Title:Xcel-RAM: Accelerating Binary Neural Networks in High-Throughput SRAM Compute Arrays

Abstract:Deep neural networks are a biologically-inspired class of algorithms that have recently demonstrated state-of-the-art accuracies involving large-scale classification and recognition tasks. Indeed, a major landmark that enables efficient hardware accelerators for deep networks is the recent advances from the machine learning community that have demonstrated aggressively scaled deep binary networks with state-of-the-art accuracies. In this paper, we demonstrate how deep binary networks can be accelerated in modified von-Neumann machines by enabling binary convolutions within the SRAM array. In general, binary convolutions consist of bit-wise XNOR followed by a population-count (popcount). We present a charge sharing XNOR and popcount operation in 10 transistor SRAM cells. We have employed multiple circuit techniques including dual-read-worldines (Dual-RWL) along with a dual-stage ADC that overcomes the inaccuracies of a low precision ADC, to achieve a fairly accurate popcount. In addition, a key highlight of the present work is the fact that we propose sectioning of the SRAM array by adding switches onto the read-bitlines, thereby achieving improved parallelism. This is beneficial for deep networks, where the kernel size grows and requires to be stored in multiple sub-banks. As such, one needs to evaluate the partial popcount from multiple sub-banks and sum them up for achieving the final popcount. For n-sections per sub-array, we can perform n convolutions within one particular sub-bank, thereby improving overall system throughput as well as the energy efficiency. Our results at the array level show that the energy consumption and delay per-operation was 1.914pJ and 45ns, respectively. Moreover, an energy improvement of 2.5x, and a performance improvement of 4x was achieved by using the proposed sectioned-SRAM, compared to a non-sectioned SRAM design.