Nischay Ram Mamidi, D. Saxena, K. Prasun, Anil Nemili, Bharatkumar Sharma, S. Deshpande
{"title":"Performance analysis of GPU accelerated meshfree q-LSKUM solvers in Fortran, C, Python, and Julia","authors":"Nischay Ram Mamidi, D. Saxena, K. Prasun, Anil Nemili, Bharatkumar Sharma, S. Deshpande","doi":"10.1109/HiPC56025.2022.00031","DOIUrl":null,"url":null,"abstract":"This paper presents a comprehensive analysis of the performance of Fortran, C, Python, and Julia based GPU accelerated meshfree solvers for compressible flows. The programming model CUDA is used to develop the GPU codes. The meshfree solver is based on the least squares kinetic upwind method with entropy variables (q-LSKUM). To measure the performance of baseline codes, benchmark calculations are performed. The codes are then profiled to investigate the differences in their performance. Analysing various performance metrics for the computationally expensive flux residual kernel helped identify various bottlenecks in the codes. To resolve the bottlenecks, several optimisation techniques are employed. Post optimisation, the performance metrics have improved significantly, with the C GPU code exhibiting the best performance.","PeriodicalId":119363,"journal":{"name":"2022 IEEE 29th International Conference on High Performance Computing, Data, and Analytics (HiPC)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE 29th International Conference on High Performance Computing, Data, and Analytics (HiPC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/HiPC56025.2022.00031","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
This paper presents a comprehensive analysis of the performance of Fortran, C, Python, and Julia based GPU accelerated meshfree solvers for compressible flows. The programming model CUDA is used to develop the GPU codes. The meshfree solver is based on the least squares kinetic upwind method with entropy variables (q-LSKUM). To measure the performance of baseline codes, benchmark calculations are performed. The codes are then profiled to investigate the differences in their performance. Analysing various performance metrics for the computationally expensive flux residual kernel helped identify various bottlenecks in the codes. To resolve the bottlenecks, several optimisation techniques are employed. Post optimisation, the performance metrics have improved significantly, with the C GPU code exhibiting the best performance.
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
