|
aegir
Phlex-based simulation framework for the SHiP experiment.
|
Phlex-based simulation framework for the SHiP experiment.
An automatic class reference is built using Doxygen from comments in the C++ code.
| Plugin | Type | Description |
|---|---|---|
particle_gun_source | Source | Configurable single-particle gun with Random123 RNG |
pythia8_source | Source | Pythia8 fixed-target p-p (serial or PythiaParallel) |
fixed_target_source | Source | Dual-target Pythia8 (p-p/p-n) with interaction point sampling |
geometry_builtin_provider | Provider | W target + Si scoring planes (test geometry) |
geometry_gdml_provider | Provider | GDML file loader |
geometry_geomodel_provider | Provider | GeoModel .db via SHiPGeometryService (optional) |
geant4_module | Transform | Geant4 simulation (direct worker, configurable concurrency) |
sim_output_module | Observer | RNTuple parallel writer + validation histograms |
Pixi resolves everything below from prefix.dev/ship (the SHiP-specific packages live there, with rattler-build recipes in ship-conda-recipes) and conda-forge. The list is reference; you do not need to install these manually when using pixi.
Install pixi, then:
pixi run build depends on configure, so you do not need to run them separately. Available tasks (see pixi.toml):
configure — run CMake with CMAKE_PREFIX_PATH=$CONDA_PREFIX.build — cmake --build build -j. Depends on configure.install — cmake --install build, populating $CONDA_PREFIX/lib and bin.test — ctest --test-dir build --output-on-failure (no-op until tests are added).smoke — quick end-to-end check used by CI (scripts/smoke.sh).clean — rm -rf build.pixi run <task> (or pixi run <command>) activates the environment for the duration of one command. Use it for scripted invocations and CI.pixi shell drops you into an interactive subshell with the environment active until you exit. Use it for iterative development — just targets, ad-hoc phlex invocations, debugging.activate.sh sets PHLEX_PLUGIN_PATH, LD_LIBRARY_PATH, SHIPGEOMETRY_ROOT, and AEGIR_ROOT automatically in both modes.
For environments without pixi (e.g. existing aliBuild setups), the project is a plain CMake build. You are responsible for providing the dependencies; see ship-conda-recipes for the canonical version pins.
Pass -DSHiPDataModel_ROOT=/path/to/data-model/install if a dependency is not in the default search path.
workflows/ ships a sampler of jsonnet configs; the snippets below pick five common ones. ls workflows/ shows the full set (multi-threaded variants, no-op sinks, GeoModel-based geometry, etc.).
If you built manually, point PHLEX_PLUGIN_PATH at the build tree yourself:
Three geometry providers are available, selectable via workflow configuration:
geometry_builtin_provider) — Tungsten target + Si scoring planes (test geometry). Used by workflows/gun_st.jsonnet and friends. No setup needed.geometry_gdml_provider) — load from a .gdml file. Set gdml_file in the workflow to an absolute path. Example: workflows/gun_st_gdml.jsonnet.GeoModel (geometry_geomodel_provider) — load from a GeoModel .db file via SHiPGeometryService. The shipgeometry conda package ships a build_geometry tool but does not bundle a pre-built DB; you generate one yourself:
Workflows reference the DB by bare filename (‘db_file: 'ship_geometry.db’); the provider resolves it against the CWD first and, failing that, against $SHIPGEOMETRY_ROOT/share/geometry/.activate.shsetsSHIPGEOMETRY_ROOT to$CONDA_PREFIXby default, so dropping a DB under $CONDA_PREFIX/share/geometry/` makes it discoverable from any CWD.
See docs/geant4_integration.md for details on the direct worker integration strategy.
See docs/benchmarks.md for details. The benchmark targets need hyperfine and just, which the bench pixi feature provides:
See CONTRIBUTING.md for details on our development workflow and coding standards.
-DGEANT4_USE_SYSTEM_ZLIB=ON — only relevant for manual builds; the conda-forge geant4 package already takes care of this.LGPL-3.0-or-later. See LICENSES/ for the full text.
1.9.8