Anatomy Of High-Performance Deep Learning Convolutions On SIMD Architectures