1 | /* Copyright 2020 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 <complex> |
17 | |
18 | #include "tensorflow/lite/c/common.h" |
19 | #include "tensorflow/lite/kernels/internal/tensor.h" |
20 | #include "tensorflow/lite/kernels/internal/tensor_ctypes.h" |
21 | #include "tensorflow/lite/kernels/kernel_util.h" |
22 | |
23 | namespace tflite { |
24 | namespace ops { |
25 | namespace builtin { |
26 | namespace complex { |
27 | |
28 | static const int kInputTensor = 0; |
29 | static const int kOutputTensor = 0; |
30 | |
31 | TfLiteStatus Prepare(TfLiteContext* context, TfLiteNode* node) { |
32 | TF_LITE_ENSURE_EQ(context, NumInputs(node), 1); |
33 | TF_LITE_ENSURE_EQ(context, NumOutputs(node), 1); |
34 | |
35 | const TfLiteTensor* input = GetInput(context, node, kInputTensor); |
36 | |
37 | TF_LITE_ENSURE(context, input->type == kTfLiteComplex64 || |
38 | input->type == kTfLiteComplex128); |
39 | |
40 | TfLiteTensor* output = GetOutput(context, node, kOutputTensor); |
41 | |
42 | if (input->type == kTfLiteComplex64) { |
43 | TF_LITE_ENSURE_TYPES_EQ(context, output->type, kTfLiteFloat32); |
44 | } else { |
45 | TF_LITE_ENSURE_TYPES_EQ(context, output->type, kTfLiteFloat64); |
46 | } |
47 | |
48 | TfLiteIntArray* output_shape = TfLiteIntArrayCopy(input->dims); |
49 | return context->ResizeTensor(context, output, output_shape); |
50 | } |
51 | |
52 | template <typename T, typename ExtractF> |
53 | void (const TfLiteTensor* input, ExtractF , |
54 | TfLiteTensor* output) { |
55 | const std::complex<T>* input_data = GetTensorData<std::complex<T>>(input); |
56 | T* output_data = GetTensorData<T>(output); |
57 | const int input_size = NumElements(input); |
58 | for (int i = 0; i < input_size; ++i) { |
59 | *output_data++ = extract_func(*input_data++); |
60 | } |
61 | } |
62 | |
63 | TfLiteStatus EvalReal(TfLiteContext* context, TfLiteNode* node) { |
64 | const TfLiteTensor* input = GetInput(context, node, kInputTensor); |
65 | |
66 | TfLiteTensor* output = GetOutput(context, node, kOutputTensor); |
67 | |
68 | switch (input->type) { |
69 | case kTfLiteComplex64: { |
70 | ExtractData<float>( |
71 | input, |
72 | static_cast<float (*)(const std::complex<float>&)>(std::real<float>), |
73 | output); |
74 | break; |
75 | } |
76 | case kTfLiteComplex128: { |
77 | ExtractData<double>(input, |
78 | static_cast<double (*)(const std::complex<double>&)>( |
79 | std::real<double>), |
80 | output); |
81 | break; |
82 | } |
83 | default: { |
84 | TF_LITE_KERNEL_LOG(context, |
85 | "Unsupported input type, Real op only supports " |
86 | "complex input, but got: " , |
87 | TfLiteTypeGetName(input->type)); |
88 | return kTfLiteError; |
89 | } |
90 | } |
91 | |
92 | return kTfLiteOk; |
93 | } |
94 | |
95 | TfLiteStatus EvalImag(TfLiteContext* context, TfLiteNode* node) { |
96 | const TfLiteTensor* input = GetInput(context, node, kInputTensor); |
97 | |
98 | TfLiteTensor* output = GetOutput(context, node, kOutputTensor); |
99 | |
100 | switch (input->type) { |
101 | case kTfLiteComplex64: { |
102 | ExtractData<float>( |
103 | input, |
104 | static_cast<float (*)(const std::complex<float>&)>(std::imag<float>), |
105 | output); |
106 | break; |
107 | } |
108 | case kTfLiteComplex128: { |
109 | ExtractData<double>(input, |
110 | static_cast<double (*)(const std::complex<double>&)>( |
111 | std::imag<double>), |
112 | output); |
113 | break; |
114 | } |
115 | default: { |
116 | TF_LITE_KERNEL_LOG(context, |
117 | "Unsupported input type, Imag op only supports " |
118 | "complex input, but got: " , |
119 | TfLiteTypeGetName(input->type)); |
120 | return kTfLiteError; |
121 | } |
122 | } |
123 | |
124 | return kTfLiteOk; |
125 | } |
126 | |
127 | TfLiteStatus EvalAbs(TfLiteContext* context, TfLiteNode* node) { |
128 | const TfLiteTensor* input = GetInput(context, node, kInputTensor); |
129 | TfLiteTensor* output = GetOutput(context, node, kOutputTensor); |
130 | |
131 | switch (input->type) { |
132 | case kTfLiteComplex64: { |
133 | ExtractData<float>( |
134 | input, |
135 | static_cast<float (*)(const std::complex<float>&)>(std::abs<float>), |
136 | output); |
137 | break; |
138 | } |
139 | case kTfLiteComplex128: { |
140 | ExtractData<double>(input, |
141 | static_cast<double (*)(const std::complex<double>&)>( |
142 | std::abs<double>), |
143 | output); |
144 | break; |
145 | } |
146 | default: { |
147 | TF_LITE_KERNEL_LOG(context, |
148 | "Unsupported input type, ComplexAbs op only supports " |
149 | "complex input, but got: " , |
150 | TfLiteTypeGetName(input->type)); |
151 | return kTfLiteError; |
152 | } |
153 | } |
154 | |
155 | return kTfLiteOk; |
156 | } |
157 | |
158 | } // namespace complex |
159 | |
160 | TfLiteRegistration* Register_REAL() { |
161 | static TfLiteRegistration r = {/*init=*/nullptr, /*free=*/nullptr, |
162 | complex::Prepare, complex::EvalReal}; |
163 | return &r; |
164 | } |
165 | |
166 | TfLiteRegistration* Register_IMAG() { |
167 | static TfLiteRegistration r = {/*init=*/nullptr, /*free=*/nullptr, |
168 | complex::Prepare, complex::EvalImag}; |
169 | return &r; |
170 | } |
171 | |
172 | TfLiteRegistration* Register_COMPLEX_ABS() { |
173 | static TfLiteRegistration r = {/*init=*/nullptr, /*free=*/nullptr, |
174 | complex::Prepare, complex::EvalAbs}; |
175 | return &r; |
176 | } |
177 | |
178 | } // namespace builtin |
179 | } // namespace ops |
180 | } // namespace tflite |
181 | |