实时嵌入式多媒体系统模糊阈值合并内存管理算法  

Fuzzy threshold coalescence memory algorithm for embedded real-time multimedia systems

作　　者：姜艳[1,2,3] 曾学文[2,3] 孙鹏[2,3] 朱小勇[2,3] 

机构地区：[1]中国科学院研究生院,北京100049 [2]中国科学院声学研究所,北京100190 [3]中国科学院国家网络新媒体工程技术研究中心,北京100190

出　　处：《西安电子科技大学学报》2012年第5期174-180,共7页Journal of Xidian University

基　　金：国家科技支撑项目课题资助项目(2011BAH08B01);863重大项目课题资助项目(2011AA01A102);中国科学院战略性先导科技专项课题资助项目(XDA06030500)

摘　　要：为了提高实时嵌入式多媒体系统动态内存分配的效率,提出一种基于两级离散列表(Two-Level Segregate Fit,TLSF)的模糊阈值合并算法.根据内存区间的使用概率和空闲块数统计,运用模糊规则计算合并系数,并与由碎片率决定的动态阈值相比较,决定是否推迟合并.实验结果表明,该算法在内存大小随机分配时与TLSF算法的性能相当;在实际的视频点播系统应用中,分配和释放效率分别提高了12.6%和9.5%。In order to raise the memory management efficiency in real-time embedded multimedia systems, a fuzzy threshold coalescence memory algorithm based on TLSF （Two-Level Segregate Fit） is proposed. This algorithm employs fuzzy rules combined with the probability of free memory regions and statistics of free memory blocks in every region to obtain the coalescence coefficient, which is compared with dynamic threshold adjusted by memory fragmentation to decide whether the coalescence needs to be postponed. Experimental results show that the performance of this algorithm is similar to that of TLSF when the allocated memory size is random, but allocation and release efficiencies are increased by 12.6 % and 9.5%, respectively, in a real video on the demand system, by adopting this algorithm in an embedded system.

关 键 词：嵌入式系统 内存管理 模糊控制 动态阈值 推迟合并 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]