基于PCM和DRAM的混合主存仿真系统研究与实现

发布时间：2018-03-02 00:09

本文关键词： 相变存储器混合主存仿真系统时序模型功耗模型　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：相变存储器(Phase Change Memory,PCM)是一种具有吸引力和发展潜力的新型存储技术,具有非易失性、高读取速度、高存储密度以及低静态功耗等优点。这些优良特性使的 PCM有望成为新一代主存系统的继任者。然而,与当前的动态随机访问存储器(Dynamic Random Access Memory,DRAM)相比,PCM具有写速度慢以及耐写性差等劣势。这意味着目前在计算机系统中,使用PCM完全替代DRAM作为主存还不切实际。而这也引发了许多研究热点:如何将PCM用于主存系统之中;如何从PCM和DRAM中各取优势以实现快响应、大容量、高可靠的主存系统;如何实现基于PCM和DRAM的混合主存数据管理等等。然而,目前缺乏可用的PCM设备,更无从构建基于PCM和DRAM的真实主存系统,这使得相关研究得不到有效实现和验证。更进一步来说,即使未来出现可用PCM设备,由于其结构固定,参数无法灵活配置等问题,亦会导致一些主存内部算法无法应用其中得以验证。因此,为了解决PCM研究火热与相关实验环境缺乏的矛盾,本文的核心研究工作便是:设计并构建一个配置灵活、使用便捷、结果精确的基于PCM和DRAM的混合主存仿真系统。本文首先概述了选题背景以及研究意义,介绍了现有主存仿真系统的特点,总结分析了其不能完全满足当前主存研究的实际需求,在此背景下提出并构建了一个基于PCM和DRAM的混合主存仿真系统。然后,本文着重阐述了该仿真系统的实现架构、功能模块、时序模型、功耗模型、混合主存分配接口以及可视化用户界面等重要组成部分。最后,通过验证性实验以及应用场景示例,说明了所构建仿真系统的有效性及可用性。本文主要贡献可归纳如下:(1)设计并实现了一个支持基于多种主存架构的仿真系统,它构建了主存系统的各个功能模块,并实现了可视化用户界面交互,具有配置灵活丰富、运行简单便捷、仿真结果详尽等特点。(2)分别构建了与DRAM和PCM运行机制相符的时序模型及功耗模型,并通过验证实验证明了其仿真结果的准确性。(3)基于该混合主存仿真系统,提出了一个应用可控的混合主存分配接口,弥补了当前主存仿真系统一大缺陷:即不具备对异质存储区分的功能。(4)探讨了所构建混合主存仿真系统的多种应用场景,指出其为主存相关研究工作带来的实际便利。
[Abstract]:Phase Change memory (PCM) is a new type of storage technology with attractive and potential development. The advantages of high memory density and low static power consumption make PCM a promising successor for a new generation of main-memory systems. Compared with the current dynamic Random Access memory DRAM, PCMs have the disadvantages of slow writing speed and poor writeability, which means that they are currently in computer systems. It is impractical to use PCM to replace DRAM as main memory completely, and this has caused many research hotspots: how to use PCM in main memory system, how to take advantage of PCM and DRAM to realize fast response, large capacity and high reliability main memory system; How to realize the mixed main memory data management based on PCM and DRAM, etc. However, there is a lack of available PCM devices at present, and there is no way to build real main memory system based on PCM and DRAM. Further, even if the PCM devices are available in the future, because of their fixed structure, the parameters can not be flexibly configured, and so on. Therefore, in order to solve the contradiction between the hot PCM research and the lack of related experimental environment, the core research work of this paper is: design and build a flexible configuration, easy to use, Results the hybrid main memory simulation system based on PCM and DRAM is accurate. Firstly, this paper summarizes the background and significance of the research, introduces the characteristics of the existing main memory simulation system, and analyzes that it can not fully meet the actual needs of the current main memory research. In this background, a hybrid main memory simulation system based on PCM and DRAM is proposed and constructed. Hybrid memory allocation interface and visual user interface, etc. Finally, through validation experiments and application scenario examples, The main contributions of this paper can be summarized as follows: 1) Design and implementation of a simulation system based on multiple main memory architectures, which constructs various functional modules of the main memory system. The visual user interface interaction is realized, which has the characteristics of flexible configuration, simple and convenient operation, detailed simulation results, etc.) the timing model and the power consumption model which are consistent with the running mechanism of DRAM and PCM are constructed, respectively. The accuracy of the simulation results is proved by the verification experiment. Based on the hybrid main memory simulation system, a controllable hybrid main memory allocation interface is proposed. This paper makes up for a big defect of the current main memory simulation system, that is, it does not have the function of differentiating heterogeneous memory. It discusses the various application scenarios of the hybrid main memory simulation system, and points out the practical convenience it brings for the research work on the main memory.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP333

【相似文献】