ClScaleKernel Class Reference

Interface for the scale kernel. More...

#include <ClScaleKernel.h>

Collaboration diagram for ClScaleKernel:

Public Member Functions
	ClScaleKernel ()
	ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE (ClScaleKernel)
void	configure (const CLCompileContext &compile_context, ITensorInfo *src, ITensorInfo *dst, const ScaleKernelInfo &info)
	Initialise the kernel's inputs, output and interpolation policy. More...
void	run_op (ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override
	Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue. More...
Public Member Functions inherited from ICLKernel
	ICLKernel ()
	Constructor. More...
cl::Kernel &	kernel ()
	Returns a reference to the OpenCL kernel of this object. More...
CLKernelType	type () const
	Returns the CL kernel type. More...
template<typename T >
void	add_1D_array_argument (unsigned int &idx, const ICLArray< T > *array, const Strides &strides, unsigned int num_dimensions, const Window &window)
	Add the passed 1D array's parameters to the object's kernel's arguments starting from the index idx. More...
void	add_1D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, const Window &window)
	Add the passed 1D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
void	add_1D_tensor_argument_if (bool cond, unsigned int &idx, const ICLTensor *tensor, const Window &window)
	Add the passed 1D tensor's parameters to the object's kernel's arguments starting from the index idx if the condition is true. More...
void	add_2D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, const Window &window)
	Add the passed 2D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
void	add_2D_tensor_argument_if (bool cond, unsigned int &idx, const ICLTensor *tensor, const Window &window)
	Add the passed 2D tensor's parameters to the object's kernel's arguments starting from the index idx if the condition is true. More...
void	add_3D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, const Window &window)
	Add the passed 3D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
void	add_4D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, const Window &window)
	Add the passed 4D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
void	add_5D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, const Window &window)
	Add the passed 5D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
void	add_3d_tensor_nhw_argument (unsigned int &idx, const ICLTensor *tensor)
	Add the passed NHW 3D tensor's parameters to the object's kernel's arguments by passing strides, dimensions and the offset to the first valid element in bytes. More...
void	add_4d_tensor_nhwc_argument (unsigned int &idx, const ICLTensor *tensor)
	Add the passed NHWC 4D tensor's parameters to the object's kernel's arguments by passing strides, dimensions and the offset to the first valid element in bytes. More...
virtual void	run (const Window &window, cl::CommandQueue &queue)
	Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue. More...
virtual void	run_composite_op (ITensorPack &tensors, const Window &window, cl::CommandQueue &queue, const experimental::dynamic_fusion::ClExecutionDescriptor &exec_desc)
	The execution is carried out through run_op method. But the run_op method needs to be extended to include ClExecutionDescriptor as now LWS GWS tuning will be separated from the IKernel. More...
template<typename T >
void	add_argument (unsigned int &idx, T value)
	Add the passed parameters to the object's kernel's arguments starting from the index idx. More...
void	set_lws_hint (const cl::NDRange &lws_hint)
	Set the Local-Workgroup-Size hint. More...
cl::NDRange	lws_hint () const
	Return the Local-Workgroup-Size hint. More...
void	set_wbsm_hint (const cl_int &wbsm_hint)
	Set the workgroup batch size modifier hint. More...
cl_int	wbsm_hint () const
	Return the workgroup batch size modifier hint. More...
const std::string &	config_id () const
	Get the configuration ID. More...
void	set_target (GPUTarget target)
	Set the targeted GPU architecture. More...
void	set_target (cl::Device &device)
	Set the targeted GPU architecture according to the CL device. More...
GPUTarget	get_target () const
	Get the targeted GPU architecture. More...
size_t	get_max_workgroup_size ()
	Get the maximum workgroup size for the device the CLKernelLibrary uses. More...
template<unsigned int dimension_size>
void	add_tensor_argument (unsigned &idx, const ICLTensor *tensor, const Window &window)
template<typename T , unsigned int dimension_size>
void	add_array_argument (unsigned &idx, const ICLArray< T > *array, const Strides &strides, unsigned int num_dimensions, const Window &window)
	Add the passed array's parameters to the object's kernel's arguments starting from the index idx. 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 *src, const ITensorInfo *dst, const ScaleKernelInfo &info)
	Static function to check if given info will lead to a valid configuration. More...
Static Public Member Functions inherited from ICLKernel
static constexpr unsigned int	num_arguments_per_3d_tensor_nhw ()
	Returns the number of arguments enqueued per NHW 3D Tensor object. More...
static constexpr unsigned int	num_arguments_per_4d_tensor_nhwc ()
	Returns the number of arguments enqueued per NHWC 4D Tensor object. More...
static constexpr unsigned int	num_arguments_per_1D_array ()
	Returns the number of arguments enqueued per 1D array object. More...
static constexpr unsigned int	num_arguments_per_1D_tensor ()
	Returns the number of arguments enqueued per 1D tensor object. More...
static constexpr unsigned int	num_arguments_per_2D_tensor ()
	Returns the number of arguments enqueued per 2D tensor object. More...
static constexpr unsigned int	num_arguments_per_3D_tensor ()
	Returns the number of arguments enqueued per 3D tensor object. More...
static constexpr unsigned int	num_arguments_per_4D_tensor ()
	Returns the number of arguments enqueued per 4D tensor object. More...
static cl::NDRange	gws_from_window (const Window &window)
	Get the global work size given an execution window. More...

