1/* Copyright 2015 The TensorFlow Authors. All Rights Reserved.
2
3Licensed under the Apache License, Version 2.0 (the "License");
4you may not use this file except in compliance with the License.
5You may obtain a copy of the License at
6
7 http://www.apache.org/licenses/LICENSE-2.0
8
9Unless required by applicable law or agreed to in writing, software
10distributed under the License is distributed on an "AS IS" BASIS,
11WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12See the License for the specific language governing permissions and
13limitations under the License.
14==============================================================================*/
15
16#ifndef TENSORFLOW_CORE_KERNELS_BROADCAST_TO_OP_H_
17#define TENSORFLOW_CORE_KERNELS_BROADCAST_TO_OP_H_
18
19#include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
20#include "tensorflow/core/framework/op_kernel.h"
21#include "tensorflow/core/framework/tensor.h"
22#include "tensorflow/core/framework/tensor_shape.h"
23#include "tensorflow/core/framework/tensor_types.h"
24#include "tensorflow/core/framework/types.h"
25#include "tensorflow/core/kernels/fill_functor.h"
26#include "tensorflow/core/util/bcast.h"
27
28namespace tensorflow {
29
30namespace functor {
31
32template <typename Device, typename T>
33struct BroadcastTo {
34 template <int NDIMS>
35 void DoBCast(
36 const Device &device, typename TTypes<T, NDIMS>::Tensor out,
37 typename TTypes<T, NDIMS>::ConstTensor in,
38 const typename Eigen::array<Eigen::DenseIndex, NDIMS> &bcast) const {
39 MaybeWith32BitIndexing<Device>(
40 [&](auto out32, auto in32, const auto &bcast32) {
41 out32.device(device) = in32.broadcast(bcast32);
42 },
43 out, in, bcast);
44 }
45
46 template <int NDIMS>
47 void ReshapeAndBCast(const Device &device, Tensor &output_tensor,
48 const Tensor &input_tensor, const BCast &bcast) const {
49 DoBCast<NDIMS>(
50 device, output_tensor.template shaped<T, NDIMS>(bcast.result_shape()),
51 input_tensor.template shaped<T, NDIMS>(bcast.x_reshape()),
52 BCast::ToIndexArrayType<Eigen::DenseIndex, NDIMS>(bcast.x_bcast()));
53 }
54
55 // PRECONDITION: rank(input_shape) > 0 &&
56 // rank(input_shape) <= rank(output_shape) &&
57 // output_shape.num_elements() > 0.
58 void operator()(const Device &device, OpKernelContext *ctx,
59 Tensor &output_tensor, const TensorShape &output_shape,
60 const Tensor &input_tensor, const TensorShape &input_shape,
61 const BCast &bcast) const {
62 const int ndims = bcast.y_reshape().size();
63 switch (ndims) {
64 case 1:
65 ReshapeAndBCast<1>(device, output_tensor, input_tensor, bcast);
66 break;
67 case 2:
68 ReshapeAndBCast<2>(device, output_tensor, input_tensor, bcast);
69 break;
70 case 3:
71 ReshapeAndBCast<3>(device, output_tensor, input_tensor, bcast);
72 break;
73 case 4:
74 ReshapeAndBCast<4>(device, output_tensor, input_tensor, bcast);
75 break;
76 case 5:
77 ReshapeAndBCast<5>(device, output_tensor, input_tensor, bcast);
78 break;
79 default:
80 ctx->SetStatus(errors::Unimplemented(
81 "Broadcast between ", input_shape.DebugString(), " and ",
82 output_shape.DebugString(), " is not supported yet."));
83 break;
84 }
85 }
86};
87
88} // namespace functor
89} // namespace tensorflow
90
91#endif // TENSORFLOW_CORE_KERNELS_BROADCAST_TO_OP_H_
92