Compute Library
 21.02
NEStackLayerKernel.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2018-2021 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  */
25 
26 #include "arm_compute/core/Error.h"
32 #include "arm_compute/core/Utils.h"
38 
39 using namespace arm_compute;
41 
42 namespace
43 {
44 Status validate_arguments(const ITensorInfo *input, unsigned int axis, unsigned int idx_input, unsigned int num_tensors, const ITensorInfo *output)
45 {
47  // Note: ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input) is not needed here as this kernel doesn't use Neon FP16 instructions.
49  ARM_COMPUTE_RETURN_ERROR_ON(idx_input >= num_tensors);
52 
53  if(output->total_size() != 0)
54  {
58  }
59 
60  return Status{};
61 }
62 
63 std::pair<Status, Window> validate_and_configure_window(ITensorInfo *input, unsigned int axis, unsigned int num_tensors, ITensorInfo *output)
64 {
65  // Output auto inizialitation if not yet initialized
66  auto_init_if_empty(*output, input->clone()->set_tensor_shape(compute_stack_shape(*input, axis, num_tensors)));
67 
68  // Configure kernel window
69  Window win = calculate_max_window(*input);
70 
71  return std::make_pair(Status{}, win);
72 }
73 
74 inline Coordinates shift_from_axis_and_replace_coordinate(const Coordinates &id, unsigned int axis, unsigned int idx_input)
75 {
76  constexpr int max_out_coord = 5; // Input shape is max a 4D shape, output is max 5D
77  Coordinates id_out = id;
78  for(unsigned int i = max_out_coord - 1; i > axis; --i)
79  {
80  id_out.set(i, id[i - 1]);
81  }
82  id_out.set(axis, idx_input);
83  return id_out;
84 }
85 } // namespace
86 
88  : _input(nullptr), _output(nullptr), _axis(), _idx_input()
89 {
90 }
91 
92 void NEStackLayerKernel::configure(const ITensor *input, unsigned int axis, unsigned int idx_input, unsigned int num_tensors, ITensor *output)
93 {
94  ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
95  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), axis, idx_input, num_tensors, output->info()));
96 
97  _input = input;
98  _output = output;
99  _axis = axis;
100  _idx_input = idx_input;
101 
102  // Configure kernel window
103  auto win_config = validate_and_configure_window(input->info(), axis, num_tensors, output->info());
104 
105  ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
106  INEKernel::configure(win_config.second);
107 }
108 
109 Status NEStackLayerKernel::validate(const ITensorInfo *input, unsigned int axis, unsigned int idx_input, unsigned int num_tensors, const ITensorInfo *output)
110 {
111  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, axis, idx_input, num_tensors, output));
112  ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), axis, num_tensors, output->clone().get()).first);
113  return Status{};
114 }
115 
117 {
118  ARM_COMPUTE_UNUSED(info);
121 
122  Window window_out;
123  window_out.use_tensor_dimensions(_output->info()->tensor_shape());
124 
125  Iterator input(_input, window);
126  Iterator output(_output, window_out);
127 
128  const int stride_x = _output->info()->strides_in_bytes()[0];
129  const int stride_y = _output->info()->num_dimensions() >= 1 ? _output->info()->strides_in_bytes()[1] : 0;
130  const int stride_z = _output->info()->num_dimensions() >= 2 ? _output->info()->strides_in_bytes()[2] : 0;
131  const int stride_w = _output->info()->num_dimensions() >= 3 ? _output->info()->strides_in_bytes()[3] : 0;
132  const int stride_k = _output->info()->num_dimensions() >= 4 ? _output->info()->strides_in_bytes()[4] : 0;
133 
134  execute_window_loop(window, [&](const Coordinates & id)
135  {
136  Coordinates id_out = shift_from_axis_and_replace_coordinate(id, _axis, _idx_input);
137  const int idx = id_out[0] * stride_x + id_out[1] * stride_y + id_out[2] * stride_z + id_out[3] * stride_w + id_out[4] * stride_k;
138  std::memcpy(output.ptr() + idx, input.ptr(), _input->info()->element_size());
139  },
140  input);
141 }
void set(size_t dimension, T value, bool increase_dim_unit=true)
Accessor to set the value of one of the dimensions.