Detailed Description

Interface for the scale kernel.

Definition at line 39 of file ClScaleKernel.h.

Constructor & Destructor Documentation

ClScaleKernel()

ClScaleKernel

(

)

Definition at line 87 of file ClScaleKernel.cpp.

References arm_compute::ELEMENTWISE.

{
    _type = CLKernelType::ELEMENTWISE;
}

Member Function Documentation

ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE()

ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE

(

ClScaleKernel

)

configure()

void configure	(	const CLCompileContext &	compile_context,
		ITensorInfo *	src,
		ITensorInfo *	dst,
		const ScaleKernelInfo &	info
	)

Initialise the kernel's inputs, output and interpolation policy.

Parameters

[in]	compile_context	The compile context to be used.
[in]	src	Source tensor info. Data types supported: U8/QASYMM8/QASYMM8_SIGNED/S16/F16/F32
[out]	dst	Destination tensor info. Data types supported: Same as src All but the lowest two dimensions must be the same size as in the input tensor, i.e. scaling is only performed within the XY-plane.
[in]	info	ScaleKernelInfo Kernel descriptor to be used to configure.

Definition at line 92 of file ClScaleKernel.cpp.

References CLBuildOptions::add_option(), CLBuildOptions::add_option_if(), CLBuildOptions::add_option_if_else(), arm_compute::adjust_vec_size(), ScaleKernelInfo::align_corners, arm_compute::AREA, ARM_COMPUTE_ERROR_ON, ARM_COMPUTE_ERROR_THROW_ON, arm_compute::BILINEAR, ScaleKernelInfo::border_mode, arm_compute::calculate_max_window(), arm_compute::CENTER, arm_compute::CHANNEL, arm_compute::CONSTANT, ScaleKernelInfo::constant_border_value, arm_compute::create_kernel(), ScaleKernelInfo::data_layout, ITensorInfo::data_layout(), ITensorInfo::data_type(), ITensorInfo::dimension(), arm_compute::F32, arm_compute::float_to_string_with_full_precision(), arm_compute::get_cl_type_from_data_type(), arm_compute::get_data_layout_dimension_index(), arm_compute::get_padding_info(), arm_compute::has_padding_changed(), arm_compute::HEIGHT, ScaleKernelInfo::interpolation_policy, arm_compute::is_data_type_float(), arm_compute::is_data_type_quantized_asymmetric(), kernel_name, arm_compute::lower_string(), arm_compute::NCHW, arm_compute::NEAREST_NEIGHBOR, arm_compute::NHWC, ICLKernel::num_arguments_per_4d_tensor_nhwc(), ITensorInfo::num_dimensions(), UniformQuantizationInfo::offset, CLBuildOptions::options(), arm_compute::test::validation::qinfo, ITensorInfo::quantization_info(), arm_compute::REPLICATE, ScaleKernelInfo::sampling_policy, UniformQuantizationInfo::scale, arm_compute::test::validation::scale_x, arm_compute::test::validation::scale_y, arm_compute::test::validation::src, arm_compute::string_from_data_layout(), arm_compute::string_from_interpolation_policy(), arm_compute::string_from_pixel_value(), arm_compute::support::cpp11::to_string(), arm_compute::utility::tolower(), QuantizationInfo::uniform(), arm_compute::UNKNOWN, arm_compute::cpu::kernels::validate_arguments(), and arm_compute::WIDTH.

