NEBatchToSpaceLayerKernel Class Reference

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

#include <NEBatchToSpaceLayerKernel.h>

Collaboration diagram for NEBatchToSpaceLayerKernel:

Public Member Functions
const char *	name () const override
	Name of the kernel. More...
	NEBatchToSpaceLayerKernel ()
	Default constructor. More...
	NEBatchToSpaceLayerKernel (const NEBatchToSpaceLayerKernel &)=delete
	Prevent instances of this class from being copied (As this class contains pointers) More...
NEBatchToSpaceLayerKernel &	operator= (const NEBatchToSpaceLayerKernel &)=delete
	Prevent instances of this class from being copied (As this class contains pointers) More...
	NEBatchToSpaceLayerKernel (NEBatchToSpaceLayerKernel &&)=default
	Allow instances of this class to be moved. More...
NEBatchToSpaceLayerKernel &	operator= (NEBatchToSpaceLayerKernel &&)=default
	Allow instances of this class to be moved. More...
	~NEBatchToSpaceLayerKernel ()=default
	Default destructor. More...
void	configure (const ITensor *input, const ITensor *block_shape, ITensor *output)
	Initialise the kernel's inputs and output. More...
void	configure (const ITensor *input, int32_t block_shape_x, int32_t block_shape_y, ITensor *output, const CropInfo &crop_info=CropInfo{})
	Initialise the kernel's inputs and output (Static block shape). 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...
virtual size_t	get_mws (const CPUInfo &platform, size_t thread_count) const
	Return minimum workload size of the relevant kernel. 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 Window &	window () const
	The maximum window the kernel can be executed on. More...
bool	is_window_configured () const
	Function to check if the embedded window of this kernel has been configured. More...

Static Public Member Functions
static Status	validate (const ITensorInfo *input, const ITensorInfo *block_shape, const ITensorInfo *output)
	Static function to check if given info will lead to a valid configuration of NEBatchToSpaceLayerKernel. More...
static Status	validate (const ITensorInfo *input, int32_t block_shape_x, int32_t block_shape_y, const ITensorInfo *output, const CropInfo &crop_info=CropInfo{})
	Static function to check if given info will lead to a valid configuration of NEBatchToSpaceLayerKernel (Static block shape). More...

Additional Inherited Members
Static Public Attributes inherited from ICPPKernel
static constexpr size_t	default_mws = 1

Detailed Description

Interface for the batch to space kernel.

Definition at line 36 of file NEBatchToSpaceLayerKernel.h.

Constructor & Destructor Documentation

NEBatchToSpaceLayerKernel() [1/3]

NEBatchToSpaceLayerKernel

(

)

Default constructor.

Definition at line 88 of file NEBatchToSpaceLayerKernel.cpp.

    : _input(nullptr),
      _block_shape(nullptr),
      _output(nullptr),
      _data_layout(DataLayout::UNKNOWN),
      _block_shape_x(),
      _block_shape_y(),
      _crop_info()
{
}

References arm_compute::UNKNOWN.

NEBatchToSpaceLayerKernel() [2/3]

NEBatchToSpaceLayerKernel ( const NEBatchToSpaceLayerKernel &  )

delete

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

NEBatchToSpaceLayerKernel() [3/3]

NEBatchToSpaceLayerKernel ( NEBatchToSpaceLayerKernel &&  )

default

Allow instances of this class to be moved.

~NEBatchToSpaceLayerKernel()

~NEBatchToSpaceLayerKernel ( )

default

Default destructor.

Member Function Documentation

configure() [1/2]

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

Initialise the kernel's inputs and output.

Parameters

[in]	input	Tensor input. Supported tensor rank: 4. Data types supported: All.
[in]	block_shape	1-D tensor with shape [M]. Data types supported: S32
[out]	output	Tensor output. Data types supported: same as input

Deprecated:

This method for dynamic block shape is not fully mature and will be removed in 23.08 release

Definition at line 99 of file NEBatchToSpaceLayerKernel.cpp.

