1 /* Copyright 2018 The TensorFlow Authors. All Rights Reserved. 2 3 Licensed under the Apache License, Version 2.0 (the "License"); 4 you may not use this file except in compliance with the License. 5 You may obtain a copy of the License at 6 7 http://www.apache.org/licenses/LICENSE-2.0 8 9 Unless required by applicable law or agreed to in writing, software 10 distributed under the License is distributed on an "AS IS" BASIS, 11 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12 See the License for the specific language governing permissions and 13 limitations under the License. 14 ==============================================================================*/ 15 16 #include <cmath> 17 #include "tensorflow/lite/c/c_api_internal.h" 18 #include "tensorflow/lite/kernels/internal/reference/reference_ops.h" 19 #include "tensorflow/lite/kernels/internal/tensor.h" 20 #include "tensorflow/lite/kernels/kernel_util.h" 21 22 namespace tflite { 23 namespace ops { 24 namespace builtin { 25 namespace elementwise { 26 namespace { 27 28 bool IsNumericSupportedType(const TfLiteType type) { 29 return type == kTfLiteFloat32; 30 } 31 32 bool IsLogicalSupportedType(const TfLiteType type) { 33 return type == kTfLiteBool; 34 } 35 36 typedef bool (*IsSupportedType)(TfLiteType); 37 template <IsSupportedType> 38 TfLiteStatus GenericPrepare(TfLiteContext* context, TfLiteNode* node) { 39 TF_LITE_ENSURE_EQ(context, NumInputs(node), 1); 40 TF_LITE_ENSURE_EQ(context, NumOutputs(node), 1); 41 const TfLiteTensor* input = GetInput(context, node, 0); 42 TfLiteTensor* output = GetOutput(context, node, 0); 43 TF_LITE_ENSURE_EQ(context, input->type, output->type); 44 if (!IsSupportedType(input->type)) { 45 context->ReportError(context, "Current data type %d is not supported.", 46 input->type); 47 return kTfLiteError; 48 } 49 return context->ResizeTensor(context, output, 50 TfLiteIntArrayCopy(input->dims)); 51 } 52 53 template <typename T> 54 inline TfLiteStatus EvalImpl(TfLiteContext* context, TfLiteNode* node, 55 T func(T), TfLiteType expected_type) { 56 const TfLiteTensor* input = GetInput(context, node, 0); 57 TfLiteTensor* output = GetOutput(context, node, 0); 58 TF_LITE_ENSURE_EQ(context, input->type, expected_type); 59 const int64_t num_elements = NumElements(input); 60 const T* in_data = GetTensorData<T>(input); 61 T* out_data = GetTensorData<T>(output); 62 for (int64_t i = 0; i < num_elements; ++i) { 63 out_data[i] = func(in_data[i]); 64 } 65 return kTfLiteOk; 66 } 67 68 inline TfLiteStatus EvalNumeric(TfLiteContext* context, TfLiteNode* node, 69 float float_func(float)) { 70 return EvalImpl<float>(context, node, float_func, kTfLiteFloat32); 71 } 72 73 inline TfLiteStatus EvalLogical(TfLiteContext* context, TfLiteNode* node, 74 bool bool_func(bool)) { 75 return EvalImpl<bool>(context, node, bool_func, kTfLiteBool); 76 } 77 78 TfLiteStatus AbsEval(TfLiteContext* context, TfLiteNode* node) { 79 return EvalNumeric(context, node, std::abs); 80 } 81 82 TfLiteStatus SinEval(TfLiteContext* context, TfLiteNode* node) { 83 return EvalNumeric(context, node, std::sin); 84 } 85 86 TfLiteStatus CosEval(TfLiteContext* context, TfLiteNode* node) { 87 return EvalNumeric(context, node, std::cos); 88 } 89 90 TfLiteStatus LogEval(TfLiteContext* context, TfLiteNode* node) { 91 return EvalNumeric(context, node, std::log); 92 } 93 94 TfLiteStatus SqrtEval(TfLiteContext* context, TfLiteNode* node) { 95 return EvalNumeric(context, node, std::sqrt); 96 } 97 98 TfLiteStatus RsqrtEval(TfLiteContext* context, TfLiteNode* node) { 99 return EvalNumeric(context, node, [](float f) { return 1.f / std::sqrt(f); }); 100 } 101 102 TfLiteStatus SquareEval(TfLiteContext* context, TfLiteNode* node) { 103 return EvalNumeric(context, node, [](float f) { return f * f; }); 104 } 105 106 TfLiteStatus LogicalNotEval(TfLiteContext* context, TfLiteNode* node) { 107 return EvalLogical(context, node, [](bool v) { return !v; }); 108 } 109 110 } // namespace 111 } // namespace elementwise 112 113 TfLiteRegistration* Register_ABS() { 114 static TfLiteRegistration r = { 115 /*init=*/nullptr, /*free=*/nullptr, 116 elementwise::GenericPrepare<elementwise::IsNumericSupportedType>, 117 elementwise::AbsEval}; 118 return &r; 119 } 120 121 TfLiteRegistration* Register_SIN() { 122 static TfLiteRegistration r = { 123 /*init=*/nullptr, /*free=*/nullptr, 124 elementwise::GenericPrepare<elementwise::IsNumericSupportedType>, 125 elementwise::SinEval}; 126 return &r; 127 } 128 129 TfLiteRegistration* Register_COS() { 130 static TfLiteRegistration r = { 131 /*init=*/nullptr, /*free=*/nullptr, 132 elementwise::GenericPrepare<elementwise::IsNumericSupportedType>, 133 elementwise::CosEval}; 134 return &r; 135 } 136 137 TfLiteRegistration* Register_LOG() { 138 static TfLiteRegistration r = { 139 /*init=*/nullptr, /*free=*/nullptr, 140 elementwise::GenericPrepare<elementwise::IsNumericSupportedType>, 141 elementwise::LogEval}; 142 return &r; 143 } 144 145 TfLiteRegistration* Register_SQRT() { 146 static TfLiteRegistration r = { 147 /*init=*/nullptr, /*free=*/nullptr, 148 elementwise::GenericPrepare<elementwise::IsNumericSupportedType>, 149 elementwise::SqrtEval}; 150 return &r; 151 } 152 153 TfLiteRegistration* Register_RSQRT() { 154 static TfLiteRegistration r = { 155 /*init=*/nullptr, /*free=*/nullptr, 156 elementwise::GenericPrepare<elementwise::IsNumericSupportedType>, 157 elementwise::RsqrtEval}; 158 return &r; 159 } 160 161 TfLiteRegistration* Register_SQUARE() { 162 static TfLiteRegistration r = { 163 /*init=*/nullptr, /*free=*/nullptr, 164 elementwise::GenericPrepare<elementwise::IsNumericSupportedType>, 165 elementwise::SquareEval}; 166 return &r; 167 } 168 169 TfLiteRegistration* Register_LOGICAL_NOT() { 170 static TfLiteRegistration r = { 171 /*init=*/nullptr, /*free=*/nullptr, 172 elementwise::GenericPrepare<elementwise::IsLogicalSupportedType>, 173 elementwise::LogicalNotEval}; 174 return &r; 175 } 176 177 } // namespace builtin 178 } // namespace ops 179 } // namespace tflite 180
