Compute Library
 21.02
CLSpaceToBatchLayerKernel.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2018-2020 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 
29 #include "src/core/CL/CLValidate.h"
32 #include "support/StringSupport.h"
33 
35 
36 namespace arm_compute
37 {
38 namespace
39 {
40 Status validate_arguments(const ITensorInfo *input, const ITensorInfo *block_info, const ITensorInfo *paddings, const ITensorInfo *output)
41 {
42  ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, block_info, paddings, output);
44  ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
45  ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
46  ARM_COMPUTE_RETURN_ERROR_ON(block_info->num_dimensions() > 1);
47  ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(block_info->tensor_shape(), TensorShape{ 2 });
48  ARM_COMPUTE_RETURN_ERROR_ON(paddings->num_dimensions() > 2);
49  ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(paddings->tensor_shape(), TensorShape{ 2, 2 });
50 
51  // Validate output if initialized
52  if(output->total_size() != 0)
53  {
54  const DataLayout data_layout = input->data_layout();
55  const int idx_channel = get_data_layout_dimension_index(data_layout, DataLayoutDimension::CHANNEL);
56  ARM_COMPUTE_RETURN_ERROR_ON(input->tensor_shape()[idx_channel] != output->tensor_shape()[idx_channel]);
59  }
60 
61  return Status{};
62 }
63 Status validate_arguments_static(const ITensorInfo *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right,
64  const ITensorInfo *output)
65 {
67  ARM_COMPUTE_RETURN_ERROR_ON(input->data_type() == DataType::UNKNOWN);
68  ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() > 4);
69  ARM_COMPUTE_RETURN_ERROR_ON(block_shape_x < 1 || block_shape_y < 1);
70 
71  // Validate output if initialized
72  if(output->total_size() != 0)
73  {
74  TensorShape expected_output_shape = misc::shape_calculator::compute_space_to_batch_shape(input, block_shape_x, block_shape_y, padding_left, padding_right);
75  ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), expected_output_shape);
78  }
79 
80  return Status{};
81 }
82 } // namespace
83 
85  : _input(nullptr), _block_shape(nullptr), _paddings(nullptr), _output(nullptr)
86 {
87 }
88 
89 void CLSpaceToBatchLayerKernel::configure(const ICLTensor *input, const ICLTensor *block_shape, const ICLTensor *paddings, ICLTensor *output)
90 {
91  configure(CLKernelLibrary::get().get_compile_context(), input, block_shape, paddings, output);
92 }
93 
94 void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const ICLTensor *block_shape, const ICLTensor *paddings, ICLTensor *output)
95 {
96  ARM_COMPUTE_ERROR_ON_NULLPTR(input, block_shape, paddings, output);
97  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), block_shape->info(), paddings->info(), output->info()));
98  auto padding_info = get_padding_info({ input, block_shape, paddings, output });
99 
100  _input = input;
101  _block_shape = block_shape;
102  _paddings = paddings;
103  _output = output;
104 
105  const DataLayout data_layout = input->info()->data_layout();
108  const int idx_batch = get_data_layout_dimension_index(data_layout, DataLayoutDimension::BATCHES);
109 
110  // Create kernel
111  CLBuildOptions build_opts;
112  build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
113  build_opts.add_option("-DWIDTH_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_width)));
114  build_opts.add_option("-DHEIGHT_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_height)));
115  build_opts.add_option("-DBATCH_SIZE=" + support::cpp11::to_string(output->info()->dimension(idx_batch)));
116  build_opts.add_option("-DWIDTH_IN=" + support::cpp11::to_string(input->info()->dimension(idx_width)));
117  build_opts.add_option("-DHEIGHT_IN=" + support::cpp11::to_string(input->info()->dimension(idx_height)));
