Abstract
Consider a parallel computer on which it is considerably cheaper to perform communications within close neighborhoods than globally. In such a context, it is not necessarily a good idea to simulate optimal PRAM algorithms. In this paper consider the list-ranking problem and present a new divide-andconquer algorithm. For lists of length N, it performs O(N · log N) work), but its communication is mainly local: there are only 6 · 2i communication operations within subnetworks of size P/2i. © Springer-Verlag 2003.
Cite
CITATION STYLE
Sibeyn, J. F. (2004). Minimizing global communication in parallel list ranking. Lecture Notes in Computer Science (Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2790, 894–902. https://doi.org/10.1007/978-3-540-45209-6_123
Register to see more suggestions
Mendeley helps you to discover research relevant for your work.
