1 Introduction
Sram have been the predominant and universal technologies used to implement storage in computer systems [1]. Fast on-chip memories are vital to increase the speed of the data flows and, hence, the speed of the entire system. With the high access speed and robustness characteristics, eight transistors (8 T) SRAM cells have been widely applied to on-chip memories (including register files (RFs) and cache memories) in state-of-the-art microprocessors [2], [3], [4]. However, on-chip memories take over 50 percent of the die area according to 2012 ITRS [5], and dissipate a dominant amount of the leakage energy [6]. This issue is expected to aggravate with continuous technology scaling, especially in battery-powered computing devices, such as laptop computers, smart phones, and medical sensors [7].