A Comparison of Sequential and Associative Computing of Priority Queues with Applications to Discrete Simulation.
Abstract
Over the past fifteen years, a number of authors have studied methods of improving data management in discrete simulation. These methods have dealt with the random access memory. Over the same period, a number of other authors have studied various forms of associative architecture. The following research combines both fields of study to focus on discrete simulation data management. The specific concern is priority queues and time flow mechanisms. The approach used is to specify three forms of associative memories and their associated algorithms and then compare them with a random access memory which uses contemporary algorithms. The following is a summary of the results. The direct algorithmic map-over of random access memory algorithms to associative architecture was unsatisfactory. The algorithmic procedures had to be reconstituted with parallel processing in mind. Further, the traditional time flow mechanisms could not be used and a new time flow mechanism, called fixed increment minimum value, was developed. With the advent of new algorithmic procedures, the various forms of the associative memories did show distinct processing potential based on parametric values established for the computer models. The advantage or disadvantage of the associative architectures with respect to the random access was most sensitive to the ratio of the memory speeds and the processing width or degree of parallelism in the associative cases. The research indicates that hybrid associative memories are superior to straight associative memories for discrete simulation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1977
- Accession Number
- ADA044213
Entities
People
- Barry M. Landson
Organizations
- Rome Laboratory