VLIW处理器的变长指令跨边界派发窗设计  

Design of Variable-length Instruction Cross-boundary Dispatch Window for VLIW Processors

作　　者：王东旭 汪东 万江华 WANG Dongxu;WANG Dong;WAN Jianghua(School of Physics and Optoelectronics,Xiangtan University,Xiangtan 411100,China;Hunan Great-Leo Microelectronics Co.,Ltd.,Changsha 410003,China)

机构地区：[1]湘潭大学物理与光电工程学院,湖南湘潭411100 [2]湖南毂梁微电子有限公司,长沙410003

出　　处：《电讯技术》2024年第12期2038-2043,共6页Telecommunication Engineering

基　　金：湖南省制造业关键产品揭榜挂帅项目(2022GXGG012)。

摘　　要：针对传统超长指令字(Very Long Instruction Word,VLIW)处理器代码体积增大会显著降低处理器性能的问题,设计了一种八流出新型变长指令跨边界派发窗。该派发窗兼容压缩指令派发功能,支持压缩指令和整字指令混合派发,有效减小了处理器代码体积。同时该派发窗引入指令跨边界派发机制,进一步排出指令间无用气泡。通过搭建派发窗仿真模型,并基于DSP/VoLIB库进行仿真,结果显示,采用新型变长指令跨边界派发窗能够充分发挥指令级并行优势。经编译器调度优化后,库中典型程序体积比传统派发窗平均降低约19.26%,处理器性能提升约15.4%。In order to solve the problem that the increase of code size of the traditional very long instruction word(VLIW)processor will significantly reduce the performance of the processor,a new eight-flow variable-length instruction cross-boundary dispatch window is designed.The dispatch window is compatible with compression instruction technology,and supports the mixed dispatch of compression instructions and integer instructions,which effectively compresses the size of processor code.At the same time,the dispatch window introduces a cross-boundary dispatch mechanism for instructions to further expel useless bubbles between instructions.By building a dispatch window simulation model and simulating it based on the DSP/VoLIB library,the results show that the new variable-length instruction cross-boundary dispatch window can give full play to the advantages of instruction-level parallelism.After compiler scheduling optimization,the size of typical programs in the library is reduced by about 19.26%on average compared with that of the traditional dispatch window,and the processor performance is improved by about 15.4%.

关 键 词：超长指令字(VLIW) 指令派发 指令压缩 跨边界派发窗 

分 类 号：TP368.1[自动化与计算机技术—计算机系统结构] TN402[自动化与计算机技术—计算机科学与技术]