摘要
目前BWDSP104X编译器的编程模型支持以字为单位的寻址方式,不支持非32位宽数据并且不兼容以字节为单位的寻址模式。通过对BWDSP104X前端相关数据类型进行修改,产生基于字节寻址模式的中间代码,而后端由此输出一系列32位机器指令来完成对应的字节地址转换或数据操作,从而实现字节寻址模式的扩展。利用寄存器对模拟实现64位数据的存取与运算,提高浮点数据运算的精度。实验结果表明,该编译器兼容字节寻址模式以及64位浮点数据类型,能较好地满足高速实时信号处理的需求。
Currently, the programming model under BWDSP104X compiler is based on the word unit addressing modes,lacking of support for non 32-bit wide data and incompatible with byte unit addressing modes. With the modification of related front-end data type,the intermediate code based on byte addressing mode is generated. A series of 32-bit machine instructions is outputted by back-end to fulfill byte address transformation or data manipulation, and the expansion of byte addressing mode is implemented. Register pairs are used to simulate 64-bit data access and operation, improving the accuracy of floating-point data operation. Experimental results show that the compiler is compatible with the byte addressing mode as well as the 64-bit floating-point data type. It can well meet the requirement of high-speed real-time signal processing.
出处
《计算机工程》
CAS
CSCD
北大核心
2016年第8期14-18,23,共6页
Computer Engineering
基金
"核高基"重大专项(2012ZX01034-001-001)
关键词
编译器
字寻址
字节寻址
字节对齐
64位数据运算
寄存器对
compiler
word addressing
byte addressing
byte alignment
64-bit data operation
register pair
