BWDSP10x上地址和数据谓词执行的编译优化.pdfVIP

计算机系统应用 h即:// 201 6 年第25 卷第 12 期 BWDSPI0x 上地址和数据谓词执行的编译优化① 樊永朝 11，郑启龙 12，耿锐气王向前3，王吴3 I(中国科学技术大学安徽省高性能计算重点实验室，合肥 230027) 气中国科学技术大学计算机科学与技术学院，合肥 23∞2η J(中国电子科技集团公司第三十八研究所，合肥 23∞88) 摘 要: 传统的谓词优化技术是在冯·诺伊曼体系结构计算机上实施的，仅对数据流进行优化，并没有考虑哈佛 体系结构下指令和数据分开的情况. BWDSPIOx是指令和数据分开的哈佛体系结构，它支持超长指令字，不仅提 供了对数据调词执行的支持也提供了对地址谓词执行的支持.特此提出了一种在区域上对两种谓词模式优化支 持的方法，在进行两种比较之前，通过判断比较操作的两个操作数类型来分别实施两种模式的谓词优化，使得对 地址的比较不用传输到通用寄存器中.实验结果表明该优化方法能显著地节省 CPU 的时间和带宽，大大减少了 分支指令，使程序性能提高了 28.4%. 关键词:地址谓词执行;数据谓词执行;区域;编译优化 Compilation Optimization ofAddress and Data Predicated Execution on BWDSPI Ox l l 3 3 FAN Yong-Chao .2, ZHENG Qi-Long .2, GENG RuP, WANG Xiang-Qian , WANG Ha0 t(A曲ui High Perforrn皿ce Computing Key Laboratory, University ofScience and Techno1ogy ofChina, Hefei 230027, China) 2(Sch∞1 ofComputer Science and Techno1ogy, University ofScience 阻d Technology ofChina, He创23∞27， China) J (No. 38th Research Institute, China Electronics Technology Group Corporation, Hefei 23∞邸， China) Abstract:白le 往aditional predicate optim泣atioo technique is based 00 the Von Newnann architecture, which considers the data flow optimization only. However, BWDSPIOx is based on the Harvard archit饵ture， which data and instructions are physically separated. It provides VLIW and supports not on1y 也ta predicatωexecution but also address predicated execution. Hence,we present an optimization method for the two predicated execution technology based on the region. In 也is method, types of the two operands of comparison operation will be identified before the two kinds of operations are executed, and when addresses are compared, the two operands don't need to 位朋sfer to general registers. Experimental result shows that the optimization method can highly reduce the time and bandwidth ofCPU, and reduce large numbers of branch instructi