Open AccessEvaluating parallel optimization on transputers
Author(s) -
AG Chalmers,
John Hearne,
Chris Scogings
Publication year - 2003
Publication title -
orion/orion
Language(s) - English
Resource type - Journals
eISSN - 2224-0004
pISSN - 0259-191X
DOI - 10.5784/5-2-482
Subject(s) - computer science , implementation , process (computing) , range (aeronautics) , parallel processing , parallel computing , cost efficiency , speedup , parallel algorithm , algorithm , materials science , composite material , programming language , operating system
The faster processing power of modern computers and the development of efficient algorithms have made it possible for operations researchers to tackle a much wider range of problems than ever before. Further improvements in processing speed can be achieved utilising relatively inexpensive transputers to process components of an algorithm in parallel. The Davidon-Fletcher-Powell method is one of the most successful and widely used optimisation algorithms for unconstrained problems. This paper examines the algorithm and identifies the components that can be processed in parallel. The results of some experiments with these components are presented which indicates under what conditions parallel processing with an inexpensive configuration is likely to be faster than the traditional sequential implementations. The performance of the whole algorithm with its parallel components is then compared with the original sequential algorithm. The implementation serves to illustrate the practicalities of speeding up typical OR algorithms in terms of difficulty, effort and cost. The results give an indication of the savings in time a given parallel implementation can be expected to yield