1 /* 2 * Copyright (c) 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_CL_ELEMENTWISE_OPERATIONS_H 25 #define ARM_COMPUTE_CL_ELEMENTWISE_OPERATIONS_H 26 27 #include "src/gpu/cl/ClCompileContext.h" 28 #include "src/gpu/cl/IClOperator.h" 29 30 namespace arm_compute 31 { 32 namespace opencl 33 { 34 /** Basic function to run @ref opencl::kernels::ClArithmeticKernel for division 35 * 36 * @note The tensor data type for the inputs must be F16/F32. 37 * @note The function performs an arithmetic division between two tensors. 38 */ 39 class ClElementwiseDivision : public IClOperator 40 { 41 public: 42 /** Configure function for a given list of arguments. 43 * 44 * @param[in] compile_context The compile context to be used. 45 * @param[in] src1 First source tensor info. Data types supported: F16/F32. 46 * @param[in] src2 Second source tensor info. same as @p src1. 47 * @param[out] dst Destination tensor info. Data types supported: same as @p src1. 48 * @param[in] act_info (Optional) Activation layer information in case of a fused activation. 49 */ 50 void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 51 /** Static function to check if given info will lead to a valid configuration 52 * 53 * Similar to @ref ClElementwiseDivision::configure() 54 * 55 * @return a status 56 */ 57 static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 58 }; 59 60 /** Basic function to run @ref opencl::kernels::ClArithmeticKernel for max 61 * 62 * @note The tensor data type for the inputs must be U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. 63 * @note The function performs a max operation between two tensors. 64 */ 65 class ClElementwiseMax : public IClOperator 66 { 67 public: 68 /** Configure function for a given list of arguments. 69 * 70 * @param[in] compile_context The compile context to be used. 71 * @param[in] src1 First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32. 72 * @param[in] src2 Second source tensor info. Data types supported: same as @p src1. 73 * @param[out] dst Destination tensor info. Data types supported: same as @p src1. 74 * @param[in] act_info (Optional) Activation layer information in case of a fused activation. 75 */ 76 void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 77 /** Static function to check if given info will lead to a valid configuration 78 * 79 * Similar to @ref ClElementwiseMax::configure() 80 * 81 * @return a status 82 */ 83 static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 84 }; 85 86 /** Basic function to run @ref opencl::kernels::ClArithmeticKernel for min 87 * 88 * @note The tensor data type for the inputs must be U8/QASYMM8/S16/QSYMM16/S32/U32/F16/F32. 89 * @note The function performs a max operation between two tensors. 90 */ 91 class ClElementwiseMin : public IClOperator 92 { 93 public: 94 /** Configure function for a given list of arguments. 95 * 96 * @param[in] compile_context The compile context to be used. 97 * @param[in] src1 First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/S32/U32/F16/F32. 98 * @param[in] src2 Second source tensor info. Data types supported: same as @p src1. 99 * @param[out] dst Destination tensor info. Data types supported: same as @p src1. 100 * @param[in] act_info (Optional) Activation layer information in case of a fused activation. 101 */ 102 void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 103 /** Static function to check if given info will lead to a valid configuration 104 * 105 * Similar to @ref ClElementwiseMin::configure() 106 * 107 * @return a status 108 */ 109 static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 110 }; 111 112 /** Basic function to run @ref opencl::kernels::ClArithmeticKernel for squared difference 113 * 114 * @note The tensor data type for the inputs must be QASYMM8/U8/S16/QSYMM16/F16/F32. 115 * @note The function performs a squared different operation between two tensors (i.e., out[i] = (in1[i] - in2[i])^2 116 */ 117 class ClElementwiseSquaredDiff : public IClOperator 118 { 119 public: 120 /** Configure function for a given list of arguments. 121 * 122 * @param[in] compile_context The compile context to be used. 123 * @param[in] src1 First source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/QSYMM16/F16/F32. 124 * @param[in] src2 Second source tensor info. Data types supported: same as @p src1. 125 * @param[out] dst Destination tensor info. Data types supported: same as @p src1. 126 * @param[in] act_info (Optional) Activation layer information in case of a fused activation. 127 */ 128 void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 129 /** Static function to check if given info will lead to a valid configuration 130 * 131 * Similar to @ref ClElementwiseSquaredDiff::configure() 132 * 133 * @return a status 134 */ 135 static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 136 }; 137 138 /** Basic function to run @ref opencl::kernels::ClArithmeticKernel for power 139 * 140 * @note The tensor data type for the inputs must be F16/F32. 141 * @note The function performs an elementwise power of in1 to in2 (i.e., out[i] = in1[i] ^ in2[i]) 142 */ 143 class ClElementwisePower : public IClOperator 144 { 145 public: 146 /** Configure function for a given list of arguments. 147 * 148 * @param[in] compile_context The compile context to be used. 149 * @param[in] src1 First source tensor info. Data types supported: F16/F32. 150 * @param[in] src2 Second source tensor info. Data types supported: F16/F32. 151 * @param[out] dst Destination tensor info. Data types supported:F16/F32. 152 * @param[in] act_info (Optional) Activation layer information in case of a fused activation. 153 */ 154 void configure(const ClCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 155 /** Static function to check if given info will lead to a valid configuration 156 * 157 * Similar to @ref ClElementwisePower::configure() 158 * 159 * @return a status 160 */ 161 static Status validate(const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, const ActivationLayerInfo &act_info = ActivationLayerInfo()); 162 }; 163 } // namespace opencl 164 } // namespace arm_compute 165 #endif /* ARM_COMPUTE_CL_ELEMENTWISE_OPERATIONS_H */ 166