Compute Library
 21.02
NESpaceToBatchLayerKernel Class Reference

Interface for the space to batch kernel. More...

#include <NESpaceToBatchLayerKernel.h>

Collaboration diagram for NESpaceToBatchLayerKernel:
[legend]

Public Member Functions

const char * name () const override
 Name of the kernel. More...
 
 NESpaceToBatchLayerKernel ()
 Default constructor. More...
 
 NESpaceToBatchLayerKernel (const NESpaceToBatchLayerKernel &)=delete
 Prevent instances of this class from being copied (As this class contains pointers) More...
 
NESpaceToBatchLayerKerneloperator= (const NESpaceToBatchLayerKernel &)=delete
 Prevent instances of this class from being copied (As this class contains pointers) More...
 
 NESpaceToBatchLayerKernel (NESpaceToBatchLayerKernel &&)=default
 Allow instances of this class to be moved. More...
 
NESpaceToBatchLayerKerneloperator= (NESpaceToBatchLayerKernel &&)=default
 Allow instances of this class to be moved. More...
 
 ~NESpaceToBatchLayerKernel ()=default
 Default destructor. More...
 
void configure (const ITensor *input, const ITensor *block_shape, const ITensor *paddings, ITensor *output)
 Initialise the kernel's inputs and output. More...
 
void configure (const ITensor *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right, ITensor *output)
 Initialise the kernel's input and output. More...
 
void run (const Window &window, const ThreadInfo &info) override
 Execute the kernel on the passed window. More...
 
- Public Member Functions inherited from ICPPKernel
virtual ~ICPPKernel ()=default
 Default destructor. More...
 
virtual void run_nd (const Window &window, const ThreadInfo &info, const Window &thread_locator)
 legacy compatibility layer for implemantions which do not support thread_locator In these cases we simply narrow the interface down the legacy version More...
 
virtual void run_op (ITensorPack &tensors, const Window &window, const ThreadInfo &info)
 Execute the kernel on the passed window. More...
 
- Public Member Functions inherited from IKernel
 IKernel ()
 Constructor. More...
 
virtual ~IKernel ()=default
 Destructor. More...
 
virtual bool is_parallelisable () const
 Indicates whether or not the kernel is parallelisable. More...
 
virtual BorderSize border_size () const
 The size of the border for that kernel. More...
 
const Windowwindow () const
 The maximum window the kernel can be executed on. More...
 

Static Public Member Functions

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 NESpaceToBatchLayerKernel. More...
 
static Status validate (const ITensorInfo *input, const int block_shape_x, const int block_shape_y, const Size2D &padding_left, const Size2D &padding_right, const ITensorInfo *output)
 Static function to check if given info will lead to a valid configuration of NESpaceToBatchLayerKernel (Static block shape and paddings) More...
 

Detailed Description

Interface for the space to batch kernel.

Definition at line 36 of file NESpaceToBatchLayerKernel.h.

Constructor & Destructor Documentation

◆ NESpaceToBatchLayerKernel() [1/3]

Default constructor.

Definition at line 88 of file NESpaceToBatchLayerKernel.cpp.

Referenced by NESpaceToBatchLayerKernel::name().

89  : _input(nullptr), _block_shape(nullptr), _paddings(nullptr), _output(nullptr), _data_layout(DataLayout::UNKNOWN), _padding_left(), _block_shape_x(), _block_shape_y()
90 {
91 }

◆ NESpaceToBatchLayerKernel() [2/3]

Prevent instances of this class from being copied (As this class contains pointers)

◆ NESpaceToBatchLayerKernel() [3/3]

Allow instances of this class to be moved.

◆ ~NESpaceToBatchLayerKernel()

Default destructor.

Referenced by NESpaceToBatchLayerKernel::name().

Member Function Documentation

◆ configure() [1/2]

void configure ( const ITensor input,
const ITensor block_shape,
const ITensor paddings,
ITensor output 
)

Initialise the kernel's inputs and output.

