本文選題：MPI + 聚合通信　； 參考：《國(guó)防科學(xué)技術(shù)大學(xué)》2014年碩士論文

【摘要】：在現(xiàn)代MPI并行應(yīng)用中,聚合通信操作被廣泛使用。聚合通信操作不僅在科學(xué)計(jì)算上占據(jù)大部分時(shí)間消耗(有時(shí)可高達(dá)70%),同時(shí)也為程序員提供更便捷編程接口。然而軟件實(shí)現(xiàn)的聚合通信操作是基于點(diǎn)對(duì)點(diǎn)操作實(shí)現(xiàn)的,當(dāng)系統(tǒng)增大時(shí),由于通信步驟、計(jì)算數(shù)據(jù)量和通信距離都增加了,聚合通信的耗時(shí)增加的很快,因此其可擴(kuò)展性很差。但應(yīng)用程序卻隨著系統(tǒng)的增大對(duì)于可擴(kuò)展性有了越來(lái)越高的要求,因此,對(duì)聚合通信進(jìn)行優(yōu)化變得也更加重要了。加速聚合通信操作的一個(gè)有效的方法是在網(wǎng)卡端使用卸載模式(offload),即在網(wǎng)卡端加入一些特殊部件來(lái)輔助處理機(jī)進(jìn)行聚合通信的數(shù)據(jù)移動(dòng)或者數(shù)據(jù)計(jì)算,也稱為軟硬件結(jié)合的聚合通信操作�！疤旌印备咚倩ミB網(wǎng)絡(luò)使用在網(wǎng)卡端加入的觸發(fā)邏輯來(lái)卸載處理機(jī)端的數(shù)據(jù)移動(dòng)任務(wù),本文即以“天河”高速互連網(wǎng)絡(luò)為基礎(chǔ),研究MPI聚合通信優(yōu)化方法,主要取得了如下成果:1)對(duì)α-β模型進(jìn)行擴(kuò)展,通過擴(kuò)展的模型計(jì)算出基于點(diǎn)對(duì)點(diǎn)的聚合通信操作時(shí)延,與基于offload實(shí)現(xiàn)的聚合通信進(jìn)行對(duì)比。經(jīng)典的α-β模型只能適合聚合通信操作的定性分析,而不能進(jìn)行定量的分析,擴(kuò)展后的α-β模型能夠?qū)酆贤ㄐ挪僮鬟M(jìn)行定量分析,為后面與offload模式的聚合通信操作進(jìn)行對(duì)比;2)提出基于offload實(shí)現(xiàn)的聚合通信評(píng)價(jià)模型,通過測(cè)試,能較好的預(yù)測(cè)實(shí)測(cè)數(shù)據(jù),并為后面同步和廣播操作的算法設(shè)計(jì)提供理論分析�；趏ffload實(shí)現(xiàn)的聚合通信操作評(píng)價(jià)模型的提出為之后我們?cè)凇疤旌印备咚倩ミB網(wǎng)絡(luò)上進(jìn)行聚合通信優(yōu)化提供了理論依據(jù),不僅如此,也指導(dǎo)我們后面的聚合通信操作的優(yōu)化;3)對(duì)基于offload實(shí)現(xiàn)的柵欄同步和廣播操作進(jìn)行算法優(yōu)化和測(cè)試分析并且對(duì)基于offload實(shí)現(xiàn)的收集操作進(jìn)行算法設(shè)計(jì)。柵欄同步操作和廣播操作為最常用的聚合通信操作,也可以應(yīng)用于許多聚合通信,我們實(shí)現(xiàn)了k-ary樹和k-nomial樹兩種算法的柵欄同步操作和廣播操作,同時(shí)從理論和實(shí)際測(cè)試給出兩種操作在性能取得最優(yōu)值時(shí)的k值,根據(jù)理論模型,基于觸發(fā)的聚合通信操作擁有良好的可擴(kuò)展性;本文對(duì)擴(kuò)展的α-β模型和基于offload實(shí)現(xiàn)的聚合通信操作評(píng)價(jià)模型進(jìn)行了實(shí)驗(yàn)驗(yàn)證,實(shí)驗(yàn)結(jié)果表明兩種模型均能較好的定量分析軟件和基于offload方式實(shí)現(xiàn)的聚合通信操作,并且模型也為后面進(jìn)行聚合通信操作優(yōu)化提供理論依據(jù)。同時(shí)在64個(gè)節(jié)點(diǎn)規(guī)模時(shí),基于offload的柵欄同步操作比基于點(diǎn)對(duì)點(diǎn)的柵欄同步操作性能提高了2.17倍,對(duì)offload實(shí)現(xiàn)的柵欄同步和廣播操作進(jìn)行優(yōu)化,優(yōu)化后的柵欄同步操作性能進(jìn)一步提高1.1倍,廣播操在性能提高1.46倍。
[Abstract]:In modern MPI parallel applications, aggregate communication operations are widely used. Aggregate communication not only takes up most of the time (sometimes up to 70) in scientific computing, but also provides a more convenient programming interface for programmers. However, the aggregate communication operation implemented by software is based on point-to-point operation. When the system increases, the computation data and communication distance increase because of the communication step, and the time consuming of aggregate communication increases quickly, so its scalability is very poor. However, the application has more and more high requirements for scalability as the system increases, so it is more important to optimize the aggregate communication. An effective way to speed up aggregate communication is to use the offloading mode on the network card side, that is, to add some special components to the network card to assist the processor in the data movement or data calculation of the aggregate communication. Also known as a combination of hardware and software communication operations. The "Tianhe" high-speed interconnection network uses the trigger logic added to the network card to unload the data movement task of the