本文選題：眾核處理器　切入點(diǎn)：Cache一致性　出處：《上海交通大學(xué)》2013年碩士論文　論文類型：學(xué)位論文

【摘要】：隨著集成電路技術(shù)的進(jìn)步和處理器體系結(jié)構(gòu)技術(shù)的發(fā)展，眾核處理器已經(jīng)成為未來微處理器的一個(gè)重要發(fā)展方向，是當(dāng)前學(xué)術(shù)界和工業(yè)界的研究熱點(diǎn)。由于處理器核數(shù)眾多，，如何利用龐大的計(jì)算資源，解決核數(shù)增加帶來的存儲(chǔ)器瓶頸，應(yīng)對(duì)任務(wù)的動(dòng)態(tài)、階段性資源需求是眾核處理器需要解決的重大問題。 論文涉及的主要工作有：1）提出動(dòng)態(tài)子網(wǎng)劃分的方案處理眾核處理器遇到的資源利用問題和動(dòng)態(tài)資源需求問題。2）為支持動(dòng)態(tài)子網(wǎng)劃分的方案，設(shè)計(jì)基于Token的動(dòng)態(tài)可重構(gòu)的Cache一致性協(xié)議。3）在Gem5仿真器中實(shí)現(xiàn)基于Token的動(dòng)態(tài)可重構(gòu)的Cache一致性協(xié)議并進(jìn)行驗(yàn)證。 在本課題動(dòng)態(tài)子網(wǎng)劃分的環(huán)境中，基于Token的動(dòng)態(tài)可重構(gòu)Cache一致性協(xié)議將在程序執(zhí)行時(shí)間、網(wǎng)絡(luò)通信代價(jià)等主要性能指標(biāo)上優(yōu)于目錄協(xié)議和Token協(xié)議。實(shí)驗(yàn)結(jié)果顯示，在16核Mesh網(wǎng)絡(luò)中，基于Token的動(dòng)態(tài)可重構(gòu)Cache一致性協(xié)議在程序執(zhí)行時(shí)間上比Token協(xié)議降低10%，而通信代價(jià)比目錄協(xié)議降低30%以上。鑒于這一結(jié)果，基于Token的動(dòng)態(tài)可重構(gòu)Cache一致性協(xié)議將能夠有效的應(yīng)用到眾核處理器結(jié)構(gòu)中。
[Abstract]:With the progress of integrated circuit technology and the development of processor architecture technology, multicore processor has become an important development direction of microprocessor in the future. How to make use of the huge computing resources to solve the memory bottleneck caused by the increasing number of cores and to deal with the dynamic of the task are the major problems that need to be solved by the multi-core processor. The main work involved in this paper is: (1) A scheme of dynamic subnet partitioning is proposed to deal with the resource utilization problem and dynamic resource requirement problem encountered by multi-core processors. A dynamic reconfigurable Cache consistency protocol based on Token. 3) is designed to implement and verify the dynamically reconfigurable Cache consistency protocol based on Token in the Gem5 simulator. In the dynamic subnet partition environment, the dynamic reconfigurable Cache consistency protocol based on Token will outperform directory protocol and Token protocol in the main performance indexes such as program execution time, network communication cost and so on. In 16-core Mesh networks, the dynamic reconfigurable Cache conformance protocol based on Token reduces the program execution time by 10 times, while the communication cost is more than 30% lower than that of the directory protocol. Dynamic reconfigurable Cache conformance protocol based on Token can be effectively applied to multi-core processor architecture.
【學(xué)位授予單位】：上海交通大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2013
【分類號(hào)】：TP332

【相似文獻(xiàn)】

相關(guān)期刊論文 前10條

1 戴梅萼,史嘉權(quán);計(jì)算機(jī)系統(tǒng)存儲(chǔ)體設(shè)計(jì)技術(shù)的探討[J];微型機(jī)與應(yīng)用;1995年03期

2 周軍;病毒能藏在CMOS和CACHE中嗎?[J];電腦愛好者;1998年12期

3 尚毅,向旭光;怎樣使奔騰586奔起來[J];實(shí)驗(yàn)技術(shù)與管理;1996年02期

4 周

本文編號(hào)：1630305