{
    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, dst, info));
    auto padding_info = get_padding_info({ src, dst });

    // Info required for the static tuning
    _data_layout = info.data_layout == DataLayout::UNKNOWN ? src->data_layout() : info.data_layout;

    const bool is_nhwc = _data_layout == DataLayout::NHWC;

    float scale_x = 0.f;
    float scale_y = 0.f;
    std::tie(scale_x, scale_y) = calculate_scale_factors(src, dst, _data_layout, info.align_corners);
    const bool is_qasymm_bilinear = is_data_type_quantized_asymmetric(src->data_type()) && info.interpolation_policy == InterpolationPolicy::BILINEAR;

    // Area interpolation behaves as Nearest Neighbour in case of up-sampling
    auto interpolation_policy_to_use = info.interpolation_policy;
    if(info.interpolation_policy == InterpolationPolicy::AREA && scale_x <= 1.f && scale_y <= 1.f)
    {
        interpolation_policy_to_use = InterpolationPolicy::NEAREST_NEIGHBOR;
    }

    // Create kernel
    const int          idx_width         = get_data_layout_dimension_index(_data_layout, DataLayoutDimension::WIDTH);
    const int          idx_height        = get_data_layout_dimension_index(_data_layout, DataLayoutDimension::HEIGHT);
    const int          idx_channel       = get_data_layout_dimension_index(_data_layout, DataLayoutDimension::CHANNEL);
    const unsigned int src_width         = src->dimension(idx_width);
    const unsigned int src_height        = src->dimension(idx_height);
    const unsigned int dst_width         = dst->dimension(idx_width);
    const unsigned int dst_channels      = dst->dimension(idx_channel);
    unsigned int       vec_size          = 0;
    unsigned int       vec_size_leftover = 0;

