虚拟环境下可重构流媒体处理的应用分析方法

Virtual Environment Application Analysis Method for Reconfigurable Streaming Media Processing

流媒体类应用现已在嵌入式移动设备上得到广泛应用,可重构计算方式为此类应用提供了灵活和高效的解决方案。但目前对于应用程序,尤其是核心循环的分析仍缺乏有效方法和工具支持。面向嵌入式可重构流媒体处理以应用中核心循环作为主体进行程序分析。结合应用存储需求等特征信息,提出一种基于PIA-CDTG图模型的启发式任务映射方法,以实现流媒体类数据密集型应用的有效映射。随后实现了可视化应用分析工具EVA2并支持虚拟环境下多领域专家的分布式协作优化,将上述方法集成其中并将分析结果以图形界面形式显示出来。方法及工具对于流媒体应用算法在嵌入式设备上的执行优化、性能评价等方面具有创新性和实用价值。

Streaming media applications are now wildly used in embedded portable devices, in which reconfigurable computing provides flexible and efficient solutions. However, there is lack of effective methodology and tool for application analysis, especially for kernel loops. The program feature analysis on the kernel loops of media applications was first taken. With these features such as storage requirement, a task mapping method based on PIA-CDTG model was proposed for efficient application mapping. A visualized analysis tool called EVA2 was implemented under a distributed collaborative optimization environment for multi-expert decision, which took the analysis results shown with graphical interfaces. Considering optimized execution and performance evaluation of embedded stream media process, the research method and tool are creative and practical.