Distributed memory multiprocessing offers a cost-effective and scalable for a large class of scientific and numeric applications. Unfortunately, the performance of current distributed memory programming environments suffers because the frequency of communication between processors can exceed that required ensure a correctly functioning program.
MW is a shard memory parallel programming system which addresses the problem of excessive communication in a distributed memory multiprocessor. MW programs are written using a conventional MIMD-style programming model executing within a single globally shared local memory. Local memories on each processor cache recently used a data to counter the effects of network latency.
MW is based on a new model of memory consistency called entry consistency. Entry consistency exploits the relationship between synchronization objects and the data they protect. Updates to shared data are communicated between processors only when not doing so would result in an inconsistent state given a properly synchronizing program.
MW implements its consistency protocols in software and has no dependencies on any specific hardware characteristic other than the ability to send messages between processors. A strictly software solution is attractive because it allows us to exploit application specific information at the lowest levels of the system, and because it ensures portability across a wide range of multiprocessor architectures.
PLEASE NOTE: The authors suggest that people interested in this topic should first obtain the paper: The Midway Distributed Shared Memory System. CMU Technical Report CMU-CS 93-119. The abstract is listed in this bibliography.