How 2018uch0002/abeille works
Project Type and Purpose
• Abeille is a command-line, full-domain Monte-Carlo neutron transport simulation code written in C++20.
• It ingests a single YAML input file describing geometry, materials, sources, and simulation settings, then solves neutron transport (fixed-source, k-eigenvalue, or noise problems).
• Key features: multiple particle tracking strategies (surface, delta, Carter), weight cancellation, tallies, plotting, OpenMP/MPI parallelism.
High-Level Architectural Components
• CLI & Configuration Layer
• “abeille” executable (main.cpp)
• YAML Input File & Parser (utils/parser, yaml-cpp)
• Core Data Models
• Geometry Module (include/geometry, surfaces, lattices, universes)
• Materials Module (include/materials, nuclides, cross sections)
• Source Module (include/source, spatial/energy/direction distributions)
• Simulation Engine
• Simulation Orchestrator (simulation/simulation.hpp & .cpp)
• Particle Mover & Tracker (particle_mover, transport_operators: surface, delta, Carter)
• Collision Operators (branching, fission, noise variants, cancellation strategies in cancelator)
• Tally & Output
• Tallies (include/tallies: filters & estimators)
• Output (.npy via ndarray, GUI/text plotters)
• Plotting Module (include/plotting: slice, pixel, GUI)
• Utilities & Infrastructure
• RNG & PCG (utils/rng, pcg)
• MPI & OpenMP (utils/mpi + compiler flags)
• Settings, error handling, timers, logging, majorant calculations
• Build & External Dependencies
• CMake-based build (CMakeLists.txt, vendor downloads: papillon-ndl, yaml-cpp, docopt, ndarray, ImApp)
• Python scripts for preprocessing (abeille.py, input file generators)
Relationships and Data Flow
• The user invokes the CLI with a YAML file
→ Parser reads settings, geometry, materials, source specs
→ Core Data Models instantiated in memory
→ Simulation Orchestrator selects tracking strategy (strategy pattern)
→ Particle Mover loops over particles: invokes Transport Operator → Collision Operator → optionally Cancellation Module → updates tallies
→ When simulation completes, Tally Module aggregates results and writes .npy files
→ Optional Plotter reads mesh outputs and produces geometry or result visualizations
→ Logs, timers and statistics emitted via Utils
Architectural Patterns and Principles
• Modular design: each functional area is its own namespace/module (geometry, materials, source, simulation, tally, plotting, utils).
• Strategy pattern for tracking algorithms (surface, delta, Carter).
• Policy-based design for collision operators and cancellation techniques.
• Separation of concerns: parsing, data modeling, simulation logic, I/O, and visualization are decoupled.
• Parallelism layers: shared-memory (OpenMP) intrinsics in inner loops and optional distributed-memory (MPI) in utils/mpi.
• Clear labels for each module (e.g., “Geometry Module”, “Collision Operators”).
• Core simulation (blue rectangles)
• External libs & build (gray).