一种面向VLIW指令压缩的寄存器分配算法  

An Algorithm of Register Allocation for Instruction Compression in VLIW Architecture

作　　者：朱少波[1] 姚庆栋[1] 洪享[1] 史册[1] 

机构地区：[1]浙江大学信息与通信工程研究所,杭州310027

出　　处：《计算机工程》2003年第20期154-156,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(69972043);浙江省自然科学基金资助项目(602031)

摘　　要：针对VLIW结构的指令压缩方法，通过对编译中间代码的深入分析和总结，提出一种改进的寄存器分配算法，该算法在线性扫描的基础上，对寄存器的选择添加约束条件，应用该算法能够使得目标代码中寄存器的编号尽量靠近，从而达到更好的压缩效果。This paper designs an improved algorithm of register allocation to reduce the instruction codes space and occupy less instruction bandwidth By analysing the internal language codes, some characters are concluded. The algorithm is based on the linear scan register allocation and is appended restrained conditions on register selection. Using this algorithm makes the register number adjacent to one instruction. Therefore, a better compression result is achieved.

关 键 词：超长指令字 寄存器分配 线性扫描 指令压缩 VLIW 算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]