CPU architecture for a fast and energy-saving calculation of convolution neural networks Conference Paper