Home // ADAPTIVE 2013, The Fifth International Conference on Adaptive and Self-Adaptive Systems and Applications // View article

Self-discovery Algorithms for a Massively-Parallel Computer

Authors:
Kier J Dugan
Jeff S Reeve
Andrew D Brown

Keywords: SpiNNaker; self-discovering networks; parallel computer bootstrap procedures; self-configuration.

Abstract:
SpiNNaker is a biologically-inspired massively-parallel computer design that will contain over a million processors, distributed over more than 60,000 chips. The system bootstrap must discover how they are connected for the machine to enter a usable state. In this paper we describe a set of algorithms for discovering missing or malfunctioning inter-chip links, assigning unique identifiers to each chip, and building point-to-point network routing tables. All of the algorithms have been simulated, and will be implemented into SpiNNaker after further investigation. Our goal is to design an autonomic bootstrap stage that can operate on arbitrary machine geometries.

Pages: 36 to 39

Copyright: Copyright (c) IARIA, 2013

Publication date: May 27, 2013

Published in: conference

ISSN: 2308-4146

ISBN: 978-1-61208-274-5

Location: Valencia, Spain

Dates: from May 27, 2013 to June 1, 2013