Browse by author
Lookup NU author(s): Dr Ashur Rafiev, Mohammed Al-Hayanni, Dr Fei Xia, Professor Rishad Shafik, Emeritus Professor Alexander RomanovskyORCiD, Professor Alex Yakovlev
This is the authors' accepted manuscript of an article that has been published in its final definitive form by IEEE, 2018.
For re-use rights please refer to the publisher's terms and conditions.
Traditional speedup models, such as Amdahl’s law, Gustafson’s, and Sun and Ni’s, have helped the research communityand industry better understand system performance capabilities and application parallelizability. As they mostly target homogeneoushardware platforms or limited forms of processor heterogeneity, these models do not cover newly emerging multi-core heterogeneousarchitectures. This paper reports on novel speedup and energy consumption models based on a more general representation ofheterogeneity, referred to as the normal form heterogeneity, that supports a wide range of heterogeneous many-core architectures. Themodelling method aims to predict system power efficiency and performance ranges, and facilitates research and development at thehardware and system software levels. The models were validated through extensive experimentation on the off-the-shelf big.LITTLEheterogeneous platform and a dual-GPU laptop, with an average error of 1% for speedup and of less than 6.5% for power dissipation.A quantitative efficiency analysis targeting the system load balancer on the Odroid XU3 platform was used to demonstrate the practicaluse of the method.
Author(s): Rafiev A, Al-Hayanni MAN, Xia F, Shafik R, Romanovsky A, Yakovlev A
Publication type: Article
Publication status: Published
Journal: IEEE Transactions on Multi-Scale Computing Systems
Year: 2018
Volume: 4
Issue: 3
Pages: 436-449
Online publication date: 12/01/2018
Acceptance date: 24/12/2017
Date deposited: 25/12/2017
ISSN (electronic): 2332-7766
Publisher: IEEE
URL: https://doi.org/10.1109/TMSCS.2018.2791531
DOI: 10.1109/TMSCS.2018.2791531
Data Access Statement: http://dx.doi.org/10.17634/123238-4
Altmetrics provided by Altmetric
