Distinguishing sharing types to minimize communication in software distributed shared memory systems

Using thread migration to redistribute threads to processors is a common sc heme for minimizing communication needed to maintain data consistency in so ftware distributed shared memory (DSM) systems. In order to minimize data-c onsistency communication, the number of shared pages is used to identify th e pair of threads that will cause the most communication. This pair of thre ads is then co-located on the same node. Thread pairs sharing a given page can be classified into thee types, i.e., read/read (r/r), read/write (r/w) and write/write (w/w). Based on memory-consistency protocol, these three ty pes of sharing generate distinct amounts of data-consistency communication. Ignoring this factor will mispredict the amount of communication caused by cross-node sharing and generate wrong decisions in thread migration. This paper presents a new policy called distinguishing of types sharing (DOTS) f or DSM systems. The basic concept of this policy is to classify sharing amo ng threads as r/r, r/w or w/w, each with a different weight, and then evalu ate communication cost in terms of these weights. Experiments show that con sidering sharing types is necessary for minimization of data-consistency co mmunication in DSM. Using DOTS for thread mapping produces more communicati on reduction than considering only the number of shared pages. (C) 2000 Els evier Science Inc. All rights reserved.