    CLBuildOptions build_opts;
    if(_data_layout == DataLayout::NHWC)
    {
        vec_size          = adjust_vec_size(src->data_type() == DataType::F32 ? 4 : 8, dst_channels);
        vec_size_leftover = dst_channels % vec_size;
        build_opts.add_option("-DSRC_TENSOR_TYPE=BUFFER");
        build_opts.add_option("-DSRC_DATA_TYPE=" + get_cl_type_from_data_type(src->data_type()));
        build_opts.add_option("-DDST_TENSOR_TYPE=BUFFER");
        build_opts.add_option("-DDST_DATA_TYPE=" + get_cl_type_from_data_type(dst->data_type()));
        build_opts.add_option("-DCONSTANT_VALUE=" + string_from_pixel_value(info.constant_border_value, src->data_type()));
        build_opts.add_option("-DN0=" + support::cpp11::to_string(vec_size));
        build_opts.add_option("-DPARTIAL_N0=" + support::cpp11::to_string(vec_size_leftover));
        build_opts.add_option("-DSCALE_" + string_from_interpolation_policy(interpolation_policy_to_use));
        build_opts.add_option_if(src->num_dimensions() > 3, "-DBATCHED_EXECUTION");
        build_opts.add_option_if(info.border_mode == BorderMode::REPLICATE, "-DBORDER_MODE_REPLICATE");
        build_opts.add_option_if(info.border_mode == BorderMode::CONSTANT, "-DBORDER_MODE_CONSTANT");
        build_opts.add_option_if(info.align_corners, "-DALIGN_CORNERS");
        build_opts.add_option_if(is_data_type_float(src->data_type()), "-DIS_FLOATING_POINT");
        build_opts.add_option_if_else(info.sampling_policy == SamplingPolicy::CENTER, "-DSAMPLING_POLICY_CENTER", "-DSAMPLING_POLICY_TOP_LEFT");
        if(is_qasymm_bilinear)
        {
            const UniformQuantizationInfo qinfo = src->quantization_info().uniform();
            build_opts.add_option("-DSCALE=" + support::cpp11::to_string(qinfo.scale));
            build_opts.add_option("-DOFFSET=" + support::cpp11::to_string(qinfo.offset));
        }
        else
        {
            build_opts.add_option("-DSCALE=" + support::cpp11::to_string(1));
            build_opts.add_option("-DOFFSET=" + support::cpp11::to_string(0));
        }
    }
    else if(_data_layout == DataLayout::NCHW)
    {
        vec_size          = adjust_vec_size(4, dst_width);
        vec_size_leftover = dst_width % vec_size;
        build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(src->data_type()));
        build_opts.add_option("-DCONSTANT_VALUE=" + string_from_pixel_value(info.constant_border_value, src->data_type()));
        build_opts.add_option("-DSRC_WIDTH=" + support::cpp11::to_string(src_width));
        build_opts.add_option("-DSRC_HEIGHT=" + support::cpp11::to_string(src_height));
        build_opts.add_option("-DSCALE_X=" + float_to_string_with_full_precision(scale_x));
        build_opts.add_option("-DSCALE_Y=" + float_to_string_with_full_precision(scale_y));
        build_opts.add_option("-DVEC_SIZE=" + support::cpp11::to_string(vec_size));
        build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + ((vec_size_leftover == 0) ? support::cpp11::to_string(vec_size) : support::cpp11::to_string(vec_size_leftover)));
        build_opts.add_option_if(info.border_mode == BorderMode::REPLICATE, "-DBORDER_MODE_REPLICATE");
        build_opts.add_option_if(info.border_mode == BorderMode::CONSTANT, "-DBORDER_MODE_CONSTANT");
        build_opts.add_option_if(info.align_corners, "-DALIGN_CORNERS");
        build_opts.add_option_if_else(info.sampling_policy == SamplingPolicy::CENTER, "-DSAMPLING_POLICY_CENTER", "-DSAMPLING_POLICY_TOP_LEFT");
        if(is_qasymm_bilinear)
        {
            const UniformQuantizationInfo qinfo = src->quantization_info().uniform();
            build_opts.add_option("-DSCALE=" + support::cpp11::to_string(qinfo.scale));
            build_opts.add_option("-DOFFSET=" + support::cpp11::to_string(qinfo.offset));
        }
    }
    else
    {
        ARM_COMPUTE_ERROR_ON("Unsupported data layout");
    }

    std::string interpolation_name = string_from_interpolation_policy(interpolation_policy_to_use);
    std::transform(interpolation_name.begin(), interpolation_name.end(), interpolation_name.begin(), ::tolower);
    std::string kernel_name = "scale_" + interpolation_name + "_";
    kernel_name += lower_string(string_from_data_layout(_data_layout));

    _kernel = create_kernel(compile_context, kernel_name, build_opts.options());

    // Configure kernel window
    Window win = calculate_max_window(*dst, Steps(vec_size));
    ICLKernel::configure_internal(win);

    ARM_COMPUTE_ERROR_ON(has_padding_changed(padding_info));