processor. This paper studies the optimization method of MPI aggregation communication based on the "Tianhe" high-speed interconnection network. The main achievements are as follows: 1) the 偽-尾 model is extended, and the operation delay based on point to point aggregation is calculated by the extended model, which is compared with the aggregate communication based on offload. The classical 偽-尾 model can only be used for qualitative analysis of polymeric communication operations, but not for quantitative analysis. The extended 偽-尾 model can be used for quantitative analysis of polymeric communication operations. The evaluation model of aggregate communication based on offload is put forward for comparing with the aggregate communication operation of offload mode. Through testing, the measured data can be well predicted, and the theoretical analysis is provided for the algorithm design of synchronization and broadcast operation. The proposed operation evaluation model of aggregate communication based on offload provides a theoretical basis for the optimization of aggregate communication in Tianhe high-speed interconnection network. It also instructs us to optimize and test the algorithm of fence synchronization and broadcast operation based on offload, and to design the algorithm of collecting operation based on offload. As the most commonly used aggregate communication operation, the fence synchronous operation and broadcast operation can also be applied to many aggregate communications. We have realized the fence synchronization operation and broadcast operation of k-ary tree and k-nomial tree algorithms. At the same time, from the theoretical and practical tests, the k value of the two operations is given when the performance is optimal. According to the theoretical model, the triggered aggregate communication operation has good scalability. In this paper, the extended 偽-尾 model and the aggregate communication operation evaluation model based on offload are verified by experiments. The experimental results show that both models can be used in quantitative analysis software and aggregate communication operation based on offload. The model also provides the theoretical basis for the optimization of aggregation communication operation. At the same time, at 64 nodes, the performance of fence synchronization based on offload is 2.17 times higher than that based on point-to-point. The synchronization and broadcast operation of fence based on offload is optimized. The performance of the optimized synchronous operation of the fence is further improved by 1.1 times, and the performance of broadcast operation is improved by 1.46 times.
【學(xué)位授予單位】：國(guó)防科學(xué)技術(shù)大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2014
【分類號(hào)】：TP393.03

【參考文獻(xiàn)】

相關(guān)期刊論文 前3條

1 楊學(xué)軍;廖湘科;盧凱;胡慶豐;宋君強(qiáng);蘇金樹;;The TianHe-1A Supercomputer: Its Hardware and Software[J];Journal of Computer Science & Technology;2011年03期

2 李占勝;畢會(huì)娟;都志輝;焦青;;MPI_ALLGATHER實(shí)現(xiàn)算法的改進(jìn)與性能分析[J];計(jì)算機(jī)工程與應(yīng)用;2007年27期

3 陳靖;張?jiān)迫?張林波;袁偉;;一種新的MPI Allgather算法及其在萬(wàn)億次機(jī)群系統(tǒng)上的實(shí)現(xiàn)與性能分析[J];計(jì)算機(jī)學(xué)報(bào);2006年05期

，

本文編號(hào)：1966644