Interrupt and Cancellation as Synchronization Methods
Parallel programs require an interprocess synchronization. There are many synchronization paradigms, but almost all of them explicitly or implicitly use waiting. We propose another approach. Instead of waiting for a condition, a process can interrupt / abandon the current activity when the condition is met. We propose to use conditions reflecting the global application state and we give a proposal how to construct such conditions. We discuss questions concerning semantics of the proposed system - how and when the conditions are evaluated and in what situations an interrupt can occur. The synchronization system allows simple and compact parallel algorithm implementations. As a practical example a parallel branch and bound algorithm with dynamic load balancing is shown. The usage of interrupts and cancellations enabled us to avoid superfluous computations, which often decrease the efficiency in parallel branch and bound programs.
KeywordsParallel Application Synchronization Condition Synchronization Method Parallel Branch Dynamic Load Balance
Unable to display preview. Download preview PDF.
- 2.Attila Gursoy, Laxmikant V. Kale ”Dagger: Combining Benefits of Synchronous and Asynchronous Communication Styles”, in Proceedings of the 8th International Symposium on Parallel Processing IPPS94, Cancun, Mexico, editor Howard Jay Siegel, IEEE Computer Society Press ISBN 0-8186-3442-1, pp 590–596Google Scholar
- 3.M. L. Scott and M.M. Michael, The Topological Barrier: A Synchronization Abstraction for Regurarly-Structered Parallel Applications, Rochester University Technical Report TR605, 1996Google Scholar
- 4.M. Raynal, J-M Helart ”Synchronization and control of distributed systems and programs” ISBN 0 471 92453 9, John Wiley and Sons Ltd 1990Google Scholar
- 5.Stanislaw Walukiewicz, ”Integer Programming”, PWN and Kluwer Academic Publishers, 1991, ISBN 83-01-09512-1Google Scholar
- 6.R. Horst and H. Tuy ”Global Optimization. Deterministic Appoaches”. Springer Verlag, third edition, 1996Google Scholar
- 7.Yuji Shinano, Kenicji Harada nad Ryuichi Hirabayashi, Control Schemes in a Generalized Utility for Parallel Branch-and-Bound Algorithms, Proceedings of the 11th International Parallel Processing Symposium (IPPS’ 97) IEEE 1997Google Scholar
- 9.H. Trienekens, ”Parallel branch and bound and anomalies” Technical Report EURFEW-CS-89-01, Erasmus University Rotterdam, 1989Google Scholar