ON INTERCONNECTION NETWORKS FOR PARALLEL COMPUTERS

被引：0

作者：

Li, Lei ^{[1
]}

Li, Qian ^{[2
]}

Yoshioka, Yoshio ^{[3
]}

机构：

[1] Hosei Univ, Fac Engn, Tokyo 1848584, Japan

[2] Cores Co Ltd, Sendai Branch, Aoba Ku, Sendai, Miyagi 9800811, Japan

[3] Hirosaki Univ, Dept Informat Sci, Hirosaki, Aomori 036, Japan

来源：

INFORMATION-AN INTERNATIONAL INTERDISCIPLINARY JOURNAL | 2008年 / 11卷 / 06期

关键词：

Network; complexity; cube; ring; pie;

D O I：

暂无

中图分类号：

T [工业技术];

学科分类号：

08 ;

摘要：

In this paper, we present new results involving interconnection networks for parallel computers based on fractals of rings and fractals of r-pies. Let d be the greatest distance in the network, and let l be the number of circuits connected to an arbitrary processor. It is well known for hypercube networks (i.e., the 2-ary it-cube) that both d and l equal log(2) N, where N is the number of processors. In this paper we present alternative network topologies with lower bounds on d and l. In particular, for the 5-ary n-cube, d and l are both strictly less than 0.86 log(2) N. For the 15-pie network, d and l are both strictly less than 0.77 log(2) N.

引用

页码：739 / 748

页数：10

共 12 条

[1]

Bertsekas Dimitri P., 1989, PARALLEL DISTRIBUTED

[2]

CHEN G, 1990, SORTING SELECTIONS

[3] EXPRESS CUBES - IMPROVING THE PERFORMANCE OF K-ARY N-CUBE INTERCONNECTION NETWORKS [J].