| Summary: | 碩士 === 國立中央大學 === 資訊及電子工程研究所 === 81 === Recently a new interconnection network topology called star
graph seems to be a very attractive alternative for the
hypercube. It has been shown that the star graphs can
accommodate more processors with less interconnection and less
communication delay than the hypercubes. Many parallel
algorithms need an efficient broadcasting algorithm as a
fundamental step. In this thesis, we propose an optimal and
non-redundant distributed broadcasting algorithm in star graphs
based on the single-port model. Our algorithm runs in O(nlogn)
time steps where n is the dimension of star graphs and
guarantees that all nodes receive the message exactly once. By
recursively partitioning a given star graph into smaller
disjoint substar graphs, a message can be broadcast to all
nodes via exploiting the hierarchical property of the star
graphs. Broadcasting a stream of m messages in a pipeline
fashion, therefore, takes O(mlogn + nlogn) time steps because
of the non-redundant property of our broadcasting algorithm.
|
|---|
