function's source, destination, conv and border_size. 41 * 42 * @param[in] input Source tensor. DataType supported: QASYMM8/QASYMM8_SIGNED/F16/F32. 43 * @param[in] weights Weights tensor. A 3D tensor with dimensions [3, 3, IFM]. 44 * Data type supported: Same as @p input or QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8/QASYMM8_SIGNED. 45 * @param[in] biases Biases tensor. A 1D tensor with dimensions [IFM]. Must be nullptr if not needed. 46 * Data type supported: Same as @p input, S32 when input is QASYMM8/QASYMM8_SIGNED. 47 * @param[out] output Destination tensor. Data type supported: Same as @p input. 48 * @param[in] conv_info Padding and stride information to use for the convolution. 49 * @param[in] depth_multiplier (Optional) Multiplier to apply to the input's depth in order to retrieve the output's depth. Defaults to 1. 50 * @param[in] act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU for QASYMM8 supported. 51 * @param[in] dilation (Optional) Dilation, in elements, across x and y. Defaults to (1, 1). 52 * @param[in] output_multipliers (Optional) Output multipliers tensor for quantized computations. In case of per-channel quantization, 53 * the number of multipliers must be equal to the number of filters (IFM). Supported data types: S32 54 * @param[in] output_shifts (Optional) Output shifts tensor for quantized computations. In case of per-channel quantization, 55 * the number of multipliers must be equal to the number of filters (IFM). Supported data types: S32 56 */ 57 void configure(const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info, 58 unsigned int depth_multiplier = 1, ActivationLayerInfo act_info = ActivationLayerInfo(), const Size2D &dilation = Size2D(1U, 1U), 59 const ICLTensor *output_multipliers = nullptr, const ICLTensor *output_shifts = nullptr) override; 60 /** Initialize the function's source, destination, conv and border_size. 61 * 62 * @param[in] compile_context The compile context to be used. 63 * @param[in] input Source tensor. DataType supported: QASYMM8/QASYMM8_SIGNED/F16/F32. 64 * @param[in] weights Weights tensor. A 3D tensor with dimensions [3, 3, IFM]. 65 * Data type supported: Same as @p input or QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8/QASYMM8_SIGNED. 66 * @param[in] biases Biases tensor. A 1D tensor with dimensions [IFM]. Must be nullptr if not needed. 67 * Data type supported: Same as @p input, S32 when input is QASYMM8/QASYMM8_SIGNED. 68 * @param[out] output Destination tensor. Data type supported: Same as @p input. 69 * @param[in] conv_info Padding and stride information to use for the convolution. 70 * @param[in] depth_multiplier (Optional) Multiplier to apply to the input's depth in order to retrieve the output's depth. Defaults to 1. 71 * @param[in] act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU for QASYMM8 supported. 72 * @param[in] dilation (Optional) Dilation, in elements, across x and y. Defaults to (1, 1). 73 * @param[in] output_multipliers (Optional) Output multipliers tensor for quantized computations. In case of per-channel quantization, 74 * the number of multipliers must be equal to the number of filters (IFM). Supported data types: S32 75 * @param[in] output_shifts (Optional) Output shifts tensor for quantized computations. In case of per-channel quantization, 76 * the number of multipliers must be equal to the number of filters (IFM). Supported data types: S32 77 */ 78 void configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info, 79 unsigned int depth_multiplier = 1, ActivationLayerInfo act_info = ActivationLayerInfo(), const Size2D &dilation = Size2D(1U, 1U), 80 const ICLTensor *output_multipliers = nullptr, const ICLTensor *output_shifts = nullptr) override; 81 /** Static function to check if given info will lead to a valid configuration of @ref CLDepthwiseConvolutionLayer3x3NCHWKernel 82 * 83 * @param[in] input Source tensor info. DataType supported: QASYMM8/QASYMM8_SIGNED/F16/F32. 84 * @param[in] weights Weights tensor info. A 3D tensor with dimensions [3, 3, IFM]. 85 * Data type supported: Same as @p input or QASYMM8/QASYMM8_SIGNED/QSYMM8_PER_CHANNEL when @p input is QASYMM8/QASYMM8_SIGNED. 86 * @param[in] biases Biases tensor info. A 1D tensor with dimensions [IFM]. Must be nullptr if not needed. 87 * Data type supported: Same as @p input, S32 when input is QASYMM8/QASYMM8_SIGNED. 88 * @param[in] output Destination tensor. Data type supported: Same as @p input. 89 * @param[in] conv_info Padding and stride information to use for the convolution. 90 * @param[in] depth_multiplier (Optional) Multiplier to apply to the input's depth in order to retrieve the output's depth. Defaults to 1. 91 * @param[in] act_info (Optional) Activation layer information in case of a fused activation. Only RELU, BOUNDED_RELU and LU_BOUNDED_RELU are supported. 92 * @param[in] gpu_target (Optional) GPU target to validate the kernel for. Defaults to midgard. 93 * @param[in] dilation (Optional) Dilation, in elements, across x and y. Defaults to (1, 1). 94 * @param[in] output_multipliers (Optional) Output multipliers tensor info for quantized computations. In case of per-channel quantization, 95 * the number of multipliers must be equal to the number of filters (IFM). Supported data types: S32 96 * @param[in] output_shifts (Optional) Output shifts tensor for quantized computations. In case of per-channel quantization, 97 * the number of multipliers must be equal to the number of filters (IFM). Supported data types: S32 98 * 99 * @return a status 100 */ 101 static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info, 102 unsigned int depth_multiplier = 1, ActivationLayerInfo act_info = ActivationLayerInfo(), GPUTarget gpu_target = GPUTarget::MIDGARD, 103 const Size2D &dilation = Size2D(1U, 1U), const ITensorInfo *output_multipliers = nullptr, const ITensorInfo *output_shifts = nullptr); 104 105 void run(const Window &window, cl::CommandQueue &queue) override; 106 BorderSize border_size() const override; 107 108 private: 109 unsigned int _conv_stride_x; 110 unsigned int _conv_pad_top; 111 unsigned int _conv_pad_left; 112 }; 113 } // namespace arm_compute 114 #endif /*ARM_COMPUTE_CLDEPTHWISECONVOLUTIONNCHWKERNEL3x3_H */ 115