一种基于RRAM热串扰的奇偶重排编码算法
An Efficient Parity Rearrangement Coding Scheme for RRAM Thermal Crosstalk Effects
-
摘要: 阻变存储器(resistive random access memory, RRAM)作为未来一种高性能的非挥发性存储器, 具有面积小、操作电压低、兼容性好等特点.但是, 在高集成存储器和频繁的写操作下, 热串扰问题会严重影响RRAM的保持特性.严重情况下, 热串扰问题甚至会造成一系列的错误翻转.因此, 本文引入了一种高效的奇偶重排编码算法(parity rearrangement coding scheme, PRCoder)来有效缓解热串扰对RRAM的影响, 并在算法层和电路层上分别进行设计与仿真.试验结果表明, PRCoder算法平均降低了32.7%的误翻转率, 并同时只会在每一个存储行带来1 bit的额外开销.此外, PRCoder仅仅带来0.3%的性能增加和0.008%的面积增加.Abstract: Resistive random access memory (RRAM) is a high-performance nonvolatile memory with the feature of small area, low operation voltage, and good compatibility. However, when frequent writing operation at high density integration, the crosstalk problem will seriously influence the retention properties of RRAM, which is common to create some error flips in serious conditions. Therefore, we present an effective parity rearrangement coding scheme, that is PRCoder, to reduce the influence in RRAM. We present the PRCoder at algorithm level and emulate it at hardware level for various scenarios. Experiments indicate that PRCoder can decrease the accidentally flips by about 32.7% on average, which is caused by thermal crosstalk, with just one additional bit of redundancy per storage line. Moreover, PRCoder causes an ignorable performance spending of ~0.3% with inferior to 0.008% extra RRAM area.
下载:
