TI A Parallel Storage Scheme and its Implementation 

AV ftp thorhild.cs.ualberta.ca:pubTR91-08.ps.Z 

AV mail britta@cs.ualberta.ca  

OR ALBRT 

LT TR 91-08 

AU Zhiyong Liu 

AU Jia-Huai You 

AU Xiaobo Li 

MN April 

YR 1991 

AB The efficiency of a parallel processing system depends largely
on the parallel access  and alignment of data distributed over
the memory banks in the system.  We present a parallel storage
scheme to distribute the elements of an N x N matrix over N
memory banks, where N is any complete (odd or even) power of two,
such that any rows, columns, forward and backward diagonals, and
square or rectangular blocks can be accessed simultaneously
without memory conflict. We also present a simple scheme for
address generation (memory module number calculation and local
address determination) and network implementation for data
alignment requirements in the storage scheme.  Schemes for
processing smaller matrices efficiently on larger scale systems
are also developed.