Architecture-Aware Optimization of Layer Fusion for Latency-Optimal CNN Inference | IEEE Conference Publication | IEEE Xplore