Related Experiment Videos
Algorithmic scalability in globally constrained conservative parallel discrete event simulations of asynchronous
A Kolakowska1, M A Novotny, G Korniss
1Department of Physics and Astronomy, and the MSU ERC, PO Box 5167, Mississippi State, Mississippi 39762-5167, USA. alicjak@bellsouth.net
Summary
A new Delta-window constraint improves parallel simulations for asynchronous systems. This modification ensures the measurement phase scales efficiently, optimizing processor utilization and system performance.
Area of Science:
- Computational physics
- Parallel computing
Background:
- Asynchronous systems require careful simulation to maintain causality.
- Previous parallel simulation methods showed scaling limitations in the measurement phase.
Purpose of the Study:
- To introduce and evaluate a moving Delta-window global constraint for parallel asynchronous simulations.
- To improve the scalability of the measurement phase in these simulations.
Main Methods:
- Implementing a moving Delta-window global constraint in parallel simulations.
- Conducting systematic studies varying system size and constraint parameters.
- Analyzing processor utilization and virtual time horizon dynamics.
Main Results:
- The Delta constraint effectively bounds the virtual time horizon width.
- The constraint controls the average progress rate of the simulation.
- Processor utilization is optimized by tuning the Delta window width.
Conclusions:
- The Delta-window constraint enhances the efficiency of parallel asynchronous simulations.
- This method offers a tunable parameter for optimizing performance in dynamic Monte Carlo studies.
- Applicable to modeling spatially extended systems with short-range interactions and asynchronous dynamics.