Reducing communication overhead under parallel list processing in multicore clusters

Author

Buenabad-Chávez, Jorge ; Castro-García, Miguel Alfonso ; Quiroz-Fabián, José Luis ; Hernández-Ventura, Edgar F. ; Román-Alonso, Graciela ; Yellin, Daniel M. ; Aguilar-Cornejo, Manuel

Author_Institution

Dept. de Comput., CINVESTAV-IPN, Mexico City, Mexico

fYear

2011

fDate

26-28 Oct. 2011

Firstpage

1

Lastpage

6

Abstract

The Data List Management Library (DLML) processes data lists in parallel, balancing the workload transparently to programmers. Its first design was targeted at clusters of uniprocessor nodes, and based on multiprocess parallelism and on message-passing communication. This paper presents a multithreaded design of DLML aimed at clusters of multicore nodes to better capitalise on intra-node parallelism. On applications tested, MultiCore DLML runs twice as fast as DLML when message-passing communication is not excessive. Good performance was achieved only after addressing issues relating to MPI communication overhead, cache locality and memory consumption.

Keywords

message passing; multi-threading; multiprocessing systems; MPI communication overhead; MultiCore DLML; cache locality; communication overhead reduction; data list management library; intranode parallelism; memory consumption; message-passing communication; multicore clusters; multiprocess parallelism; multithreaded design; parallel list processing; uniprocessor nodes; workload balancing; Instruction sets; Libraries; Message passing; Message systems; Multicore processing; Parallel processing; Switches; Message Passing; Multicore Clusters; Parallel Computing; Shared Memory;

fLanguage

English

Publisher

ieee

Conference_Titel

Electrical Engineering Computing Science and Automatic Control (CCE), 2011 8th International Conference on

Conference_Location

Merida City

Print_ISBN

978-1-4577-1011-7

Type

conf

DOI

10.1109/ICEEE.2011.6106562

Filename

6106562