Name
Title | ||
|---|---|---|
Complete exchange algorithms in wormhole-routed torus networks : a divide-and-conquer strategy |
Abstract | ||
|---|---|---|
We present efficient all-to-all personalized communication algorithms for a 2D torus in wormhole-routed networks. Our complete exchange algorithm adopts divide-and-conquer approach to reduce the start-up latency significantly, which is a good metric for network performance in wormhole networks. The first algorithm recursively divides the whole network into 2×2 networks. After specially designated nodes called master nodes have collected messages, whose destinations are the rest of the basic cells, only master nodes perform complete exchange with reduced network size. Upon finishing this complete exchange among master nodes, these nodes distribute messages to the rest of the master nodes, which results in the desired complete exchange. The second algorithm, which is more practical than the first one, divides the network only once. After we present these new algorithms, we analyze time complexities and compare several algorithms. We show that our practical algorithm is efficient by a factor of 2 in the required start-up time which means that our algorithm is suitable for wormhole-routed networks |
Year | DOI | Venue |
|---|---|---|
1999 | 10.1109/ISPAN.1999.778955 | ISPAN |
Keywords | Field | DocType |
divide-and-conquer strategy,complete exchange algorithm,multiprocessor interconnection networks,practical algorithm,network performance,whole network,all-to-all personalized communication algorithms,2d torus,algorithm recursively,time complexities,computational complexity,reduced network size,complete exchange algorithms,performance evaluation,complete exchange,wormhole-routed network,new algorithm,wormhole-routed torus networks,divide and conquer methods,master node,unicast,intelligent networks,artificial intelligence,time complexity,computer science,divide and conquer,routing,message passing,topology,computer worms | Network size,Latency (engineering),Computer science,Algorithm,Computer network,Torus,Divide and conquer algorithms,Wormhole,Recursion,Computational complexity theory,Distributed computing,Network performance | Conference |
ISSN | ISBN | Citations |
1087-4089 | 0-7695-0231-8 | 2 |
PageRank | References | Authors |
0.39 | 0 | 3 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Si-Gwan Kim | 1 | 3 | 1.10 |
| Seung Ryoul Maeng | 2 | 85 | 19.68 |
| Jung Wan Cho | 3 | 64 | 79.18 |