国防科技大学学报

引用本文:	车永刚,王正华,李晓梅.基于硬件计数器的MPEG4视频应用性能特征分析.[J].国防科技大学学报,2004,26(2):46-50 ,55.[点击复制]
	CHE Yonggang,WANG Zhenghua,LI Xiaomei.Hardware Counter-based Performance Characterizing for MPEG4 Video Applications[J].Journal of National University of Defense Technology,2004,26(2):46-50 ,55[点击复制]

【打印本页】【在线阅读全文】【下载PDF全文】【查看/发表评论】【下载PDF阅读器】【关闭】

←前一篇|后一篇→

过刊浏览高级检索

本文已被：浏览 6297次下载 6648次

基于硬件计数器的MPEG4视频应用性能特征分析

车永刚¹, 王正华¹, 李晓梅²

(1.国防科技大学并行与分布处理国家重点实验室，湖南长沙 410073;2.装备技术指挥学院，北京 101416)

摘要:

通过指令集扩展来增强多媒体应用性能是微处理器发展的一个趋势，Intel MMX、SSE和SSE2是很好的例子。但应用本身的特点是否支持这些体系结构特性，应用软件是否充分发挥了多媒体指令集的潜力，都是值得关注的问题。采用动态Profiling的方法，研究MPEG4视频应用在基于Intel P4处理器的桌面机上的性能特性，使用硬件计数器工具来获取指令级的性能数据，并进行了深入分析。发现MPEG4播放软件较好地利用了Intel P4的64位MMX指令集，但对SSE和SSE2指令集几乎没有利用上；两级Cache不命中导致的性能损失很大。为提高视频播放性能，改进存储层次设计与优化播放软件比改进多媒体指令集更加有效。

关键词: MPEG4视频硬件性能监视性能特征 MMX SSE SSE2

DOI：

投稿日期：2003-09-01

基金项目:国家自然科学基金重点资助项目(69933030)

Hardware Counter-based Performance Characterizing for MPEG4 Video Applications

CHE Yonggang¹, WANG Zhenghua¹, LI Xiaomei²

(1.National Lab. for Parallel and Distributed Processing, National Univ.of Defense Technology, Changsha 410073，China;2.Institute of Equipment and Command Technology, Beijing 101416，China)

Abstract:

It is the current trend that microprocessors extend their instruction set to speedup the multimedia applications. Intel'sMMX, SSE and SSE2 are good example. But do the multimediaapplications favorthesearchitecturefeatures? And does thecurrentsoftwaresupporttheseinstruction sets and make full use of them? These questions are of muchconcerntothe researchers. Weinvestigate theperformancecharacteristics of 2MPEG4Videoplayers on an Intel P4 baseddesktop PC bymeans ofdynamicprofiling.BasedonIntelVtune, a hardwareperformancemonitoring tool, we obtainaccurateperformancedata. And we make indepthanalysis of the obtained data based on themachine'sarchitectureparameters andperformanceprofiles. We find that the 2applicationsmakegood use of 64-bit MMXinstructions. But theyrarely use SSE and SSE2instructions. Theperformancedegradation due to Cachemisses at tow levels of Caches is remarkable. Toenhance theperformance ofMPEG4 mediaapplications,improvingmemoryhierarchyperformance andoptimizing themediaplayers are moreeffective ways thanemploying moreadvancedmultimediainstruction sets.

Keywords: MPEG4 videos hardware performancemonitoring performancecharacteristics MMX SSE SSE2