1 paper
Xuan Yang, Mingyu Gao, Qiaoyi Liu +9
We show that DNN accelerator micro-architectures and their program mappings represent specific choices of loop order and hardware parallelism for computing the seven nested loops o…