Proceedings 20th International Conference Parallel Processing 1991, 3. sējums

applications approach architecture array assigned assume binary block broadcast called column combined communication complexity components computation configuration connected consider construct contains corresponding defined denote dependence described determine distributed edge efficient elements example execution factor Figure function given graph hypercube IEEE implementation increase initial input instances integers interval iteration keys linear machine mapping matrix memory merging mesh method multiprocessor node Note O(log obtained operation optimal overhead parallel algorithm partition path performance phase points port presented problem procedure processing processors proposed References represented respectively routing runs scheme sequence shared shown shows simulation solution solved sort speedup step structure Table task technique tion tokens transform tree update vector vertices