基于Linux平台的DSP编程性能优化  

Performance optimization of DSP programming based on Linux platfrom

作　　者：钟卫[1] 赵晓群[1] ZHONG Wei ZHAO Xiao-qun(College of Electronics and Information Engineering, Tongji University ,Shanghai 201840, China)

机构地区：[1]同济大学电子与信息工程学院,上海201840

出　　处：《燕山大学学报》2016年第5期431-437,470,共8页Journal of Yanshan University

基　　金：国家自然科学基金资助项目(61271248)

摘　　要：自1980年以来,DSP芯片的不断发展使其在多个领域得以成熟应用。而在实际工业应用中,一个好的算法往往需要经过大量优化才能满足系统日益提高的性能要求。本文先使用Linux平台下的性能分析工具gprof对代码整体结构进行剖析,再基于分析结果进行有针对性有侧重点的优化。其中所选优化方法适用于大部分DSP编程环境,包括编译器选项优化、Intrinsic函数优化、循环体优化和存储空间分配优化。以混合激励线性预测(MELP)语音编解码算法为例进行说明,由实验结果可以看出,这种优化方法比普遍优化效率更高,效果更好。At present, the sustainable development of DSP chip has made it been applied in many fields maturely since 1980. In practical industrial applications, a good algorithm always tends to go through lots of system optimization to meet the ever-increasing performance requirements. Performance analysis tool gprof of Linux platform is used to analyze the overall structure of the program firstly, and then based on the analysis result more focused and targeted optimization is implemented. The optimization methods include complier option optimization, instrinsie functions optimization, loop optimization and storage space allocation optimization. In this paper, MELP（ Mixed-Excitation Linear Prediction） speech codec is taken as an example to illustrate. As it can be seen from the experimental results, this kind of optimization method has higher efficiency and better effects compared with general optimization.

关 键 词：LINUX系统 性能分析工具 DSP编程 混合激励线性预测 

分 类 号：TP368.1[自动化与计算机技术—计算机系统结构]