{
    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), block_shape->info(), output->info()));
 
    _input       = input;
    _block_shape = block_shape;
    _output      = output;
    _data_layout = input->info()->data_layout();
 
    // Configure kernel window
    Window win = calculate_max_window(*output->info(), Steps());
    ICPPKernel::configure(win);
}

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

configure() [2/2]

void configure	(	const ITensor *	input,
		int32_t	block_shape_x,
		int32_t	block_shape_y,
		ITensor *	output,
		const CropInfo &	crop_info = CropInfo{}
	)

Initialise the kernel's inputs and output (Static block shape).

Parameters

[in]	input	Tensor input. Supported tensor rank: 4. Data types supported: All.
[in]	block_shape_x	Block shape x value.
[in]	block_shape_y	Block shape y value.
[out]	output	Tensor output. Data types supported: same as input
[in]	crop_info	Specifies how the output shape is cropped after batch to space is performed

Definition at line 114 of file NEBatchToSpaceLayerKernel.cpp.

{
    ARM_COMPUTE_ERROR_ON_NULLPTR(input, output);
    const TensorShape output_shape = compute_batch_to_space_shape(
        input->info()->data_layout(), input->info()->tensor_shape(), block_shape_x, block_shape_y);
    // Output auto initialization if not yet initialized
    auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(output_shape));
 
    // Perform validation step
    ARM_COMPUTE_ERROR_THROW_ON(
        validate_arguments_static(input->info(), block_shape_x, block_shape_y, output->info(), crop_info));
 
    _input         = input;
    _output        = output;
    _block_shape_x = block_shape_x;
    _block_shape_y = block_shape_y;
    _data_layout   = input->info()->data_layout();
    _crop_info     = crop_info;
 
    // Configure kernel window
    Window win = calculate_max_window(*output->info(), Steps());
    ICPPKernel::configure(win);
}

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_batch_to_space_shape(), ITensor::info(), arm_compute::test::validation::input, and arm_compute::test::validation::output_shape.

name()

const char* name ( ) const

inlineoverridevirtual

Name of the kernel.

Returns

Kernel name

Implements ICPPKernel.

Definition at line 39 of file NEBatchToSpaceLayerKernel.h.

    {
        return "NEBatchToSpaceLayerKernel";
    }

operator=() [1/2]

NEBatchToSpaceLayerKernel& operator= ( const NEBatchToSpaceLayerKernel &  )

delete

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

operator=() [2/2]

NEBatchToSpaceLayerKernel& operator= ( NEBatchToSpaceLayerKernel &&  )

default

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]	window	Region on which to execute the kernel. (Must be a region of the window returned by window())
[in]	info	Info about executing thread and CPU.

Reimplemented from ICPPKernel.

Definition at line 158 of file NEBatchToSpaceLayerKernel.cpp.

{
    ARM_COMPUTE_UNUSED(info);
    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICPPKernel::window(), window);
 
    if (_block_shape != nullptr)
    {
        // Retrieve the block shapes dynamically
        _block_shape_x = *(reinterpret_cast<const int *>(_block_shape->ptr_to_element(0)));
        _block_shape_y = *(reinterpret_cast<const int *>(_block_shape->ptr_to_element(1)));
    }
 
    const int batch_size   = _output->info()->dimension(3);
    const int element_size = _output->info()->element_size();
 
    Window slice_out = window.first_slice_window_3D();
 
    int batch_id = 0;
    // Main loop for NCHW and NHWC
    if (_data_layout == DataLayout::NCHW)
    {
        do
        {
            Iterator out(_output, slice_out);
            execute_window_loop(
                slice_out,
                [&](const Coordinates &id)
                {
                    const int x = id.x();
                    const int y = id.y();
                    const int z = id.z();
                    // Translate x, y to uncropped version
                    const int x_c = x + _crop_info.left;
                    const int y_c = y + _crop_info.top;
 
