7.3 A 28nm 38-to-102-TOPS/W 8b Multiply-Less Approximate Digital SRAM Compute-In-Memory Macro for Neural-Network Inference | IEEE Conference Publication | IEEE Xplore