首页 | 本学科首页   官方微博 | 高级检索  
     

TMS320C6X的SPLOOP技术
引用本文:方志红,常越.TMS320C6X的SPLOOP技术[J].雷达科学与技术,2014,12(4):437-440.
作者姓名:方志红  常越
作者单位:中国电子科技集团公司第三十八研究所,安徽合肥230088
基金项目:国防科研基金(No.KJ2011196)
摘    要:软件流水是一种实现循环迭代中指令级并行的指令调度技术。它可以克服多周期指令延迟对CPU处理性能的影响,保证循环核的运行效率最优。从C64X+开始,TMS320C6X系列DSP引入SPLOOP技术,软件上增加SPLOOP(D/W)、SPKERNEL等相关指令,硬件上增加软件流水缓存等专用模块,通过模调度软件流水模式,有效缩小了软件代码量,提升了执行代码效率。一般情况下,采用SPLOOP技术后机器编译输出的循环代码质量很高,编程人员无需再对代码进行进一步的手工优化。

关 键 词:软件流水  模调度  循环核  并行运算

The SPLOOP Technology of TMS320C6X
FANG Zhi-hong,CHANG Yue.The SPLOOP Technology of TMS320C6X[J].Radar Science and Technology,2014,12(4):437-440.
Authors:FANG Zhi-hong  CHANG Yue
Affiliation:(No. 38 Research Institute of CETC, Hefei 230088, China)
Abstract:Software pipelining is a type of instruction scheduling that exploits instruction level parallel- ism(ILP) across loop iterations. It prevents multi-delay-cycles instructions from decreasing the performance of CPU by optimizing the efficiency of loop kernel. From C64X+, SPLOOP technology has been implied on TMS320C6X series DSP. Special instructions such as SPLOOP(D/W), SPKERNEL have been used while special hardware such as SPLOOP buffer has been implied. With the help of modulo scheduling, the efficiency of executing code has been improved while the size of executing code has been reduced. Under normal cir eumstances, the compiler/assembly optimizer will do a good job coding SPLOOPs and it will not be necessary for the programmer to optimize code by hand.
Keywords:software pipelining  modulo scheduling  loop kernel  parallel operation
本文献已被 CNKI 维普 等数据库收录!
点击此处可从《雷达科学与技术》浏览原始摘要信息
点击此处可从《雷达科学与技术》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号