By Topic

Minimizing network contention for mapping tasks onto massively parallel computers

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$33 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
R. Perego ; CNR, Pisa, Italy ; G. De Petris

This paper presents a new framework aimed at compile-time determination of satisfactory sub-optimal solutions to the mapping problem onto modern massively parallel computing systems. The approach incorporates realistic assumptions on the models both for parallel programs and target architectures. It is refined for the K-ary n-cube family of processor networks that use a deterministic routing algorithm and the wormhole flow control strategy. The proposed mapping heuristic is guided by an evaluation function that approximates the total completion time of a given assignment by taking into account communication delays caused by network contention which are the major contributors to message latencies when network traffic is heavy or unevenly distributed. Results achieved on several program-derived graphs with up to 784 tasks prove the effectiveness of this approach

Published in:

Parallel and Distributed Processing, 1995. Proceedings. Euromicro Workshop on

Date of Conference:

25-27 Jan 1995