Parameters
[in]inputTensor input. Supported tensor rank: 4. Data types supported: All.
[in]block_shape1-D tensor with shape [M]. Supported M: 2. Data types supported: S32
[in]paddings2-D tensor with shape [2, M] (First dimension is the fastest-changing dimension). Supported M: 2. Data types supported: S32
[out]outputTensor output. Data types supported: same as input

Definition at line 93 of file NESpaceToBatchLayerKernel.cpp.

References ARM_COMPUTE_ERROR_ON_NULLPTR, ARM_COMPUTE_ERROR_THROW_ON, arm_compute::calculate_max_window(), ITensorInfo::data_layout(), ITensor::info(), arm_compute::test::validation::input, and arm_compute::validate_arguments().

Referenced by NESpaceToBatchLayerKernel::name().

94 {
95  ARM_COMPUTE_ERROR_ON_NULLPTR(input, block_shape, paddings, output);
96  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), block_shape->info(), paddings->info(), output->info()));
97 
98  _input = input;
99  _block_shape = block_shape;
100  _paddings = paddings;
101  _output = output;
102  _data_layout = input->info()->data_layout();
103 
104  // Configure kernel window
105  Window win = calculate_max_window(*output->info(), Steps());
106  ICPPKernel::configure(win);
107 }
Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps, bool skip_border, BorderSize border_size)
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
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

◆ configure() [2/2]

void configure ( const ITensor input,
const int  block_shape_x,
const int  block_shape_y,
const Size2D padding_left,
const Size2D padding_right,
ITensor output 
)

Initialise the kernel's input and output.

(Static block shape and paddings)

Parameters
[in]inputTensor input. Supported tensor rank: 4. Data types supported: All.
[in]block_shape_xBlock shape x value.
[in]block_shape_yBlock shape y value.
[in]padding_leftThe padding at the beginning of every dimension of the output tensor.
[in]padding_rightThe padding at the end of every dimension of the output tensor.
[out]outputTensor output. Data types supported: same as input

Definition at line 109 of file NESpaceToBatchLayerKernel.cpp.

References ARM_COMPUTE_ERROR_ON_NULLPTR, ARM_COMPUTE_ERROR_THROW_ON, arm_compute::auto_init_if_empty(), arm_compute::calculate_max_window(), arm_compute::misc::shape_calculator::compute_space_to_batch_shape(), ITensorInfo::data_layout(), ITensorInfo::data_type(), ITensor::info(), arm_compute::test::validation::input, arm_compute::test::validation::output_shape, and ITensorInfo::quantization_info().

111 {
113 
114  TensorShape output_shape = misc::shape_calculator::compute_space_to_batch_shape(input->info(), block_shape_x, block_shape_y, padding_left, padding_right);
115  auto_init_if_empty(*output->info(), output_shape, 1, input->info()->data_type(), input->info()->quantization_info());
116 
117  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments_static(input->info(), block_shape_x, block_shape_y, padding_left, padding_right, output->info()));
118 
119  _input = input;
120  _output = output;
121  _block_shape_x = block_shape_x;
122  _block_shape_y = block_shape_y;
123  _padding_left = padding_left;
124  _data_layout = input->info()->data_layout();
125 
126  // Configure kernel window
127  Window win = calculate_max_window(*output->info(), Steps());
128  INEKernel::configure(win);
129 }
Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps, bool skip_border, BorderSize border_size)
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_THROW_ON(status)
Definition: Error.h:455
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...
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161

◆ name()

◆ operator=() [1/2]

NESpaceToBatchLayerKernel& operator= ( const NESpaceToBatchLayerKernel )
delete

Prevent instances of this class from being copied (As this class contains pointers)

Referenced by NESpaceToBatchLayerKernel::name().

◆ operator=() [2/2]

Allow instances of this class to be moved.

◆ run()

void run ( const Window window,
const ThreadInfo info 
)
overridevirtual

Execute the kernel on the passed window.

Warning
If is_parallelisable() returns false then the passed window must be equal to window()
Note
The window has to be a region within the window returned by the window() method
The width of the window has to be a multiple of num_elems_processed_per_iteration().
Parameters
[in]windowRegion on which to execute the kernel. (Must be a region of the window returned by window())
[in]infoInfo about executing thread and CPU.

