基于硬件计数器的MPEG4视频应用性能特征分析

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,IntelMMX、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IntelP4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IntelP4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。

Hardware Counter-based Performance Characterizing for MPEG4 Video Applications

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.