3rd Euromicro Workshop on Parallel and Distributed Processing
A Shared Memory model based on the Dataflow paradigm
San Remo, Italy
January 25-January 27
ISBN: 0-8186-7031-2
In this paper a Shared Memory model designed to be efficiently implemented on a collection of autonomous processing elements is presented. The model is inspired by the Dataflow paradigm. The main features of the model are: the single assignment principle, multi-dimensional addressing scheme, and associative data access primitives. The single assignment principle limits the number of updates of a uniquely identified and atomic data object to one. In a multidimensional addressing the atomic data objects - pages are located in an n-dimensional address space. An associative data read primitive will succeed only when a specified set of memory pages becomes available. The model can be implemented on both tightly coupled shared memory MIMD machines, and distributed systems of heterogenous machines connected by a computer network.
Index Terms:
shared memory systems; data flow computing; distributed processing; computer networks; shared memory model; dataflow paradigm; autonomous processing elements; single assignment principle; multi-dimensional addressing; associative data access primitives; associative data read; tightly coupled; heterogenous machines; computer network
Citation:
G.J. Blinowski, "A Shared Memory model based on the Dataflow paradigm," pdp, pp.68, 3rd Euromicro Workshop on Parallel and Distributed Processing, 1995