本文選題：異構(gòu)多核處理器 + 實(shí)時(shí)系統(tǒng)　； 參考：《東北大學(xué)》2012年碩士論文

【摘要】：異構(gòu)多核處理器中所集成的處理核心類型的多樣性,促成了其對(duì)特殊應(yīng)用(如信號(hào)處理、圖像處理等)的良好支持,并使得在此基礎(chǔ)上的實(shí)時(shí)調(diào)度問(wèn)題成為了多核實(shí)時(shí)領(lǐng)域的主流研究方向之一。 目前,基于異構(gòu)多核處理器架構(gòu)的多資源實(shí)時(shí)任務(wù)分配算法的研究尚處于起步階段,已取得的成果也只是針對(duì)特定的情形,并不能通用于所有類型的任務(wù)集,鑒于目前硬件體系結(jié)構(gòu)的飛速發(fā)展和異構(gòu)多核架構(gòu)在實(shí)際中廣泛的應(yīng)用需求,研究能夠在多資源異構(gòu)多核處理器架構(gòu)下對(duì)任務(wù)進(jìn)行有效分配的算法已是刻不容緩。結(jié)合合理的實(shí)時(shí)任務(wù)調(diào)度算法,對(duì)指定任務(wù)集進(jìn)行分配并調(diào)度執(zhí)行,這也是目前異構(gòu)多核實(shí)時(shí)系統(tǒng)中亟需解決的問(wèn)題。 本文主要研究了異構(gòu)多核處理器架構(gòu)上的實(shí)時(shí)任務(wù)分配問(wèn)題,根據(jù)任務(wù)是否可遷移的特點(diǎn)將實(shí)時(shí)任務(wù)分成兩類,并針對(duì)這兩類任務(wù)分別設(shè)計(jì)了異構(gòu)多核處理器上的分配算法。首先,對(duì)于不能在處理器之間遷移執(zhí)行的任務(wù),本文對(duì)任務(wù)進(jìn)行了進(jìn)一步劃分,通過(guò)對(duì)各類任務(wù)分別采用不同的策略,使算法獲得了較好的算法加速比。對(duì)于可以在不同處理器之間進(jìn)行遷移的任務(wù),論文首先論證了任務(wù)在不同處理器之間遷移后的可串行性；然后對(duì)現(xiàn)有最優(yōu)調(diào)度算法Bfair(Boundary Fair algorithm)進(jìn)行了改進(jìn),使其支持任務(wù)對(duì)非整數(shù)資源的需求；同時(shí)以改進(jìn)后的Bfair算法為基礎(chǔ)設(shè)計(jì)了實(shí)時(shí)任務(wù)在不同類型處理器上的分配算法。最后,本文對(duì)提出的兩種算法加以實(shí)現(xiàn),然后以隨機(jī)任務(wù)集為基礎(chǔ)進(jìn)行了實(shí)驗(yàn),將實(shí)驗(yàn)結(jié)果與現(xiàn)有的異構(gòu)處理器分配算法進(jìn)行比較,驗(yàn)證了本文算法在算法加速比和執(zhí)行時(shí)間方面均優(yōu)于現(xiàn)有的分配算法。本文工作為異構(gòu)多核處理器架構(gòu)上的多資源實(shí)時(shí)任務(wù)分配問(wèn)題提供了兩種解決方案,為多核實(shí)時(shí)系統(tǒng)領(lǐng)域中任務(wù)分配問(wèn)題的應(yīng)用提供了兩種算法基礎(chǔ)。
[Abstract]:The diversity of the types of processing cores integrated in heterogeneous multicore processors has contributed to its good support for special applications, such as signal processing, image processing, etc. On this basis, the real-time scheduling problem has become one of the main research directions in the field of multi-verification. At present, the research of multi-resource real-time task allocation algorithm based on heterogeneous multi-core processor architecture is still in its infancy. In view of the rapid development of hardware architecture and the wide application of heterogeneous multicore architecture in practice, it is urgent to study the algorithm for efficient task allocation in multi-resource heterogeneous multi-core processor architecture. Combined with reasonable real-time task scheduling algorithm, assigned task sets are assigned and executed, which is also a problem that needs to be solved in heterogeneous multi-verification systems. In this paper, the real-time task assignment problem on heterogeneous multicore processor architecture is studied. According to whether the task is transferable or not, the real-time task is divided into two categories, and the assignment algorithms on the heterogeneous multi-core processor are designed respectively for the two kinds of tasks. Firstly, for tasks that can not be migrated between processors, the tasks are further partitioned. Different strategies are adopted to achieve a better speedup of the algorithm. For tasks that can be migrated between different processors, this paper first demonstrates the serializability of tasks migrated between different processors, and then improves the existing optimal scheduling algorithm Bfair(Boundary Fair algorithm. At the same time, based on the improved Bfair algorithm, the assignment algorithm of real-time tasks on different types of processors is designed. Finally, the two algorithms are implemented, and then the experiments are carried out on the basis of the random task set, and the experimental results are compared with the existing heterogeneous processor allocation algorithms. It is verified that the proposed algorithm is superior to the existing algorithms in terms of speedup and execution time. This paper provides two solutions to the multi-resource real-time task allocation problem in heterogeneous multi-core processor architecture, and provides two algorithms for the application of task assignment problem in the multi-verification system domain.
【學(xué)位授予單位】：東北大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2012
【分類號(hào)】：TP332

【參考文獻(xiàn)】

相關(guān)期刊論文 前2條

1 錢(qián)曉龍,唐立新,劉文新;動(dòng)態(tài)調(diào)度的研究方法綜述[J];控制與決策;2001年02期

2 張惠娟,翟鴻鳴,周利華;多處理器系統(tǒng)的實(shí)時(shí)調(diào)度算法研究[J];計(jì)算機(jī)工程與設(shè)計(jì);2004年08期

，

本文編號(hào)：1933475