Definition: Dimensions.h:76
virtual size_t num_dimensions() const =0
The number of dimensions of the tensor (rank)
Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps, bool skip_border, BorderSize border_size)
const Window & window() const
The maximum window the kernel can be executed on.
Definition: IKernel.cpp:28
TensorShape compute_stack_shape(const ITensorInfo &a, unsigned int axis, unsigned int num_tensors)
Calculate the stack output shape of a tensor.
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(...)
Definition: Validate.h:610
#define ARM_COMPUTE_RETURN_ON_ERROR(status)
Checks if a status contains an error and returns it.
Definition: Error.h:204
virtual DataType data_type() const =0
Data type used for each element of the tensor.
Store the tensor&#39;s metadata.
Definition: ITensorInfo.h:40
void run(const Window &window, const ThreadInfo &info) override
Execute the kernel on the passed window.
NEStackLayerKernel()
Default constructor.
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
Status class.
Definition: Error.h:52
#define ARM_COMPUTE_RETURN_ERROR_ON(cond)
If the condition is true, an error is returned.
Definition: Error.h:296
Interface for Neon tensor.
Definition: ITensor.h:36
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(...)
Definition: Validate.h:288
void use_tensor_dimensions(const TensorShape &shape, size_t first_dimension=Window::DimX)
Use the tensor&#39;s dimensions to fill the window dimensions.
Definition: Window.inl:276
Copyright (c) 2017-2021 Arm Limited.
#define ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(...)
Definition: Validate.h:163
#define ARM_COMPUTE_UNUSED(...)
To avoid unused variables warnings.
Definition: Error.h:152
virtual const TensorShape & tensor_shape() const =0
Size for each dimension of the tensor.
static Status validate(const ITensorInfo *input, unsigned int axis, unsigned int idx_input, unsigned int num_tensors, const ITensorInfo *output)
Static function to check if given info will lead to a valid configuration of NEStackLayerKernel.
Coordinates of an item.
Definition: Coordinates.h:37
bool auto_init_if_empty(ITensorInfo &info, const TensorShape &shape, int num_channels, DataType data_type, QuantizationInfo quantization_info=QuantizationInfo())
Auto initialize the tensor info (shape, number of channels and data type) if the current assignment i...
virtual std::unique_ptr< T > clone() const =0
Provide a clone of the current object of class T.
virtual ITensorInfo * info() const =0
Interface to be implemented by the child class to return the tensor&#39;s metadata.
constexpr uint8_t * ptr() const
Return a pointer to the current pixel.
Definition: Helpers.inl:139
virtual size_t element_size() const =0
Element size in bytes calculated as data_size() * num_channels()
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:941
ScaleKernelInfo info(interpolation_policy, default_border_mode, PixelValue(), sampling_policy, false)
Information about executing thread and CPU.
Definition: CPPTypes.h:235
virtual size_t total_size() const =0
Returns the total size of the tensor in bytes.
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(...)
Definition: Validate.h:545
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const GEMMLowpOutputStageInfo *output_stage)
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161
void execute_window_loop(const Window &w, L &&lambda_function, Ts &&... iterators)
Iterate through the passed window, automatically adjusting the iterators and calling the lambda_funct...
Definition: Helpers.inl:77
virtual const Strides & strides_in_bytes() const =0
The strides in bytes for accessing each dimension of the tensor.
Iterator updated by execute_window_loop for each window element.
Definition: Helpers.h:46
Describe a multidimensional execution window.
Definition: Window.h:39
void configure(const ITensor *input, unsigned int axis, unsigned int idx_input, unsigned int num_tensors, ITensor *output)
Initialise the kernel&#39;s inputs and output.
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:205