{"title":"Optimizing utilization across XSEDE platforms","authors":"Haihang You, Charng-Da Lu, Ziliang Zhao, Fei Xing","doi":"10.1145/2484762.2484778","DOIUrl":null,"url":null,"abstract":"HPC resources provided by XSEDE give researchers unique opportunities to carry out scientific studies. As of 2013 XSEDE consists of 16 systems with varied architectural designs and capabilities. The hardware heterogeneity and software diversity make efficient utilization of such a federation of computing resources very challenging. For example, users are constantly faced with a myriad of possibilities to build and run an application: compilers, numerical libraries, and runtime parameters. In this paper we report performance data of several popular scientific applications built with different compilers and numerical libraries available on two XSEDE systems: Kraken and Gordon, and suggest the best way to compile applications for optimal performance. By comparison, we validate SU conversion factors between the aforementioned XSEDE systems from application's viewpoint.","PeriodicalId":426819,"journal":{"name":"Proceedings of the Conference on Extreme Science and Engineering Discovery Environment: Gateway to Discovery","volume":"10 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2013-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the Conference on Extreme Science and Engineering Discovery Environment: Gateway to Discovery","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1145/2484762.2484778","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
HPC resources provided by XSEDE give researchers unique opportunities to carry out scientific studies. As of 2013 XSEDE consists of 16 systems with varied architectural designs and capabilities. The hardware heterogeneity and software diversity make efficient utilization of such a federation of computing resources very challenging. For example, users are constantly faced with a myriad of possibilities to build and run an application: compilers, numerical libraries, and runtime parameters. In this paper we report performance data of several popular scientific applications built with different compilers and numerical libraries available on two XSEDE systems: Kraken and Gordon, and suggest the best way to compile applications for optimal performance. By comparison, we validate SU conversion factors between the aforementioned XSEDE systems from application's viewpoint.
京公网安备 11010802042870号
京ICP备2023020795号-1
分享
求助内容：
应助结果提醒方式：
扫码关注我们
