Parallel Monte Carlo Simulation Using Desktop Computers

Autor: John Henry J. Scott, Robert L. Myklebust, Dale E. Newbury
Rok vydání: 2000
Předmět:
Zdroj: Microscopy Today. 8:34-35
ISSN: 2150-3583
1551-9295
DOI: 10.1017/s1551929500057485
Popis: Monte Carlo simulation of electron scattering in solids has proven valuable to electron microscopists for many years. The electron trajectories, x-ray generation volumes, and scattered electron signals produced by these simulations are used in quantitative x-ray microanalysis, image interpretation, experimental design, and hypothesis testing. Unfortunately, these simulations are often computationally expensive, especially when used to simulate an image or survey a multidimensional region of parameter space.Here we present techniques for performing Monte Carlo simulations in parallel on a cluster of existing desktop computers. The simulation of multiple, independent electron trajectories in a sample and the collateral calculation of detected x-ray and electron signals falls into a class of computational problems termed “embarrassingly parallel”, since no information needs to be exchanged between parallel threads of execution during the calculation. Such problems are ideally suited to parallel multicomputers, where a manager process distributes the computational burden over a large number of nodes.
Databáze: OpenAIRE