Performance evaluation of matrix multiplication in Virtual Machine

2017 International Conference on Communication, Computing and Digital Systems (C-CODE) Pub Date : 2017-03-01 DOI:10.1109/C-CODE.2017.7918929

Asif Muhammad, Muhammad Arshad Islam

{"title":"Performance evaluation of matrix multiplication in Virtual Machine","authors":"Asif Muhammad, Muhammad Arshad Islam","doi":"10.1109/C-CODE.2017.7918929","DOIUrl":null,"url":null,"abstract":"In this paper, we have examined various implementations of matrix-matrix multiplication using .NET platform. Matrix multiplication is considered one of the basic operation in the field of linear algebra that is used various computer science algorithms. We have used the all loop reordering of traditional n3 sequential algorithm to analyze its behavior on the .NET common language run-time over more than 10 varying sizes of matrices. Moreover we have also analyzed the blocking version of the traditional multiplication algorithm to observe the cache behavior. We have used Intel Corei5 Arandale 2.53 GHz and Haswell 3.30 GHz processors with dual channel RAM for our experiments. Our experiments show that KIJ and IKJ reordering have performed better than the rest of the loop reordering. Furthermore, blocking implementation of matrix multiplication have not been able to gain significantly on .NET platform. In future, we will utilize task parallel library included in .NET 4.5 to gauge the performance efficiency of linear algebraic operations.","PeriodicalId":344222,"journal":{"name":"2017 International Conference on Communication, Computing and Digital Systems (C-CODE)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2017-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2017 International Conference on Communication, Computing and Digital Systems (C-CODE)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/C-CODE.2017.7918929","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 1

Abstract

In this paper, we have examined various implementations of matrix-matrix multiplication using .NET platform. Matrix multiplication is considered one of the basic operation in the field of linear algebra that is used various computer science algorithms. We have used the all loop reordering of traditional n3 sequential algorithm to analyze its behavior on the .NET common language run-time over more than 10 varying sizes of matrices. Moreover we have also analyzed the blocking version of the traditional multiplication algorithm to observe the cache behavior. We have used Intel Corei5 Arandale 2.53 GHz and Haswell 3.30 GHz processors with dual channel RAM for our experiments. Our experiments show that KIJ and IKJ reordering have performed better than the rest of the loop reordering. Furthermore, blocking implementation of matrix multiplication have not been able to gain significantly on .NET platform. In future, we will utilize task parallel library included in .NET 4.5 to gauge the performance efficiency of linear algebraic operations.

查看原文本刊更多论文

虚拟机中矩阵乘法的性能评价

在本文中，我们研究了使用。net平台实现矩阵-矩阵乘法的各种方法。矩阵乘法被认为是线性代数领域的基本运算之一，被各种计算机科学算法所使用。我们使用传统n3序列算法的全循环重排序来分析它在。net公共语言运行时在超过10个不同大小的矩阵上的行为。此外，我们还分析了传统乘法算法的阻塞版本，以观察缓存行为。我们使用Intel Corei5 Arandale 2.53 GHz和Haswell 3.30 GHz双通道RAM处理器进行实验。我们的实验表明，KIJ和IKJ重排序比其他循环重排序表现更好。此外，矩阵乘法的块实现在。net平台上还没有得到显著的发展。将来，我们将利用。net 4.5中包含的任务并行库来衡量线性代数运算的性能效率。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

2017 International Conference on Communication, Computing and Digital Systems (C-CODE)

自引率

0.00%

发文量