Reimplemented from ICPPKernel.

Definition at line 143 of file NESpaceToBatchLayerKernel.cpp.

References ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW, ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL, ARM_COMPUTE_UNUSED, arm_compute::BATCHES, ITensorInfo::dimension(), ITensorInfo::element_size(), arm_compute::execute_window_loop(), Window::first_slice_window_3D(), arm_compute::get_data_layout_dimension_index(), arm_compute::HEIGHT, ITensor::info(), arm_compute::NCHW, Iterator::ptr(), ITensor::ptr_to_element(), Window::slide_window_slice_3D(), arm_compute::test::validation::w, arm_compute::WIDTH, IKernel::window(), Size2D::x(), Window::x(), Size2D::y(), and Window::y().

Referenced by NESpaceToBatchLayerKernel::name().

144 {
148 
149  if(_block_shape != nullptr)
150  {
151  // Retrieve the block shapes dynamically
152  _block_shape_x = *(reinterpret_cast<const int *>(_block_shape->ptr_to_element(0)));
153  _block_shape_y = *(reinterpret_cast<const int *>(_block_shape->ptr_to_element(1)));
154  }
155 
156  if(_paddings != nullptr)
157  {
158  const size_t pad_left_x = *reinterpret_cast<const size_t *>(_paddings->ptr_to_element({ 0, 0 }));
159  const size_t pad_left_y = *reinterpret_cast<const size_t *>(_paddings->ptr_to_element({ 1, 0 }));
160  _padding_left = Size2D(pad_left_x, pad_left_y);
161  }
162  const int height_idx = get_data_layout_dimension_index(_data_layout, DataLayoutDimension::HEIGHT);
163  const int width_idx = get_data_layout_dimension_index(_data_layout, DataLayoutDimension::WIDTH);
164  const int batch_idx = get_data_layout_dimension_index(_data_layout, DataLayoutDimension::BATCHES);
165  const int element_size = _input->info()->element_size();
166 
167  const size_t height = _input->info()->dimension(height_idx);
168  const size_t width = _input->info()->dimension(width_idx);
169  const size_t batch_size = _input->info()->dimension(batch_idx);
170 
171  Window slice_out = window.first_slice_window_3D();
172 
173  int batch_id = 0;
174 
175  // Main loop for NCHW and NHWC
176  if(_data_layout == DataLayout::NCHW)
177  {
178  do
179  {
180  Iterator out(_output, slice_out);
181  execute_window_loop(slice_out, [&](const Coordinates & id)
182  {
183  const size_t out_x = id.x();
184  const size_t out_y = id.y();
185  const size_t z = id.z();
186  const size_t pos_x = out_x * _block_shape_x + (batch_id / batch_size) % _block_shape_x;
187  const size_t pos_y = out_y * _block_shape_y + (batch_id / batch_size) / _block_shape_x;
188  if(pos_y >= _padding_left.y() && pos_y < _padding_left.y() + height && pos_x >= _padding_left.x() && pos_x < _padding_left.x() + width)
189  {
190  const int w = batch_id % batch_size;
191  const int in_x = pos_x - _padding_left.x();
192  const int in_y = pos_y - _padding_left.y();
193  Coordinates input_coords{ in_x, in_y, z, w };
194  memcpy(out.ptr(), _input->ptr_to_element(input_coords), element_size);
195  }
196  },
197  out);
198  ++batch_id;
199  }
200  while(window.slide_window_slice_3D(slice_out));
201  }
202  else
203  {
204  do
205  {
206  Iterator out(_output, slice_out);
207  execute_window_loop(slice_out, [&](const Coordinates & id)
208  {
209  const size_t out_x = id.y();
210  const size_t out_y = id.z();
211  const size_t z = id.x();
212  const size_t pos_x = out_x * _block_shape_x + (batch_id / batch_size) % _block_shape_x;
213  const size_t pos_y = out_y * _block_shape_y + (batch_id / batch_size) / _block_shape_x;
214  if(pos_y >= _padding_left.y() && pos_y < _padding_left.y() + height && pos_x >= _padding_left.x() && pos_x < _padding_left.x() + width)
215  {
216  const int w = batch_id % batch_size;
217  const int in_x = pos_x - _padding_left.x();
218  const int in_y = pos_y - _padding_left.y();
219  Coordinates input_coords{ z, in_x, in_y, w };
220  memcpy(out.ptr(), _input->ptr_to_element(input_coords), element_size);
221  }
222  },
223  out);
224  ++batch_id;
225  }
226  while(window.slide_window_slice_3D(slice_out));
227  }
228 }
SimpleTensor< float > w
Definition: DFT.cpp:156
const Window & window() const
The maximum window the kernel can be executed on.
Definition: IKernel.cpp:28
uint8_t * ptr_to_element(const Coordinates &id) const
Return a pointer to the element at the passed coordinates.
Definition: ITensor.h:63
virtual size_t dimension(size_t index) const =0
Return the size of the requested dimension.
size_t x() const
Semantic accessor for width as x.
Definition: Size2D.h:74
#define ARM_COMPUTE_UNUSED(...)
To avoid unused variables warnings.
Definition: Error.h:152
virtual ITensorInfo * info() const =0
Interface to be implemented by the child class to return the tensor&#39;s metadata.
virtual size_t element_size() const =0
Element size in bytes calculated as data_size() * num_channels()
bool slide_window_slice_3D(Window &slice) const
Slide the passed 3D window slice.
Definition: Window.h:335
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:941
Num samples, channels, height, width.
size_t y() const
Semantic accessor for height as y.
Definition: Size2D.h:83
ScaleKernelInfo info(interpolation_policy, default_border_mode, PixelValue(), sampling_policy, false)
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
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
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:205