    // Pass scale kernel arguments
    if(is_nhwc)
    {
        unsigned int idx = 2 * num_arguments_per_4d_tensor_nhwc();
        _kernel.setArg<cl_float>(idx++, scale_x);
        _kernel.setArg<cl_float>(idx++, scale_y);
    }
    // Set config_id for enabling LWS tuning
    _config_id = "scale_";
    _config_id += (info.border_mode == BorderMode::REPLICATE ? "Bord_rep" : "");
    _config_id += (info.sampling_policy == SamplingPolicy::CENTER ? "center" : "topleft");
    _config_id += (is_nhwc ? "nhwc" : "nchw");
    _config_id += "_";
    _config_id += support::cpp11::to_string(dst->dimension(0));
    _config_id += "_";
    _config_id += support::cpp11::to_string(dst->dimension(1));
    _config_id += "_";
    _config_id += support::cpp11::to_string(dst->dimension(2));
    _config_id += "_";
    _config_id += support::cpp11::to_string(dst->dimension(3));
}

run_op()

void run_op ( ITensorPack &  tensors, const Window &  window, cl::CommandQueue &  queue  )

overridevirtual

Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue.

Note

The queue is not flushed by this method, and therefore the kernel will not have been executed by the time this method returns.

Parameters

[in]	tensors	A vector containing the tensors to operato on.
[in]	window	Region on which to execute the kernel. (Must be a valid region of the window returned by window()).
[in,out]	queue	Command queue on which to enqueue the kernel.

Reimplemented from ICLKernel.

Definition at line 219 of file ClScaleKernel.cpp.

References arm_compute::ACL_DST, arm_compute::ACL_SRC, ICLKernel::add_2D_tensor_argument(), ICLKernel::add_4d_tensor_nhwc_argument(), ARM_COMPUTE_ERROR, ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW, ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL, Window::collapse(), Window::DimZ, arm_compute::enqueue(), Window::first_slice_window_2D(), Window::first_slice_window_4D(), ITensorPack::get_const_tensor(), ITensorPack::get_tensor(), ICLKernel::lws_hint(), arm_compute::NCHW, arm_compute::NHWC, arm_compute::test::validation::reference::slice(), Window::slide_window_slice_2D(), and IKernel::window().

{
    ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
    ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICLKernel::window(), window);

    auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
    auto dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));

    switch(_data_layout)
    {
        case DataLayout::NCHW:
        {
            Window slice = window.first_slice_window_2D();

            do
            {
                unsigned int idx = 0;
                add_2D_tensor_argument(idx, src, slice);
                add_2D_tensor_argument(idx, dst, slice);
                enqueue(queue, *this, slice, lws_hint());
            }
            while(window.slide_window_slice_2D(slice));
            break;
        }
        case DataLayout::NHWC:
        {
            Window collapsed = window.collapse(ICLKernel::window(), Window::DimZ);
            Window slice     = collapsed.first_slice_window_4D();

            unsigned int idx = 0;
            add_4d_tensor_nhwc_argument(idx, src);
            add_4d_tensor_nhwc_argument(idx, dst);
            enqueue(queue, *this, slice, lws_hint());
            break;
        }
        default:
            ARM_COMPUTE_ERROR("Data layout not supported");
    }
}

validate()

Status validate ( const ITensorInfo *  src, const ITensorInfo *  dst, const ScaleKernelInfo &  info  )

static

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

Similar to ClScaleKernel::configure()

Returns

a status

Definition at line 81 of file ClScaleKernel.cpp.

References ARM_COMPUTE_RETURN_ON_ERROR, and arm_compute::cpu::kernels::validate_arguments().

Referenced by ClScale::validate().

{
    ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src, dst, info));
    return Status{};
}

---

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

• src/gpu/cl/kernels/ClScaleKernel.h
• src/gpu/cl/kernels/ClScaleKernel.cpp