FAST: A Fully-Concurrent Access SRAM Topology for High Row-Wise Parallelism Applications Based on Dynamic Shift Operations

Yiming Chen, Yushen Fu, Mingyen Lee, Sumitha George, Yongpan Liu, Vijaykrishnan Narayanan, Huazhong Yang, Xueqing Li

Abstract: This brief proposes a fully-concurrent access SRAM topology to handle high-concurrency operations on multiple rows in an SRAM array. Such high-concurrency operations are widely seen in both conventional and emerging applications where high parallelism is preferred, e.g., the table update in a database and the parallel feature update in graph computing. The proposed shift-based parallel access and compute architecture is enabled by integrating the shifter function into each SRAM cell, and by creating a datapath that exploits the high-parallelism of shift operations in multiple rows. An example of a 128-row 16-column shiftable SRAM in 65nm CMOS is designed. Post-layout SPICE simulations show improvements of 5.5x energy efficiency and 27.2x speed in average over a conventional digital near-memory computing scheme. In addition, the design has been fabricated and the measurement results show support of up to 800MHz clock at 1.0V and 1.2GHz at 1.2V.