Exploiting processor features to implement error detection in reduced precision matrix multiplications | Publicación