Parallel Monte Carlo transport modeling in the context of a time-dependent, three-dimensional multi-physics code
- Lawrence Livermore National lab., CA (United States)
The fine-scale, multi-space resolution that is envisioned for accurate simulations of complex weapons systems in three spatial dimensions implies flop-rate and memory-storage requirements that will only be obtained in the near future through the use of parallel computational techniques. Since the Monte Carlo transport models in these simulations usually stress both of these computational resources, they are prime candidates for parallelization. The MONACO Monte Carlo transport package, which is currently under development at LLNL, will utilize two types of parallelism within the context of a multi-physics design code: decomposition of the spatial domain across processors (spatial parallelism) and distribution of particles in a given spatial subdomain across additional processors (particle parallelism). This implementation of the package will utilize explicit data communication between domains (message passing). Such a parallel implementation of a Monte Carlo transport model will result in non-deterministic communication patterns. The communication of particles between subdomains during a Monte Carlo time step may require a significant level of effort to achieve a high parallel efficiency.
- Research Organization:
- Sandia National Lab. (SNL-NM), Albuquerque, NM (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 332746
- Report Number(s):
- SAND-98-1591; CONF-9709141-PROC.; ON: DE99000778; TRN: IM9916%%48
- Resource Relation:
- Conference: 5. joint Russian-American computational mathematics conference, Albuquerque, NM (United States), 2-5 Sep 1997; Other Information: PBD: [1997]; Related Information: Is Part Of Proceedings of the 5. joint Russian-American computational mathematics conference; PB: 312 p.
- Country of Publication:
- United States
- Language:
- English
Similar Records
Hybrid and Parallel Domain-Decomposition Methods Development to Enable Monte Carlo for Reactor Analyses
Parallel MCNP Monte Carlo transport calculations with MPI