RT Journal Article
JF Parallel Algorithms / Architecture Synthesis, AIZU International Symposium on
TI Communication algorithms on alternating group graphs
A1 Jyh-Jong Tsay,
A1 Chih-Ming Lai,
K1 multiprocessor interconnection networks; communication algorithms; alternating group graphs; all-to-all broadcast; all-port routing; store-and-forward routing; running time; lower bound; Hamiltonian paths; single-node scattering; time bound; multiprocessor interconnection networks; parallel algorithms; distributed memory systems
AB We study the problem of performing all-to-all broadcast on an n-alternating group graph AG/sub n/ with all-port and store-and-forward routing. The running time is [(n/sup 1/-2)/(4(n-2))+1] that is one step more than the trivial lower bound [(n/sup 2/-2)/(4(n-2))]. Our algorithm is based on some new properties of Hamiltonian paths of AG/sub n/ that are identified in this paper and are of independent interest. Similar properties have been used to design an algorithm for single-node scattering that achieves the same time bound.
PB IEEE Computer Society, [URL:http://www.computer.org]