摘要: 深入研究了YHFT_Matrix高性能DSP中的一种多线程机制,重点介绍了其循环指令缓冲的读写机制、单线程与多线程之间的模式切换机制。在基于65nm工艺下,经过综合,代码面积、功耗都有减少,关键路径优化0.07ns。对程序的执行评估测试的分析结果表明:多线程工作模式相比单线程工作模式,其处理器性能IPC(Instructions Per Cycle)平均提高了9.64%。
[1] Merten M C,et al.Modulo Scheduling Buffer[C]∥Proceeding of the 34th annual ACM international symposium on Microarchitecture.Texas,Dec.2001 [2] Hennessy J L,Patterson D A.Computer Architecture:A Quantitative approach(4th Edition)[M].Elsevier,2007:274-363 [3] Ramakrishna B,Schlansker M S,Tirumalai P.Code Generation Schema for Modulo Scheduled Loops[C]∥Proceeding of MICRO-25.1992:297-326 [4] Grunewald W,Ungerer T.Towards Extremely Fast ContextSwitching in a Block Multithreaded Processor[C]∥Proceedings of the 22nd EUROMICRO Conference.September 1996 [5] Gwennap L.MAJC Gives VLIW a New Twist[J].Microprocessor Report,1999,13(12) [6] Aa T V,Jayapala M,Barat F,et al.Instruction Buffer Exploration for Low Energy VLIWs with Instruction Clusters[C]∥Proceedings of the 2004conference on Asia South Pacific design automation.Yokohama,Japan,Jan.2004:163-247 [7] Tullsen T D,Eggers S,Emer J,et al.Exploiting Choice:Instruction Fetch and Issue on an Implementable Simultaneous Multithreading Processor[C]∥Proceedings of the 23rd Annual International Symposium on Computer Architecture.Philadelphia,May 1996 |
No related articles found! |
|