118  build_opts.add_option("-DBATCH_IN=" + support::cpp11::to_string(input->info()->dimension(idx_batch)));
119  _kernel = create_kernel(compile_context, "space_to_batch_" + lower_string(string_from_data_layout(input->info()->data_layout())), build_opts.options());
120 
121  // Configure kernel window
122  Window win = calculate_max_window(*output->info(), Steps());
123  ICLKernel::configure_internal(win);
125 }
126 
127 void CLSpaceToBatchLayerKernel::configure(const ICLTensor *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right,
128  ICLTensor *output)
129 {
130  configure(CLKernelLibrary::get().get_compile_context(), input, block_shape_x, block_shape_y, padding_left, padding_right, output);
131 }
132 
133 void CLSpaceToBatchLayerKernel::configure(const CLCompileContext &compile_context, const ICLTensor *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left,
134  const Size2D &padding_right,
135  ICLTensor *output)
136 {
137  ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
138 
139  TensorShape output_shape = misc::shape_calculator::compute_space_to_batch_shape(input->info(), block_shape_x, block_shape_y, padding_left, padding_right);
140  auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->quantization_info());
141 
142  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments_static(input->info(), block_shape_x, block_shape_y, padding_left, padding_right, output->info()));
143 
144  _input = input;
145  _output = output;
146 
147  const DataLayout data_layout = input->info()->data_layout();
150  const int idx_batch = get_data_layout_dimension_index(data_layout, DataLayoutDimension::BATCHES);
151 
152  // Create kernel
153  CLBuildOptions build_opts;
154  build_opts.add_option("-DDATA_TYPE=" + get_cl_unsigned_type_from_element_size(data_size_from_type(input->info()->data_type())));
155  build_opts.add_option("-DWIDTH_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_width)));
156  build_opts.add_option("-DHEIGHT_OUT=" + support::cpp11::to_string(output->info()->dimension(idx_height)));
157  build_opts.add_option("-DBATCH_SIZE=" + support::cpp11::to_string(output->info()->dimension(idx_batch)));
158  build_opts.add_option("-DWIDTH_IN=" + support::cpp11::to_string(input->info()->dimension(idx_width)));
159  build_opts.add_option("-DHEIGHT_IN=" + support::cpp11::to_string(input->info()->dimension(idx_height)));
160  build_opts.add_option("-DBATCH_IN=" + support::cpp11::to_string(input->info()->dimension(idx_batch)));
161  build_opts.add_option("-DBLOCK_SHAPE_X=" + support::cpp11::to_string(block_shape_x));
162  build_opts.add_option("-DBLOCK_SHAPE_Y=" + support::cpp11::to_string(block_shape_y));
163  build_opts.add_option("-DPAD_LEFT_X=" + support::cpp11::to_string(padding_left.x()));
164  build_opts.add_option("-DPAD_RIGHT_X=" + support::cpp11::to_string(padding_right.x()));
165  build_opts.add_option("-DPAD_LEFT_Y=" + support::cpp11::to_string(padding_left.y()));
166  build_opts.add_option("-DPAD_RIGHT_Y=" + support::cpp11::to_string(padding_right.y()));
167  _kernel = create_kernel(compile_context, "space_to_batch_static_" + lower_string(string_from_data_layout(input->info()->data_layout())), build_opts.options());
168 
169  // Configure kernel window
170  Window win = calculate_max_window(*output->info(), Steps());
171  ICLKernel::configure_internal(win);
172 }
173 
174 Status CLSpaceToBatchLayerKernel::validate(const ITensorInfo *input, const ITensorInfo *block_shape, const ITensorInfo *paddings, const ITensorInfo *output)
175 {
176  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, block_shape, paddings, output));
177  return Status{};
178 }
179 Status CLSpaceToBatchLayerKernel::validate(const ITensorInfo *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right,
180  const ITensorInfo *output)
181 {
182  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments_static(input, block_shape_x, block_shape_y, padding_left, padding_right, output));