◆ validate() [1/2]

Status validate ( const ITensorInfo input,
const ITensorInfo block_shape,
const ITensorInfo paddings,
const ITensorInfo output 
)
static

Static function to check if given info will lead to a valid configuration of NESpaceToBatchLayerKernel.

Parameters
[in]inputTensor input. Supported tensor rank: 4. Data types supported: All.
[in]block_shape1-D tensor with shape [M]. Supported M: 2. Data types supported: S32
[in]paddings2-D tensor with shape [2, M] (First dimension is the fastest-changing dimension). Supported M: 2. Data types supported: S32
[in]outputTensor output. Data types supported: same as input
Returns
a status

Definition at line 131 of file NESpaceToBatchLayerKernel.cpp.

References ARM_COMPUTE_RETURN_ON_ERROR, and arm_compute::validate_arguments().

Referenced by NESpaceToBatchLayerKernel::name(), and NESpaceToBatchLayer::validate().

132 {
133  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, block_shape, paddings, output));
134  return Status{};
135 }
#define ARM_COMPUTE_RETURN_ON_ERROR(status)
Checks if a status contains an error and returns it.
Definition: Error.h:204
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const GEMMLowpOutputStageInfo *output_stage)

◆ validate() [2/2]

Status validate ( const ITensorInfo input,
const int  block_shape_x,
const int  block_shape_y,
const Size2D padding_left,
const Size2D padding_right,
const ITensorInfo output 
)
static

Static function to check if given info will lead to a valid configuration of NESpaceToBatchLayerKernel (Static block shape and paddings)

Parameters
[in]inputTensor input. Supported tensor rank: 4. Data types supported: All.
[in]block_shape_xBlock shape x value.
[in]block_shape_yBlock shape y value.
[in]padding_leftThe padding at the beginning of every dimension of the output tensor.
[in]padding_rightThe padding at the end of every dimension of the output tensor.
[in]outputTensor output. Data types supported: same as input
Returns
a status

Definition at line 136 of file NESpaceToBatchLayerKernel.cpp.

References ARM_COMPUTE_RETURN_ON_ERROR.

138 {
139  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments_static(input, block_shape_x, block_shape_y, padding_left, padding_right, output));
140  return Status{};
141 }
#define ARM_COMPUTE_RETURN_ON_ERROR(status)
Checks if a status contains an error and returns it.
Definition: Error.h:204

The documentation for this class was generated from the following files: