Managing the recursive generation of tasks in a transputer network
Computational constraints in image processing and image analysis, take advantage of a parallel implementation performed on a distributed memory parallel architecture; the following one is based on tranputers. This paper proposes a parallel implementation based on the “processors farm” method, of the splitting step of Kd-tree algorithms (split & merge); it is computationally expensive, but involves a large number of independant tasks. The project sets out to implement it on a on-board real time system.
firstly, when a starvation point appears, a total starvation is created;
secondly, the master does not stop its “farming”, but manages the network temporal comportment and “farm” new sub-tasks to idle workers.
In the end, some experimentals results, performed on several networks composed from 2 till 12 transputers, are presented.
KeywordsParallel architectures Processors Farm Transputers Recursive algorithms
Unable to display preview. Download preview PDF.
- H. SAMET “The Quadtree and related hierarchical data structures” Computer surveys, Vol.16, no2, June 1984, p.187–260Google Scholar
- L.HOROWITZ, T.PAVLIDIS “Picture segmentation by a directed split-and-merge procedure”, 1974Google Scholar
- INMOS limited “The transputer data book”; Second edition INMOS, 1989Google Scholar
- P.M. DEW, HAN WANG “Data parallelism and the processor farm model for image processing and synthesis on a transputer array” 1988 Proc SPIE Int. Soc. Opt. Eng. Vol 977Google Scholar
- D.MAY, R.SHEPHERD, C.KEANE “Communicating process architecture: Transputer and OCCAM” INMOS limited, 1987Google Scholar
- P.VINCENT “Nouvelles architectures d'ordinateurs”; Edition test Chap 6Google Scholar
- S.NICOLLE, F.CHANTEMARGHUE, JP.DERUTIN, P.BONTON “Formalizing Architecture Based on the “Processors Farm” Method, For implementation of Kd-tree Algorithms” IASTED, June 18–21, 1990, Lugano, SWITZERLAND; p31–35Google Scholar
- M.GONDRAN, M.MINOUX “Graphes et algorithmes”; Edit EYROLLES, 1979Google Scholar
- D.A.NICOLE, E.K.LLYOD, J.S.WARD “Switching networks for transputer links”, 1987Google Scholar
- S.NICOLLE “Formalisation d'une architecture utilisant la méthode ferme de processeurs: parallélisation d'algorithmes de vision de type Kd-tree”. Internal report 1990Google Scholar
- R.D DOWSING “Introduction to concurrency using OCCAM” Van Nostrand Reinhold (International) 1988Google Scholar
- R.C. HOLT “Some deadlock properties of computer system” ACM computer Surveys 1972, 4, 179–96Google Scholar
- D.TRYSTAN, F.VINCENT “Programmation avancée du transputer”, Lettre du transputer, 2, 1989.Google Scholar
- P.LEGRAND, S.NICOLLE, JP.DERUTIN “Parallélisation de la phase de division d'un algorithme de type division-fusion selon la méthode ferme de processeurs” Internal report June 90Google Scholar
- D.J.PRITCHARD “Mathematical models of distributed compilation” 7th OPPT, Sept 14–16, 1987, Ed Traian MUNTEANGoogle Scholar
- P.BOUTHEMY, J.SANTILLANA RIVERO “Region segmentation according to motion-based criteria” Rapport de recherche INRIA no629, February 1987.Google Scholar
- F.CHANTEMARGUE, S.NICOLLE, P.BONTON, JP.DERUTIN “Parallelization of the splitting step of a quadtree algorithm” IASTED, June 18–21, 1990, Lugano, SWITZERLAND; p183–188Google Scholar
- J.P. DERUTIN, B.BESSERER “Transvision: an heterogeneous machine for real time image processing”, Internal report 90.Google Scholar