Automatic Granularity Control for Load Balancing of Concurrent Particle Simulations
نویسندگان
چکیده
This paper demonstrates the use of automatic gran-ularity control as part of dynamic load balancing for irregular, particle-based simulations. Performance optimization techniques are considered in the context of a concurrent Direct Simulation Monte Carlo method used to study the rareeed gas ow inside three-dimensional plasma reactors. Several computational techniques are used to reduce the overall time to deliver realistic three-dimensional results. The eeectiveness of dynamic load balancing and granularity control are presented for large-scale simulations on distributed-memory multi-computers.
منابع مشابه
An Improved Algorithm for Distributed Robotic Control System based on Granularity Division and Entropy Weight Evaluation
Abstract To settle the module granularity division problem of robotic distributed control and the problem that the system function can not be extended caused by Name server losing control, an improved method integrating granularity division and entropy weight evaluation is proposed. Firstly, the system is divided into several minimum particle module sample sets and these sets are analyzed from ...
متن کاملPySPH: A Python Framework for Smoothed Particle Hydrodynamics
[PySPH] is a Python-based open source parallel framework for Smoothed Particle Hydrodynamics (SPH) simulations. It is distributed under a BSD license. The performance critical parts are implemented in [Cython]. The framework provides a load balanced, parallel execution of solvers. It is designed to be easy to extend. In this paper we describe the architecture of PySPH and how it can be used. At...
متن کاملDistributed Load Balancing of District Heating Systems - A Small-Scale Experiment
We present results from an experiment where the effects of automatic flow control at a single substation is compared to automatic cooperative concurrent flow control at multiple substations. The latter approach is made possible by equipping individual substations with some computing power and integrating them into a communications network. Software agents, whose purpose is to cooperate with oth...
متن کاملImplementation of a 2D Electrostatic Particle in Cell algorithm in Uni ed Parallel C with dynamic load-balancing
Since space plasmas are colisionless, reliable space weather simulations will require to take into account a full kinetic description of their dynamics. The Particle in Cell (PIC) method is a particle-mesh technique that e ciently allows such description. Still, realistic plasma simulations require huge data sets and are computationally expensive. High scalability must thus be achieved in order...
متن کاملVIPP: Visual Interactive Parallel Performance Tool
VIPP is an instructional tool for exploring performance tradeoos that arise when load balancing various particle methods on distributed memory MIMD computers. VIPP enables its user to interactively visualize factors aaecting performance including: granularity, geometry, and the relative costs of communication and computation. VIPP employs trace driven, application-level simulations in lieu of l...
متن کامل