本文關(guān)鍵詞：片上網(wǎng)絡(luò)有環(huán)通信集的實(shí)時(shí)調(diào)度研究　出處：《東北大學(xué)》2012年碩士論文　論文類型：學(xué)位論文

  更多相關(guān)文章： 實(shí)時(shí)系統(tǒng) 片上網(wǎng)絡(luò) 有環(huán)通信集 實(shí)時(shí)調(diào)度算法 虛通道 時(shí)間隔離

【摘要】：傳統(tǒng)的依靠提升處理器頻率來獲得高計(jì)算性能的方法已接近極限,已不能滿足人們對高計(jì)算性能的要求,所以多核成為新的發(fā)展方向。而隨著處理核心的增多,傳統(tǒng)的總線通信架構(gòu)成為影響系統(tǒng)性能提升的瓶頸,所以片上網(wǎng)絡(luò)(Network-on-Chip)應(yīng)運(yùn)而生。片上網(wǎng)絡(luò)讓不同任務(wù)可以進(jìn)行并行通信,極大提升了系統(tǒng)的通信效率。然而當(dāng)任務(wù)間通信十分頻繁時(shí),仍會發(fā)生對通信資源的爭用,特別在實(shí)時(shí)系統(tǒng)中對任務(wù)的完成時(shí)間有嚴(yán)格限定,通信資源的爭用會導(dǎo)致實(shí)時(shí)任務(wù)完成時(shí)間的不確定性增加,所以需要對運(yùn)行在在片上網(wǎng)絡(luò)的實(shí)時(shí)系統(tǒng)的通信進(jìn)行研究來解決這一問題。 由于存在計(jì)算和通信的雙重并行性,所以將實(shí)時(shí)系統(tǒng)應(yīng)用于NoC(Network-on-Chip)需要一個(gè)新的理論框架。與傳統(tǒng)的實(shí)時(shí)系統(tǒng)不同,片上網(wǎng)絡(luò)結(jié)構(gòu)的多核SoC(System-on-Chip)對實(shí)時(shí)通信任務(wù)集進(jìn)行調(diào)度并使片上網(wǎng)絡(luò)的利用率最大化是一個(gè)新的挑戰(zhàn)。盡管和多資源調(diào)度有類似的形式,片上網(wǎng)絡(luò)實(shí)時(shí)通信集調(diào)度有其不同之處。因?yàn)橐粋�(gè)通信任務(wù)可能同時(shí)多個(gè)物理鏈路,面臨物理鏈路的競爭。 以往對片上網(wǎng)絡(luò)實(shí)時(shí)通信的研究大都集中在對無環(huán)通信任務(wù)集可調(diào)度性分析,而很少有對有環(huán)通信任務(wù)集的研究,然而實(shí)際應(yīng)用中有環(huán)通信任務(wù)集是普遍存在的。通信任務(wù)集的環(huán)是指片上網(wǎng)絡(luò)中的通信任務(wù)的路由路徑相互疊加而形成的環(huán)路。 本論文中針對有環(huán)的通信任務(wù)集提出了兩種調(diào)度算法：基于虛通道的實(shí)時(shí)調(diào)度算法(CS-VC)和基于時(shí)間隔離的實(shí)時(shí)調(diào)度算法(CS-TI),并給出了通信任務(wù)集用這兩種算法可調(diào)度的通信任務(wù)集利用率界限(也是通信任務(wù)集可調(diào)度的充分條件)。實(shí)驗(yàn)部分將CS-VC和CS-TI與目前最好的有環(huán)通信任務(wù)集實(shí)時(shí)調(diào)度算法FPA (Fixed Priority Algorithm)進(jìn)行了比較,實(shí)驗(yàn)結(jié)果表明CS-VC和CS-TI有更高的接收率,并且當(dāng)通信任務(wù)集和片上網(wǎng)絡(luò)規(guī)模越小時(shí),CS-VC和CS-TI的接收率越高。CS-TI與CS-VC相比雖然接收率較低,但CS-TI無需額外的片上網(wǎng)絡(luò)結(jié)構(gòu)作為支持更便于實(shí)現(xiàn)。
[Abstract]:The traditional method of improving processor frequency to achieve high computing performance is approaching the limit, which can not meet the requirements of high computing performance, so multi-core has become a new direction of development, and with the increase of processing cores. Traditional bus communication architecture has become the bottleneck to improve the performance of the system, so Network-on-Chip emerges as the times require. The on-chip network allows different tasks to communicate in parallel. The communication efficiency of the system is greatly improved. However, when the communication between tasks is very frequent, there will still be contention for communication resources, especially in real-time systems, the completion time of tasks is strictly limited. The contention of communication resources will lead to the uncertainty of the completion time of real-time task, so it is necessary to study the communication of real-time system running in on-chip network to solve this problem. Due to the dual parallelism of computing and communication, the application of real-time systems to NoCU Network-on-Chip requires a new theoretical framework, which is different from the traditional real-time systems. Multi-core SoCn System-on-Chip with on-Chip Network structure. Scheduling real-time communication task sets and maximizing the utilization of on-chip networks is a new challenge, although it has a similar form to multi-resource scheduling. The scheduling of real-time communication set in on-chip network has its differences, because a communication task may have multiple physical links at the same time, so it faces the competition of physical links. In the past, the research of real-time communication in on-chip network mostly focused on the schedulability analysis of the task set of acyclic communication, but there was little research on the task set of ring communication. However, in practical applications, ring communication task sets are ubiquitous, and the ring of communication task set refers to the loop formed by the superposition of the routing paths of communication tasks in the on-chip network. In this paper, we propose two scheduling algorithms for the cyclic communication task set: virtual channel based real-time scheduling algorithm (CS-VC) and time isolation based real-time scheduling algorithm (CS-TI). The limit of the utilization ratio of the communication task set is given, which is also a sufficient condition for the schedulability of the communication task set. In the experiment part, CS-VC and CS-TI are compared with FPA fixed Priority algorithm, which is the best real-time scheduling algorithm for ring communication task set. A comparison was made. The experimental results show that CS-VC and CS-TI have higher reception rate, and when the communication task set and on-chip network scale is smaller. The higher the receiving rate of CS-VC and CS-TI is, the lower the receiving rate of CS-TI is compared with that of CS-VC, but it is easier to implement CS-TI without additional on-chip network structure.
【學(xué)位授予單位】：東北大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2012
【分類號】：TP332;TN47

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 王崢;顧華璽;楊燁;樂天助;;片上網(wǎng)絡(luò)交換機(jī)制的研究[J];中國集成電路;2007年12期

2 王芳莉;杜慧敏;;片上網(wǎng)絡(luò)路由算法綜述[J];西安郵電學(xué)院學(xué)報(bào);2011年01期

，

本文編號：1427849