【摘要】：片上網(wǎng)絡(luò)以其優(yōu)越的性能逐步成為高性能片上通信的解決方案,而溫度作為片上網(wǎng)絡(luò)工作過(guò)程中的一個(gè)重要參數(shù),影響芯片的整體性能。局部“熱點(diǎn)”的產(chǎn)生會(huì)使片上網(wǎng)絡(luò)性能降低,甚至造成器件損壞。針對(duì)局部“熱點(diǎn)”的問(wèn)題,本文基于動(dòng)靜態(tài)映射的方法,設(shè)計(jì)出一種監(jiān)測(cè)與任務(wù)加載電路,實(shí)時(shí)監(jiān)測(cè)片上網(wǎng)絡(luò)的工作狀態(tài),并重新配置映射任務(wù),以降低“熱點(diǎn)”溫度。此外針對(duì)任務(wù)間一對(duì)多的通訊關(guān)系,提出一種基于層次化網(wǎng)絡(luò)結(jié)構(gòu)的多播數(shù)據(jù)包拆分傳輸策略,并設(shè)計(jì)實(shí)現(xiàn)出多播片上網(wǎng)絡(luò)。本文的主要工作如下:(1)基于映射的動(dòng)態(tài)熱管理方法針對(duì)片上網(wǎng)絡(luò)處理任務(wù)時(shí),所產(chǎn)生的局部“熱點(diǎn)”問(wèn)題,基于動(dòng)態(tài)與靜態(tài)映射相結(jié)合的方法,設(shè)計(jì)實(shí)現(xiàn)了監(jiān)測(cè)與任務(wù)加載硬件電路。片上網(wǎng)絡(luò)工作時(shí),硬件電路實(shí)時(shí)監(jiān)測(cè)片上網(wǎng)絡(luò)的運(yùn)行狀態(tài)并收集相關(guān)參數(shù),上傳到上位機(jī)。上位機(jī)處理收集到的數(shù)據(jù),得出局部“熱點(diǎn)”位置,然后重新配置任務(wù)在網(wǎng)絡(luò)中的分布,以達(dá)到“熱點(diǎn)”的消除和轉(zhuǎn)移效果。實(shí)驗(yàn)結(jié)果表明,這種方法能夠有效降低“熱點(diǎn)”溫度,最大可降低31.6%。(2)基于層次化網(wǎng)絡(luò)結(jié)構(gòu)的多播數(shù)據(jù)包拆分傳輸在已有基于層次化網(wǎng)絡(luò)結(jié)構(gòu)的多播路由算法的基礎(chǔ)上,加入了一種多播數(shù)據(jù)包兩級(jí)拆分機(jī)制。該機(jī)制針對(duì)上行路徑上以及上行路徑相鄰的目的節(jié)點(diǎn)。為了避免引入新的死鎖情況且兼顧多播數(shù)據(jù)包的整體延遲,采用空閑拆分策略。只有當(dāng)對(duì)應(yīng)的拆分輸出端口空閑時(shí),才拆分傳輸。在不增加算法實(shí)現(xiàn)復(fù)雜度的前提下,降低了多播數(shù)據(jù)包的傳輸延遲,并設(shè)計(jì)實(shí)現(xiàn)出相應(yīng)的多播片上網(wǎng)絡(luò)。實(shí)驗(yàn)結(jié)果表明,相對(duì)于不拆分傳輸方式,傳輸延遲最大降低了50.3%。
[Abstract]:With its superior performance, the on-chip network has gradually become the solution of high-performance on-chip communication, and temperature, as an important parameter in the on-chip network operation process, affects the overall performance of the chip. The generation of local "hot spots" will degrade the performance of the on-chip network and even cause device damage. Aiming at the problem of local "hot spot", based on the method of dynamic and static mapping, this paper designs a monitoring and task loading circuit to monitor the working state of the on-chip network in real time, and reconfigure the mapping task in order to reduce the "hot spot" temperature. In addition, aiming at the one-to-many communication relationship between tasks, a multicast packet split transmission strategy based on hierarchical network structure is proposed, and a multicast on-chip network is designed and implemented. The main work of this paper is as follows: (1) the mapping based dynamic thermal management method is based on the method of combining dynamic and static mapping to solve the local "hot spot" problem caused by on-chip network processing. The hardware circuit of monitoring and task loading is designed and implemented. When the on-chip network works, the hardware circuit monitors the running state of the on-chip network in real time, collects the relevant parameters, and uploads it to the host computer. The host computer processes the collected data, obtains the local "hot spot" position, and then reconfigures the distribution of tasks in the network in order to achieve the elimination and transfer effect of "hot spot". The experimental results show that this method can effectively reduce the "hot spot" temperature, and can reduce the maximum value of 31.6b. (2) the multicast packet split transmission based on hierarchical network structure is based on the existing multicast routing algorithm based on hierarchical network structure. A two-level split mechanism for multicast packets is added. The mechanism is aimed at the destination nodes on the uplink path and adjacent to the uplink path. In order to avoid introducing new deadlock and taking into account the overall delay of multicast packets, an idle split strategy is adopted. Split the transfer only when the corresponding split output port is idle. Without increasing the complexity of the algorithm, the transmission delay of multicast packets is reduced, and the corresponding multicast on-chip network is designed and implemented. The experimental results show that the maximum transmission delay is reduced by 50.3% compared with the non-split transmission mode.
【學(xué)位授予單位】：合肥工業(yè)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：TN47