183  return Status{};
184 }
185 
186 void CLSpaceToBatchLayerKernel::run(const Window &window, cl::CommandQueue &queue)
187 {
190 
191  Window slice_out = window.first_slice_window_3D();
192 
193  Window slice_in = window.first_slice_window_4D();
194  slice_in.set(Window::DimX, Window::Dimension(0, 0, 0));
195  slice_in.set(Window::DimY, Window::Dimension(0, 0, 0));
196  slice_in.set(Window::DimZ, Window::Dimension(0, 0, 0));
197  slice_in.set(3, Window::Dimension(0, 0, 0));
198 
199  Window vector_slice = window.first_slice_window_1D();
200  vector_slice.set(Window::DimX, Window::Dimension(0, 0, 0));
201 
202  Window padding_slice = window.first_slice_window_2D();
203  padding_slice.set(Window::DimX, Window::Dimension(0, 0, 0));
204  padding_slice.set(Window::DimY, Window::Dimension(0, 0, 0));
205 
206  int batch_id = 0;
207  do
208  {
209  unsigned int idx = 0;
210  const bool cond = (_paddings != nullptr && _block_shape != nullptr);
211  add_4D_tensor_argument(idx, _input, slice_in);
212  add_2D_tensor_argument_if(cond, idx, _paddings, padding_slice);
213  add_1D_tensor_argument_if(cond, idx, _block_shape, vector_slice);
214 
215  add_argument(idx, batch_id);
216  add_3D_tensor_argument(idx, _output, slice_out);
217  enqueue(queue, *this, slice_out, lws_hint());
218  ++batch_id;
219  }
220  while(window.slide_window_slice_3D(slice_out));
221 }
222 } // namespace arm_compute
Window first_slice_window_2D() const
First 2D slice of the window.
Definition: Window.h:283
void add_1D_tensor_argument_if(bool cond, unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 1D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx ...
Definition: ICLKernel.h:135
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
Shape of a tensor.
Definition: TensorShape.h:39
void add_2D_tensor_argument_if(bool cond, unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 2D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx ...
Definition: ICLKernel.h:159
void enqueue(IGCKernel &kernel, const Window &window, const gles::NDRange &lws=gles::NDRange(1U, 1U, 1U))
Add the kernel to the command queue with the given window.
Definition: IGCKernel.cpp:41
virtual size_t dimension(size_t index) const =0
Return the size of the requested dimension.
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_QUANTIZATION_INFO(...)
Definition: Validate.h:610
const StringSet & options() const
Gets the current options list set.
cl::NDRange lws_hint() const
Return the Local-Workgroup-Size hint.
Definition: ICLKernel.h:276
#define ARM_COMPUTE_RETURN_ON_ERROR(status)
Checks if a status contains an error and returns it.
Definition: Error.h:204
std::string to_string(T &&value)
Convert integer and float values to string.
virtual DataType data_type() const =0
Data type used for each element of the tensor.
void add_argument(unsigned int &idx, T value)
Add the passed parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: ICLKernel.h:255
TensorShape compute_space_to_batch_shape(const ITensorInfo *input, const int block_x, const int block_y, const Size2D &padding_left, const Size2D &padding_right)
Calculate the space to batch output shape of a tensor.
#define ARM_COMPUTE_ERROR_ON(cond)
If the condition is true then an error message is printed and an exception thrown.
Definition: Error.h:466
const DataLayout data_layout
Definition: Im2Col.cpp:151
static CLKernelLibrary & get()
Access the KernelLibrary singleton.
Store the tensor&#39;s metadata.
Definition: ITensorInfo.h:40
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
Describe one of the image&#39;s dimensions with a start, end and step.
Definition: Window.h:77
void run(const Window &window, cl::CommandQueue &queue) override
Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue...
Status class.
Definition: Error.h:52
std::string lower_string(const std::string &val)
Lower a given string.
