基于动态预解压方法的嵌入式内存压缩技术及实现

发布时间：2018-03-13 13:52

本文选题：内存压缩　切入点：嵌入式系统　出处：《湖南大学》2012年硕士论文　论文类型：学位论文

【摘要】：在嵌入式系统中，，内存资源极为宝贵。增大嵌入式设备的内存容量即意味着增加其成本、封装体积和功耗。此外，当今软件对于内存容量的需求正以每年50%-100%的速度增长，同时越来越多的新应用程序需要在内存极为有限的嵌入式设备上运行。为了解决以上问题，内存压缩技术被提了出来。当系统内存不足的时候，内存压缩机制选择不活跃的内存页面，将其从内存中移出并存储在压缩内存区中。内存压缩技术可以提高系统的可用内存，但是会产生一定的时间和空间开销。一方面，压缩和解压缩页面都会带来延迟，频繁的压缩和解压缩操作甚至会引起内存系统的抖动(trashing)而导致系统崩溃。另一方面，为了能够有效的管理压缩页面，内存压缩技术需要消耗一定的内存。针对以上问题，本文提出了一种基于预解压缩方法的嵌入式内存压缩技术。本文所做的工作主要有以下几个方面：（1）针对嵌入式内存压缩技术，建立了嵌入式内存压缩的性能评价模型。之后使用此模型对嵌入式内存压缩技术的增益和损耗进行了定性分析。（2）提出了一种新的具有动态预解压机制的嵌入式内存压缩技术—CMPD（compressed memory with pre-decompression）。为了解决之前嵌入式内存压缩技术难以避免的解压缩延迟问题，本文为CMPD设计了一种基于时间局部性的动态预解压算法，并且从压缩算法、压缩内存区容量动态调整机制、内存管理算法等方面详细阐述了CMPD的设计细节。（3）将CMPD实现成了Linux系统一个可加载的内核模块。设计了不同的实验来测量CMPD的可用内存增益、缺页中断次数、运行时间等指标。实验证明CMPD能够有效的增大系统的可用内存，并且当测试程序的内存访问模式呈现较强的时间局部性的时候，CMPD能够在之前内存压缩技术的基础上大幅度减少解压缩延迟。
[Abstract]:In embedded systems, memory resources are extremely valuable. Increasing the memory capacity of embedded devices means increasing their cost, packaging volume and power consumption. In addition, today's software demand for memory capacity is growing at a rate of 50 to 100 percent a year. At the same time, more and more new applications need to run on embedded devices with extremely limited memory. In order to solve the above problem, memory compression technology has been proposed. The memory compressor system selects inactive memory pages, removes them from memory and stores them in compressed memory areas. Memory compression technology can increase the available memory of the system, but it will result in a certain amount of time and space overhead. Compression and decompression of pages can cause delays, frequent compression and decompression operations can even cause the memory system to jitter and crash. On the other hand, in order to manage compressed pages effectively, Memory compression technology needs to consume a certain amount of memory. In view of the above problems, this paper proposes an embedded memory compression technology based on pre-decompression method. The main work of this paper is as follows:. 1) aiming at the embedded memory compression technology, the performance evaluation model of embedded memory compression is established, and then the gain and loss of embedded memory compression technology are analyzed qualitatively. In order to solve the problem of decompression delay, a new embedded memory compression technology, which has dynamic pre-decompression mechanism, is proposed, which is memory compressed memory with pre-decompression. In this paper, a dynamic predecompression algorithm based on time locality is designed for CMPD, and the design details of CMPD are described in detail from the aspects of compression algorithm, dynamic adjustment mechanism of compressed memory area capacity, memory management algorithm and so on. 3) the CMPD is realized as a loadable kernel module in Linux system. Different experiments are designed to measure the available memory gain of CMPD, the number of page missing interrupts, the running time and so on. The experiment proves that CMPD can effectively increase the available memory of the system. Moreover, when the memory access mode of the test program presents a strong temporal locality, the CMP D can greatly reduce the decompression delay on the basis of the previous memory compression techniques.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP368.1

【参考文献】