Paper Profile


Asynchronous Transport Algorithm For Domain Decomposed In JMCT Monte Carlo code


Documents: Full paper available in Conference Proceedings.
 
Date: Monday October 19
Time:16:40 - 17:05
 
Session:1B3: Monte Carlo I
 
Authors: Gang Li (Institute of Applied Physics and Computational Mathematics, China)
BaoYin Zhang (Institute of Applied Physics and Computational Mathematics, China)
Li Deng (Institute of Applied Physics and Computational Mathematics, China)
Zeyao Mo (Institute of Applied Physics and Computational Mathematics, China)
Yan Ma (Institute of Applied Physics and Computational Mathematics, China)
Rui Li (CAEP Software Center for High Performance Numerical Simulation, China)
Danhua Shangguan (Institute of Applied Physics and Computational Mathematics, China)
Yuanguang Fu (CAEP Software Center for High Performance Numerical Simulation, China)
Abstract:

High-fidelity, large-scale reactor analyses with Monte Carlo can lead to memory overload for a single core processor when it faces the detailed and accurate model of the full-core reactor. Domain decomposed calculation is one of the remedies to solve this problem. In domain decomposition, particles that cross domain boundaries need to be exchanged between the processors. An efficient, robust asynchronous transport algorithm for domain decomposition is introduced in this paper. All point-to-point communications in the algorithm are asynchronous to allow maximal overlap between computational work and communication, and result in increased parallel efficiency. The exchanged particles records are used to exit the loop at the end of the cycle in a criticality calculation when all the particles have finished. Two full-core reactor models, Dayawan reactor and BEAVRS benchmark, are simulated to verify the asynchronous transport algorithm using the Monte Carlo particle transport code JMCT (J Monte Carlo Transport Code).

Back to Top