Method configure

src/gpu/cl/kernels/ClFillKernel.cpp:52–92

}

void ClFillKernel::configure(const CLCompileContext &compile_context,
                             ITensorInfo            *tensor,
                             const PixelValue       &constant_value,
                             Window                 *window)
{
    ARM_COMPUTE_ERROR_ON_NULLPTR(tensor);
    ARM_COMPUTE_ERROR_THROW_ON(validate(tensor, constant_value, window));

    const DataType data_type  = tensor->data_type();
    const int      vec_size_x = 16 / tensor->element_size();

    // Create and update the window (if needed)
    _full_window = calculate_max_window(*tensor);
    Window win   = _full_window;
    if (window != nullptr)
    {
        ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(win, *window);
        win = *window;
    }

    const int  output_width_x = win.num_iterations(0);
    const bool multi_access_x = output_width_x >= vec_size_x;
    const bool remainder_x    = output_width_x % vec_size_x > 0;

    if (multi_access_x)
    {
        win.set(Window::DimX,
                Window::Dimension(win.x().start(), ceil_to_multiple(win.x().end(), vec_size_x), vec_size_x));
    }
    ICLKernel::configure_internal(win);

    // Create kernel
    CLBuildOptions build_opts;
    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
    build_opts.add_option("-DCONSTANT_VALUE=" + string_from_pixel_value(constant_value, data_type));
    build_opts.add_option_if(multi_access_x, "-DVEC_SIZE=" + support::cpp11::to_string(vec_size_x));
    build_opts.add_option_if(multi_access_x && remainder_x,
                             "-DLAST_ACCESSED_X=" +
                                 support::cpp11::to_string(std::max<int>(output_width_x - vec_size_x, 0)));
    _kernel = create_kernel(compile_context, "memset", build_opts.options());
}

Status ClFillKernel::validate(const ITensorInfo *tensor, const PixelValue &constant_value, Window *window)
{