processor Allocation in K-ary n-cube Multiprocessors

碩士 === 淡江大學 === 電機工程研究所 === 85 ===   This paper presents an extension of a previous processor allocation scheme, the Tree Collapsing (TC) strategy, to recognize m-ary r-cubes in a K-ary n-cube system, where m□K and r□n. We assume that tasks arrive dynamically and request processors connected in spe...

Full description

Bibliographic Details
Main Authors: Wu, Chih-Ming, 吳志銘
Other Authors: Chuang, Po-Jen
Format: Others
Language:zh-TW
Published: 1997
Online Access:http://ndltd.ncl.edu.tw/handle/89141043138551811077
Description
Summary:碩士 === 淡江大學 === 電機工程研究所 === 85 ===   This paper presents an extension of a previous processor allocation scheme, the Tree Collapsing (TC) strategy, to recognize m-ary r-cubes in a K-ary n-cube system, where m□K and r□n. We assume that tasks arrive dynamically and request processors connected in specific topologies, even with bases different from that of the system, which makes processor allocation more difficult than in the hypercube system. First, we introduce the TC strategy on the hypercubes and how it can be applied to the K-ary n-cube system. According to the TC algorithm, we can achieve the search space orderly for some incoming task requesting a particular topology. Then we explain that in recognizing subsystems, this extended strategy has full recognition ability, which is a basic factor for evaluating any processor allocation strategies.   Extensive simulation runs are carried out to collect experimental performance results for different allocation strategies, such as the Sniffing strategy, the Free-list strategy on K-ary n-cube and the K-ary Partner strategy, for comparison and discussion. It is shown from analytical and experimental results that this extended TC strategy compares favorably in most of the situations with other allocation schemes.