摘要： 针对TI TMS320C6000系列数字信号处理器容易忽视的CPU指令并行、软件流水特点和编译器内联函数、线性汇编及汇编语言的高速运行特点，给出常见的指令存储器相关性问题、循环冗余问题、嵌套循环的流水性能问题和程序存储块冲突问题的分析与优化解决方法。以定点点积算法进行参照实验，结果证明代码运行速度在进行相应C程序编程优化、线性汇编编程优化和手工汇编编程优化后最高分别可以提高85.9%、86.4%和93.1%。

关键词: 数字信号处理器, 编程优化, 指令并行, 软件流水, 线性汇编, 数据存储

Abstract: Using traditional programming methods on the TMS320C6000 Digital Signal Processor(DSP) may omit the advantages of parallel instructions, software pipelining and the high-speed performance of intrinsics, linear assembly language and assembly language, so the analysis and solving methods of common problems of instruction storage correlation, circulation abundance, the performance of nested loop and interleaved memory conflicting are provided. The reference experience with fixed-point dot product programs shows that the code running speed can be separately increased by 85.9%, 86.4% and 93.1% after optimized by optimization methods of C programming, linear assembly programming and hand assembly programming.

Key words: Digital Signal Processor(DSP), programming optimization, instruction parallel, software pipelining, linear assembly, data storage

中图分类号: 

• TP391