OpenSn
  • Quick Install Guide
  • Theory Manual
  • Tutorials
  • Python API
  • Developer’s Guide
  • Gallery
    • 1. Capable of sweeps on Polyhedral meshes
    • 2. C5G7 Criticality Benchmark with 768 processors
    • 3. Real world simulations
OpenSn
  • Gallery

Gallery

1. Capable of sweeps on Polyhedral meshes

Sphere embedded within a box:

  • Concave cells induce cyclic dependencies between cells.

  • Non-cutting volumetric partitioning induces cyclic dependencies between processors.

  • Data structures allow sweeping with all forms of cycles.

Polyhedral mesh threshold

3-D polyhedral mesh generated with STAR-CCM+.

Solution slice

Slice of the solution.

KBA-style partitioning

Arbitrary, non-cutting, KBA-style partitioning used.

2. C5G7 Criticality Benchmark with 768 processors

Very old results (circa 2020) for the reactor benchmark C5G7.

  • These results did not use acceleration techniques (to be updated in Spring 2024)

  • 7 energy groups

  • 200 directions

  • 454 491 cells

  • Ran on 768 processors

  • Took only 18 minutes to complete (less than 2 min today)

  • Used 584 GB of memory

  • k_eff within 100 pcm (much less today)

C5G7 mesh materials

Close-up view of the mesh used (colors represent materials).

Energy group 0 solution

Energy group 0 solution.

Energy group 6 solution

Energy group 6 solution.

768-way ParMETIS partition

ParMETIS partitioning of the mesh (768 processors).

ParMETIS partition close-up

Close-up of the ParMETIS partitioning with the mesh visible.

3. Real world simulations

The Center for Exascale Radiation Transport (CERT) simulated—and compared to experiment—a graphite pile with a high-energy neutron source. This simulation used:

  • ~172 energy groups

  • Over 3000 directions

  • ~500 k cells

  • Over 100 k processors for some simulations

CERT simulation

CERT simulation versus experiment.

Previous

© Copyright 2023-present, OpenSn team.

Built with Sphinx using a theme provided by Read the Docs.