Optimization On Memory Management For Just-in Time Compiler

Java is the most widely used programming language, with Garbage Collector (GC) as its most important character. GC deals with objects memory allocation and automatically reclaims the memory space in heap. GC has the advantage such as reliability of memory management and convenience for programmers. However, the collection pause and the consequent overhead in CPU have become the bottleneck of the Java Virtual Machine's performance.Our research group has been working on memory management and obtained certain achievements. We propose a new object lifetime analysis algorithm for just-in time GC to combine the live variable analysis and the points-to escape analysis, and reclaim the dead object in heap by insert some explicit free instruction. This work has already been implemented in Java virtual machine. However, the deficiency of the former work is that the analysis of Java library is not satisfactory and heap utilization of the implementation platform can be further improved. This paper is committed to improving the existing system, focusing on an accurate analysis mode for Java library and covering the following tasks:1. to present a high level abstraction for points-to graph in order to improve our analysis precision. As points-to graph is a model representing the relation between variables and objects allocated in heap, we can get more optimization potential if we make abstract summary to points-to graph.2. to optimize the. We identify the allocate point in loops and try to move it outside of the loop.3. to implement the memory management in compile time on a compact garbage collector, and improve the reuse method for reclaim space of individual objects.