Welcome to the new version of CaltechAUTHORS. Login is currently restricted to library staff. If you notice any issues, please email coda@library.caltech.edu
Published January 1, 1982 | public
Report Open

The Extension of Object-Oriented Languages to a Homogenous, Concurrent Architecture

Abstract

A homogeneous machine architecture, consisting of a regular interconnection of many identical elements, exploits the economic benefits of VLSI technology, A concurrent programming model is presented that is related to object oriented languages such as Simula and Smalltalk. Techniques are developed which permit the execution of general purpose object oriented programs on a homogeneous machine. Both the hardware architecture and the supporting software algorithms are demonstrated to scale their performance with the size of the system. The program objects communicate by passing messages. Objects may move about in the system and may have an arbitrary pointer topology, A distributed, on-the-fly garbage collection algorithm is presented which operates by message passing. Simulation of the algorithm demonstrates its ability to collect obsolete objects over the entire machine with acceptable overhead costs. Algorithms for maintaining the locality of object references and for implementing a virtual object capability are also presented. To insure the absence of hardware bottlenecks, a number of interconnection strategies are discussed and simulated for use in a homogeneous machine. Of those considered, the Boolean N-cube connection is demonstrated to provide the necessary characteristics. The object oriented machine will provide increased performance as its size is increased. It can execute a general purpose, concurrent, object oriented language where the size of the machine and its interconnection topology are transparent to the programmer.

Files

5014_TR_82.pdf
Files (27.1 MB)
Name Size Download all
md5:8a7a8673a604c28896c424f4db258c3d
12.7 MB Preview Download
md5:e58d57361b70fde450ba33dbe7e95fa2
14.4 MB Download

Additional details

Created:
August 19, 2023
Modified:
December 22, 2023