Binarized Neural Network Accelerator Macro Using Ultralow-Voltage Retention SRAM for Energy Minimum-Point Operation

Abstract

A binarized neural network (BNN) accelerator based on a processing/computing-in-memory architecture using ultralow-voltage retention SRAM (ULVR-SRAM) is proposed for the energy minimum-point (EMP) operation. The BNN accelerator (BNA) macro is designed to perform stable inference operations at EMP and substantive power-gating using ULVR at an ultralow voltage (< EMP), which can be applied to fullyconnected layers with arbitrary shapes and sizes. The EMP operation of the BNA macro, which is enabled by applying the ULVR-SRAM to the macro, can dramatically improve the energy efficiency (TOPS/W) and significantly enlarge the number of parallelized multiply-accumulate (MAC) operations. In addition, the ULVR mode of the BNA macro, which also benefits from the usage of ULVR-SRAM, is effective at reducing the standby power. The proposed BNA accelerator macro can show a high energy efficiency of 65 TOPS/W for fully connected layers. This BNA macro concept using the ULVR-SRAM can be expanded to convolution layers, where the EMP operation is also expected to enhance the energy efficiency of convolution layers.