【摘要】 基于trace的即时编译技术是一种提高解释型语言性能的有效方法。然而,现有的trace探测技术都是针对程序单次执行的,无法利用服务器端程序并发执行的特点。针对并发执行的服务器端程序,提出一种基于序列模式挖掘的trace探测方法,以快速发现热点trace。将并发执行的服务器端程序看作是多个基本块序列,应用序列模式挖掘算法,对得到的序列模式进行去重与合并以发现热点trace。实验结果表明基于序列模式挖掘的trace探测能够有效地提高trace探测的效率。更多还原

【Abstract】 Trace-based JIT compilation technique is an effective method to improve the performance of interpreted language. However,the existing trace detection technology is executed on a single execution of the program,and it’s not able to use the features of the concurrent execution of server-side programs. Aiming at the concurrent execution of serverside programs,this paper proposed a trace detection method based on sequential pattern mining to quickly discover hot trace. The concurrent execution of server-side program as basic block sequences applied sequential pattern mining algorithms which combined the sequence patterns to find hot trace. The experimental results show that trace detection based on sequential pattern mining can effectively improve the efficiency of trace detection.更多还原