File-By-File GPU Audit

Generated by tools/generate_gpu_file_audit.py on 2026-05-19 11:11 UTC.

This page is a release-gate inventory for the GPU branch. It lists every top-level Fortran source, its runtime relevance, GPU status, retained environment switches, and developer notes. The status is generated from source markers plus curated role metadata; update the role map in the generator when ownership changes.

Summary

• Top-level Fortran files audited: 69
• LESGO-owned GPU checkpoints: 17
• CPU reference checkpoints preserved but excluded from GPU count: 2
• System probes: 2

Retained GPU Checkpoints

• LESGO_ATM_DIAG_TIMING
• LESGO_ATM_LB_AUTO_SELECT
• LESGO_ATM_LB_VALIDATE
• LESGO_ATM_POINT_OWNER_LB
• LESGO_ATM_POINT_OWNER_TARGETED
• LESGO_MPI_CUDA_DEBUG
• LESGO_MPI_CUDA_SYNC
• LESGO_PRESS_DIRECT_THOMAS_OUT
• LESGO_PRESS_RHS_HALO_COMBINED
• LESGO_PRESS_STAGE_TIMING
• LESGO_PRESS_TRANSPOSE_GENERIC
• LESGO_PRESS_TRANSPOSE_TIMING
• LESGO_PROJECT_STAGE_TIMING
• LESGO_SGS_CALCSIJ_EXPLICIT
• LESGO_SGS_HALO_COMBINED
• LESGO_SGS_STAGE_TIMING
• LESGO_SGS_STRICT_SYNC

CPU Reference Checkpoints

• LESGO_CPU_REF_TIME_FORCING
• LESGO_CPU_REF_TIME_TOTAL

Default Half-Channel SGS Audit Note

The latest 128^3 half-channel validation uses sgs_model=5, so it exercises the Lagrangian scale-dependent SGS path rather than only the short actuator-turbine benchmark path. The relevant runtime files are sgs_stag_util.f90, scaledep_dynamic.f90, interpolag_Sdep.f90, and lagrange_Sdep.f90; these cover the calc_Sij/Nu_t/tau construction plus the F_NN, F_QN, F_MM, and F_LM update work after model initialization.

Audit Matrix

File	Area	Role	Relevance	CMake/Option	GPU Status	Procedures	Switches	Loop Markers	GPU/MPI Markers	Developer Note
actuator_turbine_model.f90	ATM	Turbine physics, yaw/rotation, blade force logic	Runtime hot path when USE_ATM	USE_ATM, USE_TURBINES	GPU-enabled	actuator_turbine_model, atm_model_cuda_check, atm_model_cuda_sync, atm_initialize, atm_read_actuator_points, atm_read_restart, +19 more	-	60	CUF/global=6; CUDA/device=40; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
atm_base.f90	ATM	ATM type definitions	Setup/runtime support	USE_ATM	Review if runtime-active	atm_base, error, interpolate, vector_add, vector_divide, vector_multiply, +5 more	-	2	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
atm_input_util.f90	ATM	ATM input parsing	I/O/setup	USE_ATM, USE_TURBINES	GPU-enabled	atm_input_util, read_input_conf, read_turbine_model_variables, atm_print_initialize, read_airfoil, readline, +1 more	-	15	CUF/global=0; CUDA/device=34; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
atm_lesgo_interface.f90	ATM	GPU LESGO/ATM interface, sampling, gather, point-owner LB	Runtime hot path when USE_ATM	USE_ATM, USE_HIT, USE_TURBINES	GPU-enabled	atm_lesgo_interface, atm_lb_auto_record, atm_diag_event_start, atm_diag_event_stop, atm_diag_event_flush, atm_cuda_check, +43 more	LESGO_ATM_DIAG_TIMING, LESGO_ATM_LB_AUTO_SELECT, LESGO_ATM_LB_VALIDATE, LESGO_ATM_POINT_OWNER_LB, LESGO_ATM_POINT_OWNER_TARGETED	149	CUF/global=23; CUDA/device=141; MPI=198	Legacy ATM path is default; point-owner LB remains experimental and must validate force sums.
cfl_util.f90	Diagnostics	GPU CFL reduction	Runtime diagnostic	default/core	GPU-enabled	cfl_util, cfl_cuda_sync, get_max_cfl, get_cfl_dt	-	6	CUF/global=2; CUDA/device=21; MPI=10	Runtime loops have GPU coverage; preserve device-resident data flow.
clocks.f90	Timing	Wall-clock timers	Runtime support	default/core	Review if runtime-active	clock_m, start, stop	-	0	CUF/global=0; CUDA/device=0; MPI=4	Audit before using in a new active configuration.
concurrent_precursor.f90	CPS	Concurrent precursor coordination	Optional runtime	USE_CPS, USE_SCALARS	GPU-enabled	concurrent_precursor, initialize_cps, synchronize_cps, inflow_cps	-	4	CUF/global=1; CUDA/device=9; MPI=30	Runtime loops have GPU coverage; preserve device-resident data flow.
convec.f90	Convection	Production GPU convection path	Runtime hot path	default/core	GPU-enabled	convec_cuda_m, convec_cuda_impl, ensure_convec_cuda, padd_3d_dp, unpadd_3d_dp, check_convec_cuda, +3 more	-	53	CUF/global=16; CUDA/device=127; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
coriolis.f90	Forcing	GPU Coriolis forcing loops	Runtime	default/core	GPU-enabled	coriolis, coriolis_cuda_sync, coriolis_init, coriolis_finalize, coriolis_calc	-	9	CUF/global=3; CUDA/device=22; MPI=7	Runtime loops have GPU coverage; preserve device-resident data flow.
cuda_mpi_debug.f90	MPI/GPU debug	Concise CUDA-aware MPI pointer/sync diagnostics	Debug support	default/core	GPU-enabled	cuda_mpi_debug, cuda_mpi_debug_init, print_env, cuda_pre, cuda_post, probe_real, +5 more	LESGO_MPI_CUDA_DEBUG, LESGO_MPI_CUDA_SYNC	0	CUF/global=0; CUDA/device=107; MPI=41	Retained switch should be documented before changing behavior.
derivatives.f90	Derivatives	GPU spatial derivative kernels	Runtime hot path	default/core	GPU-enabled	derivatives, filt_da_cuda, xy_derivs_cuda, stress_uv_xy_derivs_cuda, stress_w_xy_derivs_cuda, stress_uv_div_cuda, +19 more	-	114	CUF/global=38; CUDA/device=305; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
divstress_uv.f90	SGS/divstress	GPU horizontal stress divergence	Runtime hot path	default/core	GPU-enabled	divstress_uv	-	5	CUF/global=2; CUDA/device=25; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
divstress_w.f90	SGS/divstress	GPU vertical stress divergence	Runtime hot path	default/core	GPU-enabled	divstress_w	-	24	CUF/global=6; CUDA/device=37; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
emul_complex.f90	Math utility	Complex arithmetic compatibility	Shared	USE_SCALARS	Review if runtime-active	emul_complex, procedure, procedure, procedure, mul_real_complex_imag_scalar, mul_real_complex_2D, +2 more	-	6	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
fft.f90	FFT	FFT plan/state support	Runtime support	default/core	Review if runtime-active	fft, padd, unpadd, init_fft, init_wavenumber	-	2	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
finalize.f90	Finalization	Shutdown and cleanup	Finalization	USE_ATM, USE_TURBINES	Review if runtime-active	finalize	-	0	CUF/global=0; CUDA/device=0; MPI=4	Audit before using in a new active configuration.
forcing.f90	Forcing/projection	GPU forcing, applied force reset, projection timing	Runtime hot path	USE_ATM, USE_HIT	GPU-enabled	forcing, forcing_cuda_sync, forcing_random, forcing_applied, forcing_induced, project, +7 more	LESGO_PROJECT_STAGE_TIMING	41	CUF/global=13; CUDA/device=88; MPI=99	Retained switch should be documented before changing behavior.
fringe.f90	Inflow	Fringe-region support	Runtime	default/core	GPU-enabled	fringe, procedure, constructor	-	2	CUF/global=0; CUDA/device=2; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
functions.f90	Utility	Shared GPU-enabled helper loops and math	Shared/runtime	default/core	GPU-enabled	functions, procedure, procedure, procedure, procedure, procedure, +15 more	-	12	CUF/global=2; CUDA/device=21; MPI=19	Runtime loops have GPU coverage; preserve device-resident data flow.
grid.f90	Grid	Grid metadata and coordinates	Runtime data	default/core	Review if runtime-active	grid_m, build	-	5	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
hit_inflow.f90	HIT	Optional HIT input/reference path	Optional	USE_HIT	GPU-enabled	hit_inflow, initialize_HIT, extract_HIT_data, compute_HIT_plane_data, inflow_HIT, hit_write_restart, +2 more	-	27	CUF/global=2; CUDA/device=60; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
hit_inflow_gpu.f90	HIT	GPU HIT inflow helpers	Optional runtime	USE_HIT	GPU-enabled	hit_inflow_gpu, hit_gpu_setup, hit_compute_plane_gpu, hit_apply_fringe_gpu	-	9	CUF/global=2; CUDA/device=131; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
inflow.f90	Inflow	GPU inflow forcing/runtime loops	Runtime	USE_CPS, USE_HIT	GPU-enabled	inflow, inflow_cuda_sync, inflow_init, apply_inflow, inflow_uniform	-	4	CUF/global=1; CUDA/device=41; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
init_random_seed.f90	Initialization	Random seed control	Setup	default/core	Review if runtime-active	init_random_seed, lcg	LESGO_RANDOM_SEED	2	CUF/global=0; CUDA/device=0; MPI=0	Retained switch should be documented before changing behavior.
initial.f90	Initialization	Initial condition setup and device state	Initialization	USE_SCALARS, USE_TURBINES	GPU-enabled	initial, initial_cuda_sync, initial_cuda_touch_velocity, ic_uniform, check_for_interp, ic_file, +3 more	-	27	CUF/global=2; CUDA/device=24; MPI=10	Runtime loops have GPU coverage; preserve device-resident data flow.
initialize.f90	Initialization	Startup allocation and setup	Initialization	USE_ATM, USE_CPS, USE_HIT, USE_LVLSET, USE_SCALARS, USE_TURBINES	Review if runtime-active	initialize	-	0	CUF/global=0; CUDA/device=0; MPI=5	Audit before using in a new active configuration.
input_util.f90	Configuration	Input parsing helpers	I/O/setup	USE_HIT, USE_SCALARS, USE_TURBINES	CPU acceptable: setup/I/O	input_util, procedure, read_input_conf, domain_block, model_block, coriolis_block, +10 more	-	13	CUF/global=0; CUDA/device=0; MPI=0	CPU is acceptable unless this path becomes repeated inside the timestep.
interpolag_Sdep.f90	SGS	Lagrangian scale-dependent interpolation	Runtime	default/core	GPU-enabled	interpolag_Sdep, interpolag_sdep_cuda_sync	-	17	CUF/global=2; CUDA/device=22; MPI=17	Exercised by the default half-channel sgs_model=5 validation; preserve device-resident Lagrangian data flow.
interpolag_Ssim.f90	SGS	Lagrangian scale-sim interpolation	Runtime	default/core	GPU-enabled	interpolag_Ssim, interpolag_ssim_cuda_sync	-	11	CUF/global=2; CUDA/device=22; MPI=13	Runtime loops have GPU coverage; preserve device-resident data flow.
io.f90	Output	Output and optional CGNS paths	I/O	USE_TURBINES	GPU-enabled	io, openfiles, energy, write_tau_wall_bot, write_tau_wall_top, write_parallel_cgns, +7 more	-	40	CUF/global=1; CUDA/device=18; MPI=33	Runtime loops have GPU coverage; preserve device-resident data flow.
iwmles.f90	Wall model	Integral wall model support	Runtime	default/core	GPU-enabled	iwmles, iwm_cuda_sync, iwm_wallstress, iwm_init, iwm_finalize, iwm_calc_lhs, +5 more	-	28	CUF/global=5; CUDA/device=38; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
lagrange_Sdep.f90	SGS	Lagrangian scale-dependent averaging	Runtime	default/core	GPU-enabled	lagrange_Sdep, lagrange_sdep_cuda_sync	-	23	CUF/global=10; CUDA/device=59; MPI=19	Covers F_NN, F_QN, F_MM, and F_LM update work in the default half-channel validation.
lagrange_Ssim.f90	SGS	Lagrangian scale-sim averaging	Runtime	default/core	GPU-enabled	lagrange_Ssim, lagrange_ssim_cuda_sync	-	21	CUF/global=9; CUDA/device=53; MPI=15	Runtime loops have GPU coverage; preserve device-resident data flow.
level_set.f90	Level set	Optional level-set runtime loops	Optional runtime	USE_HIT, USE_LVLSET	GPU-enabled	level_set, level_set_cuda_sync, level_set_init, level_set_vel_err, level_set_Cs_lag_dyn, level_set_lag_dyn, +35 more	-	106	CUF/global=8; CUDA/device=34; MPI=165	Runtime loops have GPU coverage; preserve device-resident data flow.
level_set_base.f90	Level set	Level-set shared definitions	Optional setup	USE_LVLSET	GPU-enabled	level_set_base, level_set_base_init	-	0	CUF/global=0; CUDA/device=3; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
linear_simple.f90	Math utility	Linear helper routines	Shared	default/core	Review if runtime-active	linear_simple, procedure, solve_linear, assert_eq2, assert_eq3, assert_eq4, +4 more	-	3	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
main.f90	Main driver	Timestep orchestration, timing, CPU/GPU comparison output	Runtime	USE_SCALARS, USE_TURBINES	GPU-enabled	main, main_cuda_sync	LESGO_CPU_REF_TIME_FORCING, LESGO_CPU_REF_TIME_TOTAL	33	CUF/global=13; CUDA/device=49; MPI=29	Retained switch should be documented before changing behavior.
messages.f90	Messages	User-facing message helpers	Setup	default/core	CPU acceptable: setup/I/O	messages, procedure, procedure, message_a, message_ai, message_aiai, +20 more	-	0	CUF/global=0; CUDA/device=0; MPI=0	CPU is acceptable unless this path becomes repeated inside the timestep.
mpi_defs.f90	MPI	MPI decomposition, rank/GPU binding, communication helpers	Runtime	USE_CPS	GPU-enabled	mpi_defs, initialize_mpi, bind_cuda_device, create_mpi_comms_cps, mpi_sync_real_array, sync_down, +2 more	-	1	CUF/global=0; CUDA/device=52; MPI=90	Runtime loops have GPU coverage; preserve device-resident data flow.
mpi_transpose_mod.f90	MPI transpose	Transpose support used by pressure and spectral paths	Runtime	default/core	GPU-enabled	mpi_transpose_mod, mpi_transpose	-	8	CUF/global=0; CUDA/device=1; MPI=14	Runtime loops have GPU coverage; preserve device-resident data flow.
param.f90	Configuration	Global parameters and CMake/runtime option state	Setup	default/core	CPU acceptable: setup/I/O	param	-	0	CUF/global=0; CUDA/device=0; MPI=3	CPU is acceptable unless this path becomes repeated inside the timestep.
param_output.f90	Output	Output configuration	I/O/setup	default/core	CPU acceptable: setup/I/O	param_output	-	4	CUF/global=0; CUDA/device=0; MPI=0	CPU is acceptable unless this path becomes repeated inside the timestep.
pid.f90	Utility	PID/control helper	Setup	default/core	Review if runtime-active	pid_m, procedure, constructor, advance_noset, advance_set	-	0	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
press_stag_array.f90	Pressure	GPU pressure RHS, halos, cuFFT orchestration	Runtime hot path	default/core	GPU-enabled	press_stag_array, press_cuda_sync, press_queue_event_start, press_queue_event_stop, press_queue_report, press_rhs_halo_audit, +12 more	CUDA_VISIBLE_DEVICES, LESGO_PRESS_RHS_HALO_COMBINED, LESGO_PRESS_STAGE_TIMING, MPICH_GPU_SUPPORT_ENABLED	80	CUF/global=27; CUDA/device=250; MPI=137	Pressure math and zero-mode handling are release-sensitive; validate divergence and wall stress after edits.
rmsdiv.f90	Diagnostics	GPU divergence metric/reduction	Runtime diagnostic	default/core	GPU-enabled	rmsdiv	-	6	CUF/global=1; CUDA/device=19; MPI=3	Runtime loops have GPU coverage; preserve device-resident data flow.
scalars.f90	Scalars	Optional scalar transport GPU loops	Optional runtime	USE_SCALARS	GPU-enabled	scalars, scalars_cuda_sync, scalars_copy_rhs_gpu, scalars_advective_gpu, scalars_flux_gpu, scalars_rhs_theta_gpu, +12 more	-	65	CUF/global=20; CUDA/device=58; MPI=17	Runtime loops have GPU coverage; preserve device-resident data flow.
scaledep_dynamic.f90	SGS	Scale-dependent dynamic SGS model support	Runtime	default/core	GPU-enabled	scaledep_dynamic, scaledep_dynamic_cuda_sync	-	17	CUF/global=8; CUDA/device=31; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
sgs_param.f90	SGS	SGS parameters	Setup/runtime support	default/core	GPU-enabled	sgs_param, sgs_param_init	-	0	CUF/global=0; CUDA/device=7; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
sgs_stag_util.f90	SGS	GPU SGS model/stress construction and halos	Runtime hot path	default/core	GPU-enabled	sgs_stag_util, sgs_cuda_sync, sgs_cuda_barrier, sgs_event_record, sgs_event_elapsed_seconds, sgs_diag_time, +26 more	CUDA_VISIBLE_DEVICES, LESGO_SGS_CALCSIJ_EXPLICIT, LESGO_SGS_HALO_COMBINED, LESGO_SGS_STAGE_TIMING, LESGO_SGS_STRICT_SYNC, MPICH_GPU_SUPPORT_ENABLED	122	CUF/global=33; CUDA/device=275; MPI=205	Do not change SGS formulas while editing timing or halo paths; validate wall stress and the default half-channel SGS statistics.
shifted_inflow.f90	Inflow	Shifted inflow support	Runtime	USE_SCALARS	GPU-enabled	shifted_inflow, shifted_inflow_cuda_sync, shifted_inflow_init, inflow_shifted	-	10	CUF/global=3; CUDA/device=65; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
sim_param.f90	Configuration	Simulation parameters	Setup	default/core	GPU-enabled	sim_param, sim_param_init	-	0	CUF/global=0; CUDA/device=1; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
sponge.f90	Boundary	GPU sponge damping loops	Runtime	default/core	GPU-enabled	sponge, sponge_cuda_sync, sponge_init, sponge_force	-	7	CUF/global=2; CUDA/device=19; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
stability.f90	Stability	Stability helper routines	Runtime support	default/core	Review if runtime-active	stability, calc_phi, calc_psi_m, calc_psi_h	-	0	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
stat_defs.f90	Statistics	Statistics definitions	Runtime support	USE_TURBINES	Review if runtime-active	stat_defs	-	0	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
std_dynamic.f90	SGS	Standard dynamic SGS model support	Runtime	default/core	GPU-enabled	std_dynamic, std_dynamic_cuda_sync	-	13	CUF/global=6; CUDA/device=37; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
string_util.f90	Utility	String parsing helpers	Setup	USE_HIT	Review if runtime-active	string_util, procedure, procedure, procedure, numtostr_r, numtostr_i, +36 more	-	3	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
test_filtermodule.f90	Filtering	GPU test-filter support for dynamic SGS	Runtime hot path	default/core	GPU-enabled	test_filtermodule, require_test_filter_cufft_success, test_filter_cuda_sync, test_filter_cuda_barrier, ensure_test_filter_cuda_plan, ensure_test_filter_cuda_many_plan, +10 more	-	8	CUF/global=3; CUDA/device=102; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
time_average.f90	Statistics	Time averaging and output accumulation	Runtime/I/O	USE_ATM, USE_TURBINES	Review if runtime-active	time_average, init, compute, finalize, checkpoint, write_parallel_cgns, +1 more	-	20	CUF/global=0; CUDA/device=0; MPI=39	Audit before using in a new active configuration.
trees_base_ls.f90	Level set	Tree geometry base data	Optional setup	USE_LVLSET	Review if runtime-active	trees_base_ls, cross_product, delta, grid_initialize, grid_nearest_of_pt, grid_of_pt, +2 more	-	1	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
trees_global_fmask_ls.f90	Level set	Tree/fmask preprocessing	Optional setup	USE_LVLSET	GPU-enabled	trees_global_fmask_ls, global_fmask_init, calc_global_fmask_ta, read_global_fmask, write_global_fmask, write_fmt_global_fmask, +5 more	-	34	CUF/global=5; CUDA/device=21; MPI=7	Runtime loops have GPU coverage; preserve device-resident data flow.
trees_io_ls.f90	Level set	Tree I/O	Optional I/O	USE_LVLSET	Review if runtime-active	trees_io_ls, draw_cross_section, set_h, set_radius, write_tecplot_zone_hdr, draw_tree_array, +4 more	-	9	CUF/global=0; CUDA/device=0; MPI=0	Audit before using in a new active configuration.
trees_pre_ls.f90	Level set	Tree preprocessing	Optional setup	USE_LVLSET	GPU-enabled	trees_pre_ls_mod, trees_pre_ls	-	13	CUF/global=0; CUDA/device=3; MPI=13	Runtime loops have GPU coverage; preserve device-resident data flow.
trees_setup_ls.f90	Level set	Tree setup	Optional setup	USE_LVLSET	GPU-enabled	trees_setup_ls, sdistfcn_tree_array, dist_plate, dist_circle, dist_circle_b, dist_circle_c, +16 more	-	16	CUF/global=1; CUDA/device=16; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
tridag_array.f90	Pressure	GPU tridiagonal solve and transpose-Thomas helper	Runtime hot path	default/core	GPU-enabled	tridag_transpose_gpu_mod, tridag_array, tridag_array_transpose_thomas_cuda, tridag_array_spike2_cuda, tridag_array_replicated_cuda, tridag_array	CUDA_VISIBLE_DEVICES, LESGO_PRESS_DIRECT_THOMAS_OUT, LESGO_PRESS_TRANSPOSE_GENERIC, LESGO_PRESS_TRANSPOSE_TIMING, MPICH_GPU_SUPPORT_ENABLED	89	CUF/global=34; CUDA/device=308; MPI=158	Pressure math and zero-mode handling are release-sensitive; validate divergence and wall stress after edits.
turbine_indicator.f90	Turbines	Turbine indicator setup/runtime helpers	Optional setup/runtime	USE_TURBINES	GPU-enabled	turbine_indicator, val, init	-	12	CUF/global=7; CUDA/device=40; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
turbines.f90	Turbines	Optional actuator disk/turbine routines	Optional runtime	USE_TURBINES	GPU-enabled	turbines, turbines_init, turbines_nodes, turbines_forcing, turbines_finalize, turbines_checkpoint, +3 more	-	35	CUF/global=0; CUDA/device=5; MPI=31	Runtime loops have GPU coverage; preserve device-resident data flow.
turbines_gpu.f90	Turbines	GPU helper routines for optional turbine paths	Optional runtime	USE_TURBINES	GPU-enabled	turbines_gpu, turbines_interp_w_to_uv_gpu	-	5	CUF/global=2; CUDA/device=8; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.
types.f90	Shared types	Core precision and type definitions	Shared	default/core	CPU acceptable: setup/I/O	types	-	0	CUF/global=0; CUDA/device=0; MPI=0	CPU is acceptable unless this path becomes repeated inside the timestep.
wallstress.f90	Wall model	GPU wall-stress related loops and diagnostics	Runtime	default/core	GPU-enabled	wallstress, wallstress_cuda_sync, ws_free_lbc, ws_free_ubc, ws_dns_lbc, ws_dns_ubc, +2 more	-	26	CUF/global=9; CUDA/device=39; MPI=0	Runtime loops have GPU coverage; preserve device-resident data flow.

Procedure Inventory

actuator_turbine_model.f90

• Area: ATM
• Runtime relevance: Runtime hot path when USE_ATM
• GPU status: GPU-enabled
• Procedures/modules found: actuator_turbine_model, atm_model_cuda_check, atm_model_cuda_sync, atm_initialize, atm_read_actuator_points, atm_read_restart, atm_write_restart, atm_initialize_output, atm_create_points, atm_update, atm_control_yaw, atm_computeRotorSpeed, atm_rotateBlades, atm_compute_cl_correction, atm_compute_cl_correction_gpu, s_fit, atm_calculate_variables, atm_computeBladeForce, atm_computeNacelleForce, atm_integrate_u, atm_yawNacelle, atm_output, atm_compute_power, atm_write_blade_points, atm_process_output
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

atm_base.f90

• Area: ATM
• Runtime relevance: Setup/runtime support
• GPU status: Review if runtime-active
• Procedures/modules found: atm_base, error, interpolate, vector_add, vector_divide, vector_multiply, vector_mag, rotatePoint, matrix_vector, cross_product, distance
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

atm_input_util.f90

• Area: ATM
• Runtime relevance: I/O/setup
• GPU status: GPU-enabled
• Procedures/modules found: atm_input_util, read_input_conf, read_turbine_model_variables, atm_print_initialize, read_airfoil, readline, eat_whitespace
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

atm_lesgo_interface.f90

• Area: ATM
• Runtime relevance: Runtime hot path when USE_ATM
• GPU status: GPU-enabled
• Procedures/modules found: atm_lesgo_interface, atm_lb_auto_record, atm_diag_event_start, atm_diag_event_stop, atm_diag_event_flush, atm_cuda_check, atm_cuda_sync, atm_lesgo_reset_turbine_gpu, atm_interp_w_to_uv, atm_prepare_direct_w, atm_lesgo_apply_force_gpu, atm_lesgo_convolute_force_gpu_atpoint, atm_lesgo_force_gpu_atpoint, atm_lesgo_nacelle_force_gpu_atpoint, atm_point_owner_sample_turbine, atm_point_owner_force_turbine, atm_lb_ensure_targeted_buffers, atm_lb_refresh_targeted_sample_slots, atm_lb_targeted_velocity_exchange, atm_lb_pack_force_turbine, atm_lb_unpack_force_turbine, atm_point_owner_lb_gather_targeted, atm_point_owner_lb_force, atm_point_owner_lb_gather, atm_point_owner_lb_validate, atm_lesgo_initialize, atm_lesgo_finalize, atm_lesgo_report_timing, atm_lesgo_diag_load, atm_lesgo_lb_plan, +19 more
• Retained switches: LESGO_ATM_DIAG_TIMING, LESGO_ATM_LB_AUTO_SELECT, LESGO_ATM_LB_VALIDATE, LESGO_ATM_POINT_OWNER_LB, LESGO_ATM_POINT_OWNER_TARGETED
• Developer note: Legacy ATM path is default; point-owner LB remains experimental and must validate force sums.

cfl_util.f90

• Area: Diagnostics
• Runtime relevance: Runtime diagnostic
• GPU status: GPU-enabled
• Procedures/modules found: cfl_util, cfl_cuda_sync, get_max_cfl, get_cfl_dt
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

clocks.f90

• Area: Timing
• Runtime relevance: Runtime support
• GPU status: Review if runtime-active
• Procedures/modules found: clock_m, start, stop
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

concurrent_precursor.f90

• Area: CPS
• Runtime relevance: Optional runtime
• GPU status: GPU-enabled
• Procedures/modules found: concurrent_precursor, initialize_cps, synchronize_cps, inflow_cps
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

convec.f90

• Area: Convection
• Runtime relevance: Runtime hot path
• GPU status: GPU-enabled
• Procedures/modules found: convec_cuda_m, convec_cuda_impl, ensure_convec_cuda, padd_3d_dp, unpadd_3d_dp, check_convec_cuda, convec_cuda_sync, require_convec_cufft, convec
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

coriolis.f90

• Area: Forcing
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: coriolis, coriolis_cuda_sync, coriolis_init, coriolis_finalize, coriolis_calc
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

cuda_mpi_debug.f90

• Area: MPI/GPU debug
• Runtime relevance: Debug support
• GPU status: GPU-enabled
• Procedures/modules found: cuda_mpi_debug, cuda_mpi_debug_init, print_env, cuda_pre, cuda_post, probe_real, probe_complex, mpi_dbg_sendrecv_r, mpi_dbg_sendrecv_c, mpi_dbg_send_r, mpi_dbg_recv_r
• Retained switches: LESGO_MPI_CUDA_DEBUG, LESGO_MPI_CUDA_SYNC
• Developer note: Retained switch should be documented before changing behavior.

derivatives.f90

• Area: Derivatives
• Runtime relevance: Runtime hot path
• GPU status: GPU-enabled
• Procedures/modules found: derivatives, filt_da_cuda, xy_derivs_cuda, stress_uv_xy_derivs_cuda, stress_w_xy_derivs_cuda, stress_uv_div_cuda, stress_w_div_cuda, filt_da_vel_cuda, ensure_filt_da_cuda_plan, ensure_filt_da_vel_cuda_plan, ensure_stress_uv_xy_cuda_plan, ensure_stress_w_xy_cuda_plan, derivatives_cuda_sync, require_filt_da_cufft_success, require_filt_da_cuda_success, stress_uv_xy_derivs, stress_w_xy_derivs, ddx, ddy, ddxy, filt_da, filt_da_vel, ddz_vel, ddz_uv, ddz_w
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

divstress_uv.f90

• Area: SGS/divstress
• Runtime relevance: Runtime hot path
• GPU status: GPU-enabled
• Procedures/modules found: divstress_uv
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

divstress_w.f90

• Area: SGS/divstress
• Runtime relevance: Runtime hot path
• GPU status: GPU-enabled
• Procedures/modules found: divstress_w
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

emul_complex.f90

• Area: Math utility
• Runtime relevance: Shared
• GPU status: Review if runtime-active
• Procedures/modules found: emul_complex, procedure, procedure, procedure, mul_real_complex_imag_scalar, mul_real_complex_2D, mul_real_complex_imag_2D, mul_real_complex_real_2D
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

fft.f90

• Area: FFT
• Runtime relevance: Runtime support
• GPU status: Review if runtime-active
• Procedures/modules found: fft, padd, unpadd, init_fft, init_wavenumber
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

finalize.f90

• Area: Finalization
• Runtime relevance: Finalization
• GPU status: Review if runtime-active
• Procedures/modules found: finalize
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

forcing.f90

• Area: Forcing/projection
• Runtime relevance: Runtime hot path
• GPU status: GPU-enabled
• Procedures/modules found: forcing, forcing_cuda_sync, forcing_random, forcing_applied, forcing_induced, project, project_sync_velocity_halos_cuda, project_sync_velocity_direct_halos_cuda, project_sync_velocity_direct_halos_overlap_cuda, project_stage_report, project_ensure_halo_buffers, project_pack_velocity_halos_cuda, project_unpack_velocity_halos_cuda
• Retained switches: LESGO_PROJECT_STAGE_TIMING
• Developer note: Retained switch should be documented before changing behavior.

fringe.f90

• Area: Inflow
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: fringe, procedure, constructor
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

functions.f90

• Area: Utility
• Runtime relevance: Shared/runtime
• GPU status: GPU-enabled
• Procedures/modules found: functions, procedure, procedure, procedure, procedure, procedure, procedure, tau_wall_cuda_sync, interp_to_uv_grid, interp_to_w_grid, bilinear_interp_sa_nocheck, bilinear_interp_sa, bilinear_interp_aa, linear_interp_sa_nocheck, linear_interp_sa, linear_interp_aa, cross_product, binary_search, get_tau_wall_bot, get_tau_wall_top, count_lines
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

grid.f90

• Area: Grid
• Runtime relevance: Runtime data
• GPU status: Review if runtime-active
• Procedures/modules found: grid_m, build
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

hit_inflow.f90

• Area: HIT
• Runtime relevance: Optional
• GPU status: GPU-enabled
• Procedures/modules found: hit_inflow, initialize_HIT, extract_HIT_data, compute_HIT_plane_data, inflow_HIT, hit_write_restart, hit_read_restart, interpolate3D
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

hit_inflow_gpu.f90

• Area: HIT
• Runtime relevance: Optional runtime
• GPU status: GPU-enabled
• Procedures/modules found: hit_inflow_gpu, hit_gpu_setup, hit_compute_plane_gpu, hit_apply_fringe_gpu
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

inflow.f90

• Area: Inflow
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: inflow, inflow_cuda_sync, inflow_init, apply_inflow, inflow_uniform
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

init_random_seed.f90

• Area: Initialization
• Runtime relevance: Setup
• GPU status: Review if runtime-active
• Procedures/modules found: init_random_seed, lcg
• Retained switches: LESGO_RANDOM_SEED
• Developer note: Retained switch should be documented before changing behavior.

initial.f90

• Area: Initialization
• Runtime relevance: Initialization
• GPU status: GPU-enabled
• Procedures/modules found: initial, initial_cuda_sync, initial_cuda_touch_velocity, ic_uniform, check_for_interp, ic_file, ic_interp, ic_dns, ic_les
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

initialize.f90

• Area: Initialization
• Runtime relevance: Initialization
• GPU status: Review if runtime-active
• Procedures/modules found: initialize
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

input_util.f90

• Area: Configuration
• Runtime relevance: I/O/setup
• GPU status: CPU acceptable: setup/I/O
• Procedures/modules found: input_util, procedure, read_input_conf, domain_block, model_block, coriolis_block, time_block, flow_cond_block, output_block, level_set_block, turbines_block, scalars_block, checkentry, readline, parse_vector_real, parse_vector_point3D
• Retained switches: -
• Developer note: CPU is acceptable unless this path becomes repeated inside the timestep.

interpolag_Sdep.f90

• Area: SGS
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: interpolag_Sdep, interpolag_sdep_cuda_sync
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

interpolag_Ssim.f90

• Area: SGS
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: interpolag_Ssim, interpolag_ssim_cuda_sync
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

io.f90

• Area: Output
• Runtime relevance: I/O
• GPU status: GPU-enabled
• Procedures/modules found: io, openfiles, energy, write_tau_wall_bot, write_tau_wall_top, write_parallel_cgns, write_null_cgns, output_loop, inst_write, force_tot, checkpoint, output_final, output_init
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

iwmles.f90

• Area: Wall model
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: iwmles, iwm_cuda_sync, iwm_wallstress, iwm_init, iwm_finalize, iwm_calc_lhs, iwm_slv, iwm_calc_wallstress, iwm_monitor, iwm_checkPoint, iwm_read_checkPoint
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

lagrange_Sdep.f90

• Area: SGS
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: lagrange_Sdep, lagrange_sdep_cuda_sync
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

lagrange_Ssim.f90

• Area: SGS
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: lagrange_Ssim, lagrange_ssim_cuda_sync
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

level_set.f90

• Area: Level set
• Runtime relevance: Optional runtime
• GPU status: GPU-enabled
• Procedures/modules found: level_set, level_set_cuda_sync, level_set_init, level_set_vel_err, level_set_Cs_lag_dyn, level_set_lag_dyn, modify_beta, neumann_F_MM, mpi_sync_F_MM, zero_F_LM, modify_dutdn, extrap_tau_simple, extrap_tau_log, enforce_un, enforce_log_profile, interp_scal, interp_tij_u, fill_f, interp_tij_w, fill_f, interp_phi, interp_vel, smooth_tau, level_set_smooth_vel, smooth, level_set_global_CA, level_set_Cs, mpi_sync, mpi_sync_tau, level_set_BC, +11 more
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

level_set_base.f90

• Area: Level set
• Runtime relevance: Optional setup
• GPU status: GPU-enabled
• Procedures/modules found: level_set_base, level_set_base_init
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

linear_simple.f90

• Area: Math utility
• Runtime relevance: Shared
• GPU status: Review if runtime-active
• Procedures/modules found: linear_simple, procedure, solve_linear, assert_eq2, assert_eq3, assert_eq4, ludcmp, lubksb, outerprod, swap
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

main.f90

• Area: Main driver
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: main, main_cuda_sync
• Retained switches: LESGO_CPU_REF_TIME_FORCING, LESGO_CPU_REF_TIME_TOTAL
• Developer note: Retained switch should be documented before changing behavior.

messages.f90

• Area: Messages
• Runtime relevance: Setup
• GPU status: CPU acceptable: setup/I/O
• Procedures/modules found: messages, procedure, procedure, message_a, message_ai, message_aiai, message_aiar, message_al, message_aii, message_air, message_ai_array, message_aiai_array, message_ar, message_ar_array, message_aiar_array, warn, error_a, error_ai, error_ai_array, error_aia, error_aiai, error_aiar, error_arar, error_al, error_ar, error_ar_array
• Retained switches: -
• Developer note: CPU is acceptable unless this path becomes repeated inside the timestep.

mpi_defs.f90

• Area: MPI
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: mpi_defs, initialize_mpi, bind_cuda_device, create_mpi_comms_cps, mpi_sync_real_array, sync_down, sync_up, sync_downup_nb
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

mpi_transpose_mod.f90

• Area: MPI transpose
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: mpi_transpose_mod, mpi_transpose
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

param.f90

• Area: Configuration
• Runtime relevance: Setup
• GPU status: CPU acceptable: setup/I/O
• Procedures/modules found: param
• Retained switches: -
• Developer note: CPU is acceptable unless this path becomes repeated inside the timestep.

param_output.f90

• Area: Output
• Runtime relevance: I/O/setup
• GPU status: CPU acceptable: setup/I/O
• Procedures/modules found: param_output
• Retained switches: -
• Developer note: CPU is acceptable unless this path becomes repeated inside the timestep.

pid.f90

• Area: Utility
• Runtime relevance: Setup
• GPU status: Review if runtime-active
• Procedures/modules found: pid_m, procedure, constructor, advance_noset, advance_set
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

press_stag_array.f90

• Area: Pressure
• Runtime relevance: Runtime hot path
• GPU status: GPU-enabled
• Procedures/modules found: press_stag_array, press_cuda_sync, press_queue_event_start, press_queue_event_stop, press_queue_report, press_rhs_halo_audit, press_rhs_halo_report, press_rhs_assembly_report, press_stage_report, press_pack_rhs_cuda, press_pack_rhs_halo_cuda, press_unpack_rhs_halo_cuda, press_pack_rhs_halo_combined_cuda, press_unpack_rhs_halo_combined_cuda, press_rhs_prep_cuda, press_assemble_rhs_cuda, press_assemble_rhs_range_cuda, press_zero_mode_cuda
• Retained switches: CUDA_VISIBLE_DEVICES, LESGO_PRESS_RHS_HALO_COMBINED, LESGO_PRESS_STAGE_TIMING, MPICH_GPU_SUPPORT_ENABLED
• Developer note: Pressure math and zero-mode handling are release-sensitive; validate divergence and wall stress after edits.

