A parallel algorithm for matrix multiplication of compressed Z order data structures

被引：0

作者：

Scott, K ^{[1
]}

机构：

[1] Univ Alaska Anchorage, Anchorage, AK 99508 USA

来源：

PROCEEDINGS OF THE ISCA 20TH INTERNATIONAL CONFERENCE ON COMPUTERS AND THEIR APPLICATIONS | 2005年

关键词：

matrix multiplication; Z order; P-trees;

D O I：

暂无

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Because matrix multiplication is an important and costly operation, it is a candidate for performance improvement through parallel and distributed computation. Using Z order in a new way it is possible to represent bit matrices in tree form and devise a corresponding algorithm. The advantages gained are: Matrix compression; Obviation of finding dot products for rows or columns of all 0's; The use of bit operations to do the multiplication; The ability to subdivide the problem at multiple levels.

引用

页码：453 / 458

页数：6