Every configuration input is now physical and is converted to solver units exactly once; post-processing reports every output physically under dimensionalize. This changes what several inputs mean when the reference scales are not 1 - picurv validate names them - and fixes outputs that mixed unit systems.
- Now read in physical units: particles.point_source, wall_function.roughness_height, solver.yml uniform_flow, the verification.sources profiles, field_statistics window start_time, end_time and time_cadence, ic_gen expressions (physical coordinates and values), the spectral provider velocities and wavenumbers, and a mode: file initial condition. A file saved by an earlier run declares its scales with source_case, or velocity_scale and length_scale.
- Every input records its dimension and conversion site in INPUT_QUANTITIES, and every catalogued field a FIELD_DIM_* dimension; units.nondimensionalization is now an enforced contract, checked by audit_units.py against page 19's input and field indexes.
- Loaded fields are scaled as they are read, so a field a step does not reload is no longer rescaled; Nu_t, Diffusivity, Phi and the other catalogued fields now have scales instead of staying non-dimensional with a warning. The DimensionalizeAllLoadedFields pipeline stage is retired.
- MSD columns, spectra and field-statistics history times, and ParaView collection times are physical under dimensionalize; runtime diagnostics CSVs gain a physical_time column.