2009 IEEE International Symposium on Parallel &Amp; Distributed Processing 2009
DOI: 10.1109/ipdps.2009.5161003
|View full text |Cite
|
Sign up to set email alerts
|

NewMadeleine: An efficient support for high-performance networks in MPICH2

Abstract: This paper describes how the NewMadeleine communication library has been integrated within the MPICH2 MPI implementation and the benefits brought. NewMadeleine is integrated as a Nemesis network module but the upper layers and in particular the CH3 layer has been modified. By doing so, we allow NewMadeleine to fully deliver its performance to an MPI application. NewMadeleine features sophisticated strategies for sending messages and natively supports multirail network configurations, even heterogeneous ones. I… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
5
0

Year Published

2010
2010
2021
2021

Publication Types

Select...
4
2

Relationship

1
5

Authors

Journals

citations
Cited by 6 publications
(5 citation statements)
references
References 9 publications
0
5
0
Order By: Relevance
“…We used MPICH2-nmad [14] as an MPI interface over NEW-MADELEINE, and compared against latest stable release OPENMPI 1.4.5. We evaluate the raw overhead of checksums computation as well as their impact on NAS Parallel Benchmarks.…”
Section: Discussionmentioning
confidence: 99%
“…We used MPICH2-nmad [14] as an MPI interface over NEW-MADELEINE, and compared against latest stable release OPENMPI 1.4.5. We evaluate the raw overhead of checksums computation as well as their impact on NAS Parallel Benchmarks.…”
Section: Discussionmentioning
confidence: 99%
“…Different approaches have been proposed to mitigate this burden, including Task-Aware MPI (TAMPI) [13] and a tight integration of fiber and tasking libraries with MPI implementations [14][15][16]. In a previous paper, we have argued that such an integration, while seemingly attractive due to its ease-of-use, fosters the development of a non-portable application ecosystem [17].…”
Section: Background and Motivationmentioning
confidence: 99%
“…Different approaches have been proposed to mitigate this burden, including Task-Aware MPI (TAMPI) [46] and a tight integration of user-level thread, fiber, and tasking libraries (cf. Section 2) with MPI implementations [32,34,45]. In this paper, we argue that the latter is the least preferable solution and in fact should be seen critical by the MPI community.…”
mentioning
confidence: 90%