嵌入式系统虚拟机技术研究

发布时间：2018-03-17 02:32

本文选题：嵌入式系统　切入点：虚拟机　出处：《宁波大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着电子技术的飞速发展，嵌入式系统的应用已经深入到人们生活的方方面面，而种类繁多的微处理器极大地降低了嵌入式软件的通用性，造成了许多嵌入式应用的重复开发。为了使应用程序在嵌入式系统之间具有高度的可移植性，Java虚拟机技术被广泛采用。由于Java自身的一些特征导致其存在效率低、内存占用等缺陷，因此，提高Java虚拟机性能已成为一大研究热点。目前对Java虚拟机的优化工作主要是针对桌面计算机，而很多优秀的虚拟机优化技术在硬件资源受限的嵌入式平台上不适用，导致嵌入式Java虚拟机优化相对滞后。本文以KVM为基础，对嵌入式Java虚拟机的类装载器、垃圾收集和执行引擎进行研究，并提出了一些改进虚拟机性能的方案。主要研究工作包括以下内容：首先，，深入研究了嵌入式Java虚拟机运行环境及其体系结构，主要包括运行时数据区、本地方法接口、类装载子系统、垃圾收集器以及执行引擎，着重研究了类装载、垃圾收集和执行引擎的实现原理。其次，深入分析了嵌入式Java虚拟机性能优化的技术瓶颈，并提出了对其多个模块的改进算法。在类名存储方式上，提出了包名分离的存储方法，有效消除了类名存储产生的冗余信息，降低了常量池的内存占用；在内存管理方面，提出了一种动态分代垃圾收集算法，该算法对新生代和旧生代存储区域进行动态划分，提高了内存的使用效率；在执行引擎方面，提出一种基于自适应编译技术的执行引擎优化方案，该方案将Java指令划分为上下文相关和上下文无关两种，提高了编译速度，并采用简化的热点监视器，有效提高了嵌入式Java虚拟机的性能。最后，将优化后的Java虚拟机移植到开发板，并验证了上述优化算法能提高Java虚拟机的性能。
[Abstract]:With the rapid development of electronic technology, the application of embedded system has penetrated into all aspects of people's lives, and many kinds of microprocessor greatly reduces the versatility of embedded software, resulting in duplication of development of many embedded applications. In order to make the application's portability in embedded system, Java virtual machine technology is widely used.
Due to some characteristics of Java itself due to its low efficiency, memory usage and other defects, therefore, improve the performance of Java virtual machine has become a major research focus. The optimization of the Java virtual machine is mainly for desktop computers, and many excellent virtual machine optimization of embedded platform technology in the limited hardware resource is not applicable in embedded Java virtual machine optimization is relatively backward. Based on KVM, the embedded Java virtual machine class loader, garbage collection and execution engine research, and puts forward some suggestions on the performance of virtual machines in the case. The main work is as follows:
Firstly, the embedded Java virtual machine running environment and its architecture are deeply studied, mainly including runtime data area, local method interface, class loading subsystem, garbage collector and execution engine. The implementation principle of class loading, garbage collection and execution engine is emphatically studied.
Secondly, in-depth analysis of the technical bottleneck of embedded Java virtual machine to optimize the performance of the improved algorithm is proposed for the multiple modules. In class storage mode, put forward the storage method of package name separation, effectively eliminate the redundant information storage class generated, reducing the memory footprint of the constant pool in memory management; and put forward a dynamic generational garbage collection algorithm, the algorithm of dynamic division of the new generation and the old generation storage area, improve memory efficiency; in the execution engine, presents an execution engine optimization scheme based on adaptive compiler technology, the program will be divided into Java instruction context and context free two, improve the compilation speed, and the use of simplified hot spot monitor, can effectively improve the performance of embedded Java virtual machine.
Finally, the optimized Java virtual machine is transplanted to the development board, and it is verified that the optimization algorithm can improve the performance of the Java virtual machine.

【学位授予单位】：宁波大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP368.1;TP302

【参考文献】