Georgios Pinitas | de5a1cc | 2018-02-02 12:52:07 +0000 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (c) 2018 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_CLDIRECTCONVOLUTIONLAYEROUTPUTSTAGEKERNEL_H__ |
| 25 | #define __ARM_COMPUTE_CLDIRECTCONVOLUTIONLAYEROUTPUTSTAGEKERNEL_H__ |
| 26 | |
| 27 | #include "arm_compute/core/CL/ICLKernel.h" |
| 28 | |
| 29 | namespace arm_compute |
| 30 | { |
| 31 | class ITensor; |
| 32 | /** OpenCL kernel to accumulate the biases, if provided, or downscale in case of quantized input. |
| 33 | * |
| 34 | * @note We assume bias to be shared |
| 35 | */ |
| 36 | class CLDirectConvolutionLayerOutputStageKernel : public ICLKernel |
| 37 | { |
| 38 | public: |
| 39 | /** Default constructor */ |
| 40 | CLDirectConvolutionLayerOutputStageKernel(); |
| 41 | /** Prevent instances of this class from being copied (As this class contains pointers) */ |
| 42 | CLDirectConvolutionLayerOutputStageKernel(const CLDirectConvolutionLayerOutputStageKernel &) = delete; |
| 43 | /** Prevent instances of this class from being copied (As this class contains pointers) */ |
| 44 | CLDirectConvolutionLayerOutputStageKernel &operator=(const CLDirectConvolutionLayerOutputStageKernel &) = delete; |
| 45 | /** Allow instances of this class to be moved */ |
| 46 | CLDirectConvolutionLayerOutputStageKernel(CLDirectConvolutionLayerOutputStageKernel &&) = default; |
| 47 | /** Allow instances of this class to be moved */ |
| 48 | CLDirectConvolutionLayerOutputStageKernel &operator=(CLDirectConvolutionLayerOutputStageKernel &&) = default; |
| 49 | /** Default destructor */ |
| 50 | ~CLDirectConvolutionLayerOutputStageKernel() = default; |
| 51 | /** Set the accumulate buffer and the biases of the kernel. |
| 52 | * |
| 53 | * @param[in, out] input Input to add the bias to. If @p output is not specified then accumulation is done in-place. |
| 54 | * Data type supported: S32/QS16/QS32/F16/F32 |
| 55 | * @param[in] bias (Optional) The shared bias tensor to add. It must be 1D Tensor. Data type supported: Same as @p input |
| 56 | * @param[out] output (Optional) If the output tensor is specified the accumulation is done out-of-place. (Defaults to nullptr) |
| 57 | * Required parameter if output is of QASYMM8 type. |
| 58 | * Data types supported: QS8/QASYMM8/QS16/F16/F32 |
| 59 | * @param[in] result_fixedpoint_multiplier (Optional)Fixed point value to be multiplied to each element of the input matrix when once the result_offset has been add |
| 60 | * @param[in] result_shift (Optional)Integer value used to round to nearest division by a power-of-two the result after the fixed point multiplication |
| 61 | * @param[in] result_offset_after_shift (Optional)Offset to be applied to result before converting it back to QASYMM8 |
| 62 | */ |
| 63 | void configure(ICLTensor *input, const ICLTensor *bias = nullptr, ICLTensor *output = nullptr, |
| 64 | int result_fixedpoint_multiplier = 0, int result_shift = 0, int result_offset_after_shift = 0); |
| 65 | /** Static function to check if given info will lead to a valid configuration of @ref CLDirectConvolutionLayerOutputStageKernel |
| 66 | * |
| 67 | * @param[in] input Input to add the bias to. If @p output is not specified then accumulation is done in-place. |
| 68 | * Data type supported: QS16/QS32/F16/F32 |
| 69 | * @param[in] bias (Optional) The shared bias tensor to add. It must be 1D Tensor. Data type supported: Same as @p input |
| 70 | * @param[in] output (Optional) If the output tensor is specified the accumulation is done out-of-place. (Defaults to nullptr) |
| 71 | * Data type supported: QS8/QS16/F16/F32 |
| 72 | * @return a status |
| 73 | */ |
| 74 | static Status validate(const ITensorInfo *input, const ITensorInfo *bias = nullptr, const ITensorInfo *output = nullptr); |
| 75 | |
| 76 | // Inherited methods overridden: |
| 77 | void run(const Window &window, cl::CommandQueue &queue) override; |
| 78 | |
| 79 | private: |
| 80 | ICLTensor *_input; |
| 81 | const ICLTensor *_bias; |
| 82 | ICLTensor *_output; |
| 83 | int _result_fixedpoint_multiplier; |
| 84 | int _result_shift; |
| 85 | int _result_offset_after_shift; |
| 86 | }; |
| 87 | } // namespace arm_compute |
| 88 | #endif /*__ARM_COMPUTE_CLDIRECTCONVOLUTIONLAYEROUTPUTSTAGEKERNEL_H__ */ |