Definition: Utils.cpp:350
#define ARM_COMPUTE_RETURN_ERROR_ON(cond)
If the condition is true, an error is returned.
Definition: Error.h:296
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(...)
Definition: Validate.h:288
void add_3D_tensor_argument(unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 3D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: ICLKernel.h:172
Copyright (c) 2017-2021 Arm Limited.
#define ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(...)
Definition: Validate.h:163
1 channel, 1 S32 per channel
void add_option(std::string option)
Adds option to the existing build option list.
cl::Kernel create_kernel(const CLCompileContext &ctx, const std::string &kernel_name, const std::set< std::string > &build_opts=std::set< std::string >())
Creates an opencl kernel using a compile context.
Definition: CLHelpers.cpp:403
static constexpr size_t DimX
Alias for dimension 0 also known as X dimension.
Definition: Window.h:43
Class to describe a number of elements in each dimension.
Definition: Steps.h:40
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 ITensorInfo * info() const =0
Interface to be implemented by the child class to return the tensor&#39;s metadata.
size_t data_size_from_type(DataType data_type)
The size in bytes of the data type.
Definition: Utils.h:106
void set(size_t dimension, const Dimension &dim)
Set the values of a given dimension.
Definition: Window.inl:49
bool slide_window_slice_3D(Window &slice) const
Slide the passed 3D window slice.
Definition: Window.h:335
virtual QuantizationInfo quantization_info() const =0
Get the quantization settings (scale and offset) of the tensor.
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:941
bool has_padding_changed(const std::unordered_map< const ITensorInfo *, PaddingSize > &padding_map)
Check if the previously stored padding info has changed after configuring a kernel.
Definition: Utils.cpp:528
CLCompileContext class.
static constexpr size_t DimY
Alias for dimension 1 also known as Y dimension.
Definition: Window.h:45
Interface for OpenCL tensor.
Definition: ICLTensor.h:42
const std::string & string_from_data_layout(DataLayout dl)
Convert a data layout identity into a string.
Definition: Utils.cpp:123
static constexpr size_t DimZ
Alias for dimension 2 also known as Z dimension.
Definition: Window.h:47
Class for specifying the size of an image or rectangle.
Definition: Size2D.h:34
#define ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(...)
Definition: Validate.h:545
std::string get_cl_unsigned_type_from_element_size(size_t element_size)
Translates the element size to an unsigned integer data type.
Definition: CLHelpers.cpp:103
#define ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(t, c,...)
Definition: Validate.h:792
std::unordered_map< const ITensorInfo *, PaddingSize > get_padding_info(std::initializer_list< const ITensorInfo *> infos)
Stores padding information before configuring a kernel.
Definition: Utils.cpp:513
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const GEMMLowpOutputStageInfo *output_stage)
Window first_slice_window_4D() const
First 4D slice of the window.
Definition: Window.h:299
void configure(const ICLTensor *input, const ICLTensor *block_shape, const ICLTensor *paddings, ICLTensor *output)
Initialise the kernel&#39;s inputs and output.
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161
size_t get_data_layout_dimension_index(const DataLayout data_layout, const DataLayoutDimension data_layout_dimension)
Get the index of the given dimension.
Definition: Helpers.inl:193
Window first_slice_window_3D() const
First 3D slice of the window.
Definition: Window.h:291
void add_4D_tensor_argument(unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 4D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: ICLKernel.h:182
DataLayout
[DataLayout enum definition]
Definition: Types.h:120
Describe a multidimensional execution window.
Definition: Window.h:39
static Status validate(const ITensorInfo *input, const ITensorInfo *block_shape, const ITensorInfo *paddings, const ITensorInfo *output)
Static function to check if given info will lead to a valid configuration of CLSpaceToBatchLayerKerne...
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:205
Window first_slice_window_1D() const
First 1D slice of the window.
Definition: Window.h:275
virtual DataLayout data_layout() const =0
Get the data layout of the tensor.