Authors: Stavros Souravlas Manos Roumeliotis 
              Publish Date: 2015/09/10
              Volume: 71, Issue: 11, Pages: 4215-4234 
			  
              Abstract
              This paper addresses the wellknown problem of redistributing data arrays over a multiprocessor network The blockcyclic redistribution techniques found in the literature deal effectively with this problem based on the assumption that there is a direct link between all the processors of the network Most of these techniques aim at reducing the number of messages and the total redistribution cost However an application of these techniques on nonalltoall communication networks like tori shows that these techniques suffer long delays In this work we try to solve the general blockcyclic redistribution problem on nonalltoall networks by grouping the messages into welldefined classes and transferring them with the support of a wellspecified number of virtual channels
              
			  Keywords: