config_t Derived Type

Set to .false. to use the built-in no-pivoting solver instead. Only relevant when time_scheme = 'beuler'.

'lax_friedrichs' (default), 'steger_warming', 'van_leer', 'ausm_plus'

'auto' (default) — enable per-scheme: on for weno5/weno5z/weno_cu6/weno7/ weno9/weno11/eno3/mp5/teno5, off otherwise. 'yes' — always apply eigensystem decomposition (more accurate for shocks, higher cost). 'no' — always skip eigensystem decomposition (faster; safe for scalar/smooth problems).

Valid values: 'minmod' (default), 'superbee', 'mc', 'van_leer', 'koren'. Ignored for all non-MUSCL reconstruction schemes.

When .true., reconstructed face states on the FDS path (AUSM+, HLL, HLLC, Roe) are scaled toward the adjacent cell average to ensure density and pressure remain non-negative. Ignored when a FVS flux scheme is active.

When .true., each face is classified smooth or non-smooth using the selected sensor (hybrid_sensor). Smooth faces use the linear (optimal-weight) WENO variant — same order as the primary scheme but without nonlinear dissipation. Non-smooth faces fall back to the full nonlinear primary reconstruction scheme.

Valid values: 'jameson' — Jameson-Schmidt-Turkel pressure second-derivative sensor (JST 1981); physically motivated for compressible flows. Threshold ~ 0.1. 'density_gradient' — Normalised density jump across the face; cheap and robust. Threshold ~ 0.1. 'weno_beta' — Ratio max(β)/min(β) of WENO5 smoothness indicators; only meaningful for 5-point WENO-family schemes; falls back to 'density_gradient' otherwise. Threshold ~ 50.

Default 0.1 suits the 'jameson' and 'density_gradient' sensors. Use ~50 for 'weno_beta'.

Valid values: 'sod', 'shu_osher', 'smooth_wave', 'linear_advection', 'woodward_colella', 'lax', 'from_file', 'udf'.

File must contain whitespace-separated columns: x rho u p (one line per grid point; same format as the solver output file).

When .false., a grid mismatch causes an error stop.

Used when problem_type = 'udf'. The solver compiles this file to a shared library at runtime (requires gfortran on PATH) and calls the subroutine to fill the initial state. See example/udf_sod.f90 for the required interface.

Valid values: 'dirichlet' (default), 'inflow', 'outflow', 'reflecting', 'periodic', 'nonreflecting', 'supersonic_inlet', 'subsonic_inlet', 'supersonic_outlet', 'subsonic_outlet', 'neumann', 'neumann_gradient'.

Valid values: 'dirichlet' (default), 'inflow', 'outflow', 'reflecting', 'periodic', 'nonreflecting', 'supersonic_inlet', 'subsonic_inlet', 'supersonic_outlet', 'subsonic_outlet', 'neumann', 'neumann_gradient'.

Used only when bc_left = 'nonreflecting'.

Used only when bc_right = 'nonreflecting'.

0 = fully non-reflecting; 1 = Dirichlet target pressure p_ref. See Thompson (1987), Poinsot & Lele (1992).

'pressure' (default) = isentropic pressure-relaxation (Thompson 1987). 'characteristic' = full LODI characteristic targeting (Poinsot & Lele 1992, Sec. 3); activates u_ref_, rho_ref_, sigma_nrbc_entropy.

Used only when bc_left = 'nonreflecting' and nrbc_mode = 'characteristic'.

Used only when bc_right = 'nonreflecting' and nrbc_mode = 'characteristic'.

Used only when bc_left = 'nonreflecting' and nrbc_mode = 'characteristic'.

Used only when bc_right = 'nonreflecting' and nrbc_mode = 'characteristic'.

0 = fully non-reflecting entropy wave (default); 1 = entropy wave driven to the rho_ref target. Active only when nrbc_mode = 'characteristic'. See Poinsot & Lele (1992), J. Comput. Phys. 101, Sec. 3.

Used only when bc_left = 'subsonic_inlet'. Derived from static conditions via p0 = p(1 + (gam-1)/2Ma^2)^(gam/(gam-1)).

Used only when bc_left = 'subsonic_inlet'. Derived from static conditions via rho0 = rho(1 + (gam-1)/2Ma^2)^(1/(gam-1)).

Used only when bc_right = 'subsonic_inlet'.

Used only when bc_right = 'subsonic_inlet'.

Used only when bc_left = 'subsonic_outlet'.

Used only when bc_right = 'subsonic_outlet'.

Used only when bc_left = 'neumann_gradient'. Ghost state: q_ghost = q_wall + neumann_grad_left * dx.

Used only when bc_right = 'neumann_gradient'. Ghost state: q_ghost = q_wall + neumann_grad_right * dx.

Enable detailed wall-clock timing summary output (default .false.). When .true., compute_resid() and the driver accumulate fine-grained timing and print a performance table after the run. Lightweight per-iteration iter_s / elapsed_s logging is always available. Has no effect on results.

Logger verbosity threshold (default 3 = LOGLVL_INFO). 0 = SILENT, 1 = ERROR, 2 = WARN, 3 = INFO, 4 = DEBUG. See module logger for the LOGLVL_* constants.

Log file path (default 'run.log'). Empty string disables file logging.

Write a live solution snapshot every N iterations (0 = disabled). The snapshot file is overwritten each time, so a GUI app can watch it with inotify/kqueue. Format matches result.dat with a leading comment line: "# iter=NNN t=T".

Path for the live snapshot file (default 'snapshot.dat').

Write a checkpoint every N iterations (0 = disabled). Checkpoint files are named '_NNNNNN.bin' and contain enough state to resume the run exactly: iter, t, ub, and (if BDF2) bdf2_ub_prev. The file 'latest_checkpoint' is also updated to point to the most recent checkpoint so restarts need not know the iteration.

Base name for checkpoint files (default 'checkpoint').

Path to a checkpoint file to resume from (empty = fresh start). When set, the IC is skipped and iter/t are loaded from the checkpoint.