Shared Memory Based MPI Broadcast Algorithms for NUMA Systems

Autor:	Elizaveta Tokmasheva, Mikhail G. Kurnosov
Rok vydání:	2020
Předmět:	Broadcast algorithm Speedup Shared memory Xeon Computer science Server Scalability Limit (mathematics) Parallel computing Software_PROGRAMMINGTECHNIQUES Queue
Zdroj:	Communications in Computer and Information Science ISBN: 9783030646158 RuSCDays
Popis:	MPI_Bcast collective communication operation is used by many scientific applications and tend to limit overall parallel application scalability. This paper investigates the design and optimization of broadcast operation for NUMA nodes with GNU/Linux. We describe algorithms for MPI_Bcast that take advantage of NUMA-specific placement of queues in a shared memory for message transferring. On a Xeon Nehalem and Xeon Broadwell servers, our implementation achieves on average 20–60% speedup over algorithms of Open MPI coll/sm and MVAPICH.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::a99e2a55c8cb6d25d6196b2a4a5e6167 https://doi.org/10.1007/978-3-030-64616-5_41 Zobrazit plný text záznamu