Title
An Abstraction to Support Design of Deadlock-free Routing Algorithms for Large and Hierarchical NoCs
Abstract
New techniques and tools will be required to handle the complexity of on-chip networks required for building multi-core platforms with hundreds of cores. The concepts of hierarchy and abstraction, which have been useful in the design and analysis of electronic systems, will also be useful in the network design area. In this paper we propose an abstraction for networks to reduce the complexity of design and analysis of deadlock-free routing for large and hierarchical networks. Raising the level of abstraction necessarily leads to loss of some information. This loss of information in turn often results in loss of performance of the design as compared to when the network was considered at a lower level of abstraction. We show that our proposed abstraction leads to very small loss of performance. In particular, we show that average increase in communication distance (hops) is low or comparable to non-abstract design. Our simulation-based evaluation indicates that abstract routing design has no negative impact on message latency. On the contrary, abstract routing shows a slight advantage, which increases if more network communication is local and confined within subnets.
Year
DOI
Venue
2011
10.1109/CIT.2011.32
Computer and Information Technology
Keywords
Field
DocType
abstract routing,abstract routing design,network communication,small loss,proposed abstraction,deadlock-free routing algorithms,support design,on-chip network,deadlock-free routing,communication distance,network design area,hierarchical nocs,hierarchical network,network analysis,chip,topology,algorithm design,logic design,network design,network topology,network on chip,network routing,routing,algorithm design and analysis,concrete
Logic synthesis,Abstraction,Algorithm design,Network planning and design,Computer science,Hierarchical routing,Network on a chip,Computer network,Network topology,Real-time computing,Abstraction inversion,Distributed computing
Conference
ISBN
Citations 
PageRank 
978-0-7695-4388-8
0
0.34
References 
Authors
8
2
Name
Order
Citations
PageRank
Rickard Holsmark124913.10
Shashi Kumar228216.58