Title :
Interconnect system compression analysis for multi-core architectures
Author :
Liu, Jiangjiang ; Zhang, Jianyong ; Mahapatra, Nihar
Author_Institution :
Dept. of Comp. Sci., Lamar Univ., Beaumont, TX, USA
Abstract :
One of the major problems associated with integrating multiple cores on a single chip is the performance demand it places on the interconnect system because of the combined traffic generated by multiple simultaneously executing threads. In this paper, we analyze interconnect compression potential in the two primary types of information - instructions and data - and important interconnect components in multi-core systems. In our study, lower compression ratio means better compression. We show great compression potential with 0.47-0.67 zero-information compression ratio, 0.28-0.29 zeroth-order compression ratio, and 0.11-0.14 first-order compression ratio obtained with different levels of compression specialization.
Keywords :
computer architecture; multiprocessing systems; multiprocessor interconnection networks; interconnect system compression analysis; multicore architecture; processor-memory traffic; Analytical models; Benchmark testing; Computational modeling; Entropy; Integrated circuit interconnections; Markov processes; Multicore processing;
Conference_Titel :
SOC Conference (SOCC), 2010 IEEE International
Conference_Location :
Las Vegas, NV
Print_ISBN :
978-1-4244-6682-5
DOI :
10.1109/SOCC.2010.5784654