1 | /* Copyright 2021 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 | #include "tensorflow/lite/kernels/internal/reference/broadcast_args.h" |
16 | |
17 | #include <algorithm> |
18 | #include <cstdint> |
19 | #include <memory> |
20 | |
21 | #include "tensorflow/lite/c/common.h" |
22 | #include "tensorflow/lite/kernels/internal/tensor.h" |
23 | #include "tensorflow/lite/kernels/kernel_util.h" |
24 | |
25 | namespace tflite { |
26 | namespace ops { |
27 | namespace builtin { |
28 | namespace broadcast_args { |
29 | |
30 | constexpr int kShape1Tensor = 0; |
31 | constexpr int kShape2Tensor = 1; |
32 | constexpr int kOutputTensor = 0; |
33 | |
34 | struct BroadcastArgsContext { |
35 | BroadcastArgsContext(TfLiteContext* context, TfLiteNode* node) { |
36 | shape1 = GetInput(context, node, kShape1Tensor); |
37 | shape2 = GetInput(context, node, kShape2Tensor); |
38 | output = GetOutput(context, node, kOutputTensor); |
39 | } |
40 | const TfLiteTensor* shape1; |
41 | const TfLiteTensor* shape2; |
42 | TfLiteTensor* output; |
43 | }; |
44 | |
45 | TfLiteStatus Prepare(TfLiteContext* context, TfLiteNode* node) { |
46 | TF_LITE_ENSURE(context, NumInputs(node) == 2); |
47 | TF_LITE_ENSURE_EQ(context, NumOutputs(node), 1); |
48 | |
49 | BroadcastArgsContext op_context(context, node); |
50 | TF_LITE_ENSURE(context, op_context.shape1->type == kTfLiteInt32 || |
51 | op_context.shape1->type == kTfLiteInt64); |
52 | TF_LITE_ENSURE_EQ(context, op_context.shape1->type, op_context.shape2->type); |
53 | TF_LITE_ENSURE_EQ(context, op_context.shape1->type, op_context.output->type); |
54 | |
55 | // Ensures the shapes are 1D tensor. |
56 | TF_LITE_ENSURE_EQ(context, NumDimensions(op_context.shape1), 1); |
57 | TF_LITE_ENSURE_EQ(context, NumDimensions(op_context.shape2), 1); |
58 | |
59 | // Resizing the shape of the output tensor. |
60 | TfLiteIntArray* output_shape = TfLiteIntArrayCreate(1); |
61 | output_shape->data[0] = std::max(SizeOfDimension(op_context.shape1, 0), |
62 | SizeOfDimension(op_context.shape2, 0)); |
63 | return context->ResizeTensor(context, op_context.output, output_shape); |
64 | } |
65 | |
66 | TfLiteStatus Eval(TfLiteContext* context, TfLiteNode* node) { |
67 | BroadcastArgsContext op_context(context, node); |
68 | |
69 | #define TF_LITE_BROADCAST_ARG(data_type) \ |
70 | reference_ops::BroadcastArgs(GetTensorShape(op_context.shape1), \ |
71 | GetTensorData<data_type>(op_context.shape1), \ |
72 | GetTensorShape(op_context.shape2), \ |
73 | GetTensorData<data_type>(op_context.shape2), \ |
74 | GetTensorShape(op_context.output), \ |
75 | GetTensorData<data_type>(op_context.output)) |
76 | |
77 | if (op_context.output->type == kTfLiteInt32) { |
78 | TF_LITE_BROADCAST_ARG(int32_t); |
79 | } else { |
80 | TF_LITE_BROADCAST_ARG(int64_t); |
81 | } |
82 | #undef TF_LITE_BROADCAST_ARG |
83 | |
84 | return kTfLiteOk; |
85 | } |
86 | |
87 | } // namespace broadcast_args |
88 | |
89 | TfLiteRegistration* Register_BROADCAST_ARGS() { |
90 | static TfLiteRegistration r = {nullptr, nullptr, broadcast_args::Prepare, |
91 | broadcast_args::Eval}; |
92 | return &r; |
93 | } |
94 | |
95 | } // namespace builtin |
96 | } // namespace ops |
97 | } // namespace tflite |
98 | |