DLFloat: A 16-b Floating Point Format Designed for Deep Learning Training and InferenceAnkur AgrawalBruce Fleischeret al.2019ARITH 2019