期刊文献+

超长指令字DSP处理器的共享寄存器堆设计

The Design of Shared Register File for VLIW DSP Processors
下载PDF
导出
摘要 共享数据寄存器堆设计是超长指令字DSP处理器实现的难点。它的访问延时成为处理器的关键延时之一。在一高性能超长指令字DSP处理器的设计中,通过对传统单周期读写寄存器堆的设计方案进行深入的分析和研究,优化关键路径,设计出双周期读写结构的寄存器堆。通过电路实现比较后证实,双周期方案在减少27%访问时间的同时减少23%的面积。 The design of shared data register file is the hardest for implementation of VLIW ( Very Long Instruction Word) DSP (Digital Signal Processing) Processors. Its access time could be one of the critical delays. In the design of a high performance VLIW DSP Processor, a new two-stage-access design is promoted, based on the detail analysis and study of the traditional one-stage-access design, in order to optimize the critical delay. After comparing of the performance of the implementations of two types of design:, it is clear that the using two-stage-access structure reduces the access time by 27%, while saving the area by 23%.
出处 《科学技术与工程》 2006年第13期1921-1925,1928,共6页 Science Technology and Engineering
关键词 超长指令字 数字信号处理器 寄存器堆 VLIW DSP register file
  • 相关文献

参考文献6

  • 1[1]Hennessy J L,Patterson D A,et al.Computer architecture:a quantitative approach.3 rd edition.Morgan Kaufmann Publisher,2002
  • 2[2]Fisher J A.Trace scheduling:a technique for global microcode compaction.IEEE Transactions on Computers,1981; C -30:478-490
  • 3[3]Palacharla A S,Jouppi N P,Smith J E.Complexity-effective superscalar processors,in:Eds Hill M,Jouppi N,Sohi G.Proc of Int Symp on Computer Architecture,1997:206-218
  • 4[4]Cruz J L,Gonzlez A,et al.Multiple-banked register file architectures.International Symposium on Computer Architecture,Vancouver,Canada,2000:316-325
  • 5[5]Ho R,Mai Km Horowitz M.The future of wires.Proceedings of the IEEE,2001:490-504
  • 6[6]Agarwal V,Hrishikesh M S,et al.Clock rate versus IPC:the end of the road for conventional microarchitectures.International Symposium on Computer Architecture,Vancouver,Canada,2000

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部