[EN] In this work, we assess the performance and energy efciency of high-performance codes for the convolution operator, based on the direct, explicit/implicit lowering and Winograd algorithms used for deep learning (DL) ...