DSP实时图像处理软件优化方法研究  

Research on Optimization Method of DSP Real-time Image Processing Software

作　　者：雷涛[1,2] 周进[1] 吴钦章[1] 

机构地区：[1]中国科学院光电技术研究所,成都610209 [2]中国科学院研究生院,北京100039

出　　处：《计算机工程》2012年第14期177-180,共4页Computer Engineering

摘　　要：为提高高速图像处理系统中数字信号处理器(DSP)软件的实时性,分别提出面向算法与代码2个层次的优化方法。算法级优化通过重新设计算法的实现流程,充分利用处理器资源,完成算法到处理器上的高效映射;代码级优化使用汇编语言,对固定算法的代码进行优化,使循环核形成高效的软件流水,达到实时性能要求。实验结果表明,2种优化方法均能提高DSP软件中关键模块的处理速度。In the high-speed image processing system, in order to solve real-time performance of the software based on the Digital Signal Processor（DSP）, two optimization methods are presented. One method is based on the algorithm, the other is based on the code. The method of optimization based on the algorithm means that the algorithm can be mapped to the processor efficiently through designing the flow of the algorithm. The code optimization means optimizing the given code using assemble language. Then the loop kernel can form high efficient software pipelining and satisfy strict real-time constraints on performance. Experimental results show that the optimization based on the algorithm and the optimization based on the code can both increase the processing speed of key modules in DSP software.

关 键 词：实时图像处理 算法级优化 代码级优化 数字信号处理器 软件流水 循环核 

分 类 号：TP911.73[自动化与计算机技术]