Matrix编译器中饱和算术指令映射的设计与实现  

Design and implementation of saturation arithmetic instructions mapping in Matrix complier

作　　者：淡孝强[1] 陈跃跃[1] 孙海燕[1] 阳柳[1] 罗杰[1] 辛乃军[1] 王霁[1] 

机构地区：[1]国防科学技术大学计算机学院,湖南长沙410073

出　　处：《计算机工程与科学》2013年第8期8-14,共7页Computer Engineering & Science

摘　　要：Matrix是一款面向软基站的高性能DSP,饱和算术指令是Matrix体系结构中很重要的一种指令,它使得DSP算法更加安全、准确、高效。因此,Matrix编译器对饱和算术指令的映射支持就尤为关键。Matrix编译器的开发平台是gcc,现有的gcc版本只支持Fixed-point的饱和算术指令,不支持整数和浮点的饱和算术指令。首先分析了gcc内部指令映射的机制,在此基础上提出了一种基于C扩展的饱和算术指令映射实现方案,并通过实现饱和加法指令的映射来验证所提出的方案。实验结果表明,该方案能够高效、准确地实现饱和算术指令的映射。Matrix is a high-performance DSP processor for soft base stations. The saturation arith- metic instruction is of importance in Matrix, which enables the DSP algorithms more secure, accurate and efficient. As a result, supporting saturation arithmetic instructions mapping is critical for Matrix compiler, which is developed based on GCC. However, the current versions of GCC only support satu- ration arithmetic instruction with fixed--point types rather than integer or float types. Based on the in- ternal mechanism of instructions mapping in GCC, an implementation scheme of C extensions based sat- uration arithmetic instructions is proposed. Further more, the proposed scheme is validated by imple- menting the saturation integer addition instruction mapping, and the results show that this scheme can support Matrix＇s saturation arithmetic instruction mapping efficiently and accurately.

关 键 词：GCC 饱和算术 指令映射 C扩展 

分 类 号：TP314[自动化与计算机技术—计算机软件与理论]