                    const int in_batch =
                        batch_id + ((x_c % _block_shape_x) + (y_c % _block_shape_y) * _block_shape_x) * batch_size;
                    const int   in_x = x_c / _block_shape_x;
                    const int   in_y = y_c / _block_shape_y;
                    Coordinates input_coords{in_x, in_y, z, in_batch};
                    memcpy(out.ptr(), _input->ptr_to_element(input_coords), element_size);
                },
                out);
            ++batch_id;
        } while (window.slide_window_slice_3D(slice_out));
    }
    else
    {
        // For NHWC we can perform a block copy on the Channel (first) dimension. Thus we do not need to iterate over this dimension
        slice_out.set(0U, Window::Dimension(0U, 1U, 1U));
        do
        {
            Iterator out(_output, slice_out);
            execute_window_loop(
                slice_out,
                [&](const Coordinates &id)
                {
                    const int x = id.y();
                    const int y = id.z();
 
                    // Translate x, y to uncropped version
                    const int x_c = x + _crop_info.left;
                    const int y_c = y + _crop_info.top;
 
                    const int in_batch =
                        batch_id + ((x_c % _block_shape_x) + (y_c % _block_shape_y) * _block_shape_x) * batch_size;
                    const int   in_x = x_c / _block_shape_x;
                    const int   in_y = y_c / _block_shape_y;
                    Coordinates input_coords{0, in_x, in_y, in_batch};
                    memcpy(out.ptr(), _input->ptr_to_element(input_coords),
                           element_size * _input->info()->dimension(0));
                },
                out);
            ++batch_id;
        } while (window.slide_window_slice_3D(slice_out));
    }
}

References ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW, ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL, ARM_COMPUTE_UNUSED, ITensorInfo::dimension(), ITensorInfo::element_size(), arm_compute::execute_window_loop(), Window::first_slice_window_3D(), ITensor::info(), arm_compute::test::validation::info, Padding2D::left, arm_compute::NCHW, Iterator::ptr(), ITensor::ptr_to_element(), Window::set(), Window::slide_window_slice_3D(), Padding2D::top, arm_compute::utils::cast::U, IKernel::window(), Window::x(), and Window::y().

validate() [1/2]

Status validate ( const ITensorInfo *  input, const ITensorInfo *  block_shape, const ITensorInfo *  output  )

static

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

Parameters

[in]	input	Tensor input. Supported tensor rank: 4. Data types supported: All.
[in]	block_shape	1-D tensor with shape [M]. Data types supported: S32
[in]	output	Tensor output. Data types supported: same as input

Returns

a status

Deprecated:

This method for dynamic block shape is not fully mature and will be removed in 23.08 release

Definition at line 140 of file NEBatchToSpaceLayerKernel.cpp.

{
    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, block_shape, output);
    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, block_shape, output));
    return Status{};
}

References ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR, ARM_COMPUTE_RETURN_ON_ERROR, arm_compute::test::validation::input, and arm_compute::cpu::kernels::validate_arguments().

Referenced by NEBatchToSpaceLayer::validate().

validate() [2/2]

Status validate ( const ITensorInfo *  input, int32_t  block_shape_x, int32_t  block_shape_y, const ITensorInfo *  output, const CropInfo &  crop_info = CropInfo{}  )

static

Static function to check if given info will lead to a valid configuration of NEBatchToSpaceLayerKernel (Static block shape).

Parameters

[in]	input	Tensor input. Supported tensor rank: 4. Data types supported: All.
[in]	block_shape_x	Block shape x value.
[in]	block_shape_y	Block shape y value.
[in]	output	Tensor output. Data types supported: same as input
[in]	crop_info	Specifies how the output shape is cropped after batch to space is performed

Returns

a status

Definition at line 147 of file NEBatchToSpaceLayerKernel.cpp.

{
    ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output);
    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments_static(input, block_shape_x, block_shape_y, output, crop_info));
    return Status{};
}

References ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR, ARM_COMPUTE_RETURN_ON_ERROR, and arm_compute::test::validation::input.

---

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

• src/core/NEON/kernels/NEBatchToSpaceLayerKernel.h
• src/core/NEON/kernels/NEBatchToSpaceLayerKernel.cpp