Method configure

src/gpu/cl/kernels/ClCol2ImKernel.cpp:114–157

}

void ClCol2ImKernel::configure(const CLCompileContext &compile_context,
                               ITensorInfo            *src,
                               ITensorInfo            *dst,
                               const Size2D           &convolved_dims,
                               unsigned int            num_groups)
{
    ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst);

    // Perform validation step
    ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src, dst, convolved_dims, num_groups));

    _convolved_dims = convolved_dims;

    const DataType data_type = src->data_type();

    // Create kernel
    CLBuildOptions build_opts;
    build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(data_type));
    build_opts.add_option("-DELEMENT_SIZE=" + support::cpp11::to_string(src->element_size()));
    build_opts.add_option("-DWIDTH_INPUT=" + support::cpp11::to_string(src->dimension(0)));
    build_opts.add_option("-DWIDTH_OUTPUT=" + support::cpp11::to_string(_convolved_dims.width));
    build_opts.add_option("-DNUM_GROUPS=" + support::cpp11::to_string(num_groups));

    _kernel = create_kernel(compile_context, "col2im", build_opts.options());

    // Configure kernel window
    auto win_config = validate_and_configure_window(src, dst, _convolved_dims, num_groups);
    ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
    IClKernel::configure_internal(win_config.second);

    // Set config_id for enabling LWS tuning
    _config_id = "col2im_";
    _config_id += lower_string(string_from_data_type(src->data_type()));
    _config_id += "_";
    _config_id += support::cpp11::to_string(num_groups);
    _config_id += "_";
    _config_id += support::cpp11::to_string(src->dimension(0));
    _config_id += "_";
    _config_id += support::cpp11::to_string(src->dimension(1));
    _config_id += "_";
    _config_id += support::cpp11::to_string(dst->dimension(0));
    _config_id += "_";
    _config_id += support::cpp11::to_string(dst->dimension(1));
}

Status ClCol2ImKernel::validate(const ITensorInfo *src,
                                const ITensorInfo *dst,