Proceedings of the 26th European MPI Users' Group Meeting最新文献_第2页

Persistent coarrays: integrating MPI storage windows in coarray fortran 持久队列:在队列fortran中集成MPI存储窗口

Proceedings of the 26th European MPI Users' Group Meeting Pub Date : 2019-09-11 DOI: 10.1145/3343211.3343214

Sergio Rivas-Gomez, A. Fanfarillo, Sai B. Narasimhamurthy, S. Markidis

引用次数: 4

Efficient notifications for MPI one-sided applications MPI单侧应用程序的有效通知

Proceedings of the 26th European MPI Users' Group Meeting Pub Date : 2019-09-11 DOI: 10.1145/3343211.3343216

Marc Sergent, C. Aitkaci, Pierre Lemarinier, Guillaume Papauré

{"title":"Efficient notifications for MPI one-sided applications","authors":"Marc Sergent, C. Aitkaci, Pierre Lemarinier, Guillaume Papauré","doi":"10.1145/3343211.3343216","DOIUrl":"https://doi.org/10.1145/3343211.3343216","url":null,"abstract":"MPI One-sided communications have the potential to increase applications performance by reducing the noise on remote processors. They consist in Remote Memory Accesses roughly orchestrated in two types of operations: memory synchronizations and actual RMAs. Applications such as PGAS ones benefit from RMAs operations, yet they require to get notified when such a communication has completed in order to process the data. Using heavy provided synchronization mechanisms for each communication impact performance drastically. One of the solutions is for applications to implement a notification mechanism through a supplementary communication. To alleviate the need for developers to manage this notification mechanism, we propose an extension to the MPI one-sided API that provides RMAs with accompanying notification. We present a basic implementation of this extension that relies solely on existing MPI one-sided communications and provide all the functionality, that can be ported to any MPI libraries for their supported networks. We also propose an implementation targeting performance in Open MPI over InfiniBand's network, and present its performance evaluation.","PeriodicalId":314904,"journal":{"name":"Proceedings of the 26th European MPI Users' Group Meeting","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129975684","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 2

An MPI interface for application and hardware aware cartesian topology optimization 一个MPI接口，用于应用程序和硬件感知笛卡尔拓扑优化

Proceedings of the 26th European MPI Users' Group Meeting Pub Date : 2019-09-11 DOI: 10.1145/3343211.3343217

Christoph Niethammer, R. Rabenseifner

{"title":"An MPI interface for application and hardware aware cartesian topology optimization","authors":"Christoph Niethammer, R. Rabenseifner","doi":"10.1145/3343211.3343217","DOIUrl":"https://doi.org/10.1145/3343211.3343217","url":null,"abstract":"Many scientific applications perform computations on a Cartesian grid. The common approach for the parallelization of these applications with MPI is domain decomposition. To help developers with the mapping of MPI processes to subdomains, the MPI standard provides the concept of process topologies. However, the current interface causes problems and requires too much care in its usage: MPI_Dims_create does not take into account the application topology and most implementations of MPI_Cart_create do not consider the underlying network topology and node architecture. To overcome these shortcomings, we defined a new interface that includes application-aware weights to address the communication needs of grid-based applications. The new interface provides a hardware-aware factorization of the processes together with an optimized process mapping onto the underlying hardware resources. The paper describes the underlying implementation, which uses a new multi-level factorization and decomposition approach minimizing slow inter-node communication. Benchmark results show the significant performance gains on multi node NUMA systems.","PeriodicalId":314904,"journal":{"name":"Proceedings of the 26th European MPI Users' Group Meeting","volume":"96 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126961204","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 4

Implementing efficient message logging protocols as MPI application extensions 作为MPI应用程序扩展实现高效的消息日志记录协议

Proceedings of the 26th European MPI Users' Group Meeting Pub Date : 2019-05-08 DOI: 10.1145/3343211.3343219

K. Dichev, Dimitrios S. Nikolopoulos

{"title":"Implementing efficient message logging protocols as MPI application extensions","authors":"K. Dichev, Dimitrios S. Nikolopoulos","doi":"10.1145/3343211.3343219","DOIUrl":"https://doi.org/10.1145/3343211.3343219","url":null,"abstract":"Message logging protocols are enablers of local rollback, a more efficient alternative to global rollback, for fault tolerant MPI applications. Until now, message logging MPI implementations have incurred the overheads of a redesign and redeployment of an MPI library, as well as continued performance penalties across various kernels. Successful research efforts for message logging implementations do exist, but not a single one of them can be easily deployed today by more than a few experts. In contrast, in this work we build efficient message logging capabilities on top of an MPI library with no message logging capabilities; we do so for two different send-deterministic HPC kernels, one with a global exchange pattern (CG), and one with a neighbour exchange pattern (LULESH). While our library of choice ULFM detects failure and recovers MPI communicators, we build on that to then restore the intra- and inter-process data consistency of both applications. This task follows a similar pattern across these kernels, and we present our methodology in a generic way. In the end, our extensions provide message logging capabilities for each kernel, without the need for an actual message logging runtime underneath. On the performance side, we eliminate event logging for these kernels, and design a flexible user-defined hybrid between global and local rollback. Our extensions span a few hundred lines of code for each kernel, are open-sourced, and enable local and global rollback after process failure.","PeriodicalId":314904,"journal":{"name":"Proceedings of the 26th European MPI Users' Group Meeting","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130851566","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 2

Proceedings of the 26th European MPI Users' Group Meeting 第26届欧洲MPI用户组会议记录

Proceedings of the 26th European MPI Users' Group Meeting Pub Date : 1900-01-01 DOI: 10.1145/3343211

Antonio J. Peña, P. Balaji, W. Gropp, R. Thakur

引用次数: 0