【參考文獻(xiàn)】

相關(guān)期刊論文 前9條

1 張穎;吳寧;葛芬;;基于路徑的無(wú)死鎖NoC多播路由器的設(shè)計(jì)[J];微電子學(xué)與計(jì)算機(jī);2014年11期

2 胡哲琨;陳杰;;基于層次化片上網(wǎng)絡(luò)的混合多播路由算法[J];哈爾濱工程大學(xué)學(xué)報(bào);2013年10期

3 郝曉潔;顧華璽;尚軍輝;;片上網(wǎng)絡(luò)調(diào)度算法相關(guān)研究[J];中國(guó)集成電路;2011年04期

4 趙建武;師奕兵;王志剛;;復(fù)用NoC測(cè)試SoC內(nèi)嵌IP芯核的測(cè)試規(guī)劃研究[J];計(jì)算機(jī)工程與應(yīng)用;2010年15期

5 陳芳露;陸雯青;虞志益;周曉方;;一種新型片上網(wǎng)絡(luò)互連結(jié)構(gòu)的仿真和實(shí)現(xiàn)[J];小型微型計(jì)算機(jī)系統(tǒng);2010年05期

6 林世俊;張凡;金德鵬;曾烈光;;分布式同步的GALS片上網(wǎng)絡(luò)及其接口設(shè)計(jì)[J];清華大學(xué)學(xué)報(bào)(自然科學(xué)版);2008年01期

7 張恒龍;顧華璽;王長(zhǎng)山;;片上網(wǎng)絡(luò)拓?fù)浣Y(jié)構(gòu)的研究[J];中國(guó)集成電路;2007年11期

8 王宏偉;陸俊林;佟冬;程旭;;層次化片上網(wǎng)絡(luò)結(jié)構(gòu)的簇生成算法[J];電子學(xué)報(bào);2007年05期

9 高明倫;杜高明;;NoC:下一代集成電路主流設(shè)計(jì)技術(shù)[J];微電子學(xué);2006年04期

相關(guān)博士學(xué)位論文 前1條

1 朱曉靜;片上網(wǎng)絡(luò)的結(jié)構(gòu)設(shè)計(jì)與性能分析[D];中國(guó)科學(xué)技術(shù)大學(xué);2008年

相關(guān)碩士學(xué)位論文 前3條

1 劉鑫;基于離散螢火蟲的眾核SoC映射方法研究[D];合肥工業(yè)大學(xué);2015年

2 鄭國(guó)海;多播片上網(wǎng)絡(luò)路由器的設(shè)計(jì)與實(shí)現(xiàn)[D];西安電子科技大學(xué);2014年

3 鮑培蕾;面向千核片上網(wǎng)絡(luò)拓?fù)浣Y(jié)構(gòu)和路由器的研究[D];西安電子科技大學(xué);2014年

，

本文編號(hào)：2141740