胡淑珂 龐留勇 周向前
摘 要:隨著經(jīng)濟(jì)的快速增長(zhǎng),很多產(chǎn)品的運(yùn)作生產(chǎn),往往需要不同的工藝流程。文章考慮了帶有時(shí)間懲罰的有向串并聯(lián)圖最小任務(wù)分配問(wèn)題和帶有時(shí)間懲罰的有向串并聯(lián)圖最小跨度任務(wù)分配問(wèn)題,根據(jù)有向串并聯(lián)任務(wù)優(yōu)先圖結(jié)構(gòu)構(gòu)建有向串并聯(lián)分配圖。在此基礎(chǔ)上,文章應(yīng)用時(shí)間復(fù)雜性更小的特殊結(jié)構(gòu)最短路算法以及收縮方法,分別設(shè)計(jì)了兩個(gè)時(shí)間復(fù)雜性為O(nm2k2)的多項(xiàng)式算法來(lái)解決帶時(shí)間懲罰的有向串并聯(lián)圖任務(wù)分配問(wèn)題,其中n為任務(wù)數(shù)量,m為處理器數(shù)量,k是時(shí)間段限制數(shù)量。
關(guān)鍵詞:有向串并聯(lián)圖;任務(wù)分配;多項(xiàng)式算法;時(shí)間限制懲罰;最小成本;最小跨度
中圖分類號(hào):O224? 文獻(xiàn)標(biāo)識(shí)碼:A? 文章編號(hào):1673-260X(2023)02-0014-06
3.2 模型構(gòu)建
定理2:Min-MS-PSA算法是帶有時(shí)間懲罰的有向串并聯(lián)圖最小跨度分配問(wèn)題的一個(gè)精確算法,其時(shí)間復(fù)雜性為O(nm2k2),n為任務(wù)數(shù)量,m為處理器數(shù)量,k為時(shí)間段限制數(shù)量。
證明:Min-MS-PSA算法區(qū)別于Min-PSA算法的是當(dāng)分支收縮合并為兩層有向結(jié)構(gòu)時(shí),目標(biāo)是在異構(gòu)多核處理器系統(tǒng)下使得整個(gè)過(guò)程跨度最小,算法將局部收縮的兩層有向結(jié)構(gòu)對(duì)應(yīng)頂點(diǎn)之間的弧權(quán)重置為并行最短路中權(quán)重最大值,最后能夠得到帶有時(shí)間懲罰的有向串并聯(lián)圖最小跨度分配問(wèn)題的全局最優(yōu)分配解。時(shí)間復(fù)雜性同定理1證明。
4 結(jié)論
本文提出的帶有時(shí)間懲罰的有向串并聯(lián)圖任務(wù)分配問(wèn)題靜態(tài)Min-PSA算法和Min-MS-PSA算法是通過(guò)逐步獲得局部最優(yōu)解,并收縮合并最后獲得問(wèn)題的全局最優(yōu)解。Min-PSA算法和Min-MS-PSA算法都是基于并行SHORT算法來(lái)解決帶有時(shí)間懲罰的有向串并聯(lián)圖最小成本分配問(wèn)題和帶有時(shí)間懲罰的有向串并聯(lián)圖最小跨度分配問(wèn)題,其時(shí)間復(fù)雜度均為O(nm2k2)。
本文是基于帶有時(shí)間懲罰的有向串并聯(lián)圖結(jié)構(gòu)的基礎(chǔ)上對(duì)任務(wù)分配問(wèn)題進(jìn)行分配,大大提高任務(wù)執(zhí)行的效率,從而提高生產(chǎn)生活中的效益。同樣適用于無(wú)時(shí)間懲罰的有向串并聯(lián)圖的任務(wù)分配問(wèn)題,更多特殊結(jié)構(gòu)的任務(wù)分配以及調(diào)度問(wèn)題值得進(jìn)一步探討。
——————————
參考文獻(xiàn):
〔1〕H.S. Stone. Multiprocessor scheduling with the aid of network flow algorithms[J]. IEEE Transactions on Software Engineering, 1977, 3(01): 85-93.
〔2〕V. M. Lo. Heuristic algorithms for task assignment in distributed systems[J]. IEEE Transactions on Computers, 1988,37(11): 1384-1397.
〔3〕S.H. Bokhari. On the mapping problem[J]. IEEE Transactions on Computers,1981, 30(03):207-214.
〔4〕D. Fernandez-Baca. Allocating modules to processors in a distributed system[J]. IEEE Transactions on Software Engineering, 1989, 15(11): 1427-1436.
〔5〕S.H. Bokhari. A shortest tree algorithm for optimal assignments across space and time in a distributed processor system[J]. IEEE Transactions on Software Engineering, 1981, 7(06): 583-589.
〔6〕Alain Billionnet. Allocating Tree Structured Programs in a Distributed System with Uniform Communication Costs.[J]. IEEE Trans. Parallel Distrib. Syst.,1994,5(04): 445-448.
〔7〕D. Towsley. Allocating programs containing branches and loops within a multiple processor system[J]. IEEE Transactions on Software Engineering, 1986, 12(10): 1018-1024.
〔8〕劉軼,張昕,李鶴,錢德沛.多核處理器大規(guī)模并行系統(tǒng)中的任務(wù)分配問(wèn)題及算法[J].小型微型計(jì)算機(jī)系統(tǒng),2008,29(05):972-975.
〔9〕S.H. Bokhari. Assignment Problems in Parallel and Distributed Computing[M]. Boston: Kluwer Academic Publishers, 1987.
〔10〕趙國(guó)亮,李云飛,王川.異構(gòu)多核系統(tǒng)任務(wù)調(diào)度算法研究[J].計(jì)算機(jī)工程與設(shè)計(jì),2014,35(09):3099-3106.
〔11〕Q. Zhuge, C. Xue, E.H.M. Sha. Efficient assignment and scheduling for heterogeneous DSP systems[J]. IEEE Transactions on Parallel and Distributed Systems: A Publication of the IEEE Computer Society, 2005, 16(06): 516-525.
〔12〕Jacobo Valdes,Robert E. Tarjan,Eugene L. Lawler. The Recognition of Series Parallel Digraphs[J]. SIAM Journal on Computing,1982,11(02): 298-313.
收稿日期:2022-09-15
基金項(xiàng)目:國(guó)家自然科學(xué)基金項(xiàng)目(12171193);河南省高等學(xué)校重點(diǎn)科研項(xiàng)目(23B110012;23B110009;22B110006;21A110015;20B110008);河南省科技攻關(guān)項(xiàng)目(212102310464)