rmsdiv.f90

• Area: Diagnostics
• Runtime relevance: Runtime diagnostic
• GPU status: GPU-enabled
• Procedures/modules found: rmsdiv
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

scalars.f90

• Area: Scalars
• Runtime relevance: Optional runtime
• GPU status: GPU-enabled
• Procedures/modules found: scalars, scalars_cuda_sync, scalars_copy_rhs_gpu, scalars_advective_gpu, scalars_flux_gpu, scalars_rhs_theta_gpu, buoyancy_force_gpu, scalars_init, ic_scal, ic_scal_file, ic_scal_les, ic_scal_interp, scalars_checkpoint, scalars_deriv, obukhov, scalars_transport, to_big, buoyancy_force
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

scaledep_dynamic.f90

• Area: SGS
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: scaledep_dynamic, scaledep_dynamic_cuda_sync
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

sgs_param.f90

• Area: SGS
• Runtime relevance: Setup/runtime support
• GPU status: GPU-enabled
• Procedures/modules found: sgs_param, sgs_param_init
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

sgs_stag_util.f90

• Area: SGS
• Runtime relevance: Runtime hot path
• GPU status: GPU-enabled
• Procedures/modules found: sgs_stag_util, sgs_cuda_sync, sgs_cuda_barrier, sgs_event_record, sgs_event_elapsed_seconds, sgs_diag_time, sgs_calc_diag_begin, sgs_calc_diag_start, sgs_calc_diag_stop, sgs_calc_cpu_start, sgs_calc_cpu_stop, sgs_calc_set_zrange, sgs_calc_set_audit, sgs_tau_detail_begin, sgs_tau_detail_start, sgs_tau_detail_stop, sgs_tau_detail_add_bytes, sgs_tau_detail_add_msg, sgs_dwdz_detail_begin, sgs_dwdz_detail_start, sgs_dwdz_detail_stop, sgs_dwdz_detail_add_msg, sgs_dwdz_path_audit, sgs_pointer_env_audit, sgs_pointer_env_audit_device, sgs_stag, calc_Sij, calc_Sij_nut_dynamic_cuda, sgs_sync_dwdz_down_cuda, sgs_calc_sij_detail_report, +2 more
• Retained switches: CUDA_VISIBLE_DEVICES, LESGO_SGS_CALCSIJ_EXPLICIT, LESGO_SGS_HALO_COMBINED, LESGO_SGS_STAGE_TIMING, LESGO_SGS_STRICT_SYNC, MPICH_GPU_SUPPORT_ENABLED
• Developer note: Do not change SGS formulas while editing timing or halo paths; validate tau/wall-stress behavior.

shifted_inflow.f90

• Area: Inflow
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: shifted_inflow, shifted_inflow_cuda_sync, shifted_inflow_init, inflow_shifted
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

sim_param.f90

• Area: Configuration
• Runtime relevance: Setup
• GPU status: GPU-enabled
• Procedures/modules found: sim_param, sim_param_init
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

sponge.f90

• Area: Boundary
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: sponge, sponge_cuda_sync, sponge_init, sponge_force
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

stability.f90

• Area: Stability
• Runtime relevance: Runtime support
• GPU status: Review if runtime-active
• Procedures/modules found: stability, calc_phi, calc_psi_m, calc_psi_h
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

stat_defs.f90

• Area: Statistics
• Runtime relevance: Runtime support
• GPU status: Review if runtime-active
• Procedures/modules found: stat_defs
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

std_dynamic.f90

• Area: SGS
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: std_dynamic, std_dynamic_cuda_sync
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

string_util.f90

• Area: Utility
• Runtime relevance: Setup
• GPU status: Review if runtime-active
• Procedures/modules found: string_util, procedure, procedure, procedure, numtostr_r, numtostr_i, eat_whitespace, uppercase, split_string, count_string_occur, string_concat_a, string_concat_r, string_concat_i, string_concat_ai, string_concat_ar, string_concat_aia, string_concat_ara, string_concat_aiaia, string_concat_arara, string_concat_aiai, string_concat_arar, string_concat_araia, string_concat_arai, string_concat_aiaiai, string_concat_ararar, string_concat_aiaiaia, string_concat_ararara, string_splice_aa, string_splice_ar, string_splice_ai, +12 more
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

test_filtermodule.f90

• Area: Filtering
• Runtime relevance: Runtime hot path
• GPU status: GPU-enabled
• Procedures/modules found: test_filtermodule, require_test_filter_cufft_success, test_filter_cuda_sync, test_filter_cuda_barrier, ensure_test_filter_cuda_plan, ensure_test_filter_cuda_many_plan, apply_test_filter_cuda, apply_test_filter_cuda_3, apply_test_filter_cuda_6, test_filter_init, test_filter, test_filter_3, test_filter_6, test_test_filter, test_test_filter_3, test_test_filter_6
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

time_average.f90

• Area: Statistics
• Runtime relevance: Runtime/I/O
• GPU status: Review if runtime-active
• Procedures/modules found: time_average, init, compute, finalize, checkpoint, write_parallel_cgns, write_null_cgns
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

trees_base_ls.f90

• Area: Level set
• Runtime relevance: Optional setup
• GPU status: Review if runtime-active
• Procedures/modules found: trees_base_ls, cross_product, delta, grid_initialize, grid_nearest_of_pt, grid_of_pt, mag, pt_of_grid
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

trees_global_fmask_ls.f90

• Area: Level set
• Runtime relevance: Optional setup
• GPU status: GPU-enabled
• Procedures/modules found: trees_global_fmask_ls, global_fmask_init, calc_global_fmask_ta, read_global_fmask, write_global_fmask, write_fmt_global_fmask, calc_global_fmask, filter_global_fmask, calc_global_fmask_gpu, filter_global_fmask_gpu, truncgauss_kernel_3d
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

trees_io_ls.f90

• Area: Level set
• Runtime relevance: Optional I/O
• GPU status: Review if runtime-active
• Procedures/modules found: trees_io_ls, draw_cross_section, set_h, set_radius, write_tecplot_zone_hdr, draw_tree_array, read_tree_data, read_ta_data, write_tree_data, write_ta_data
• Retained switches: -
• Developer note: Audit before using in a new active configuration.

trees_pre_ls.f90

• Area: Level set
• Runtime relevance: Optional setup
• GPU status: GPU-enabled
• Procedures/modules found: trees_pre_ls_mod, trees_pre_ls
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

trees_setup_ls.f90

• Area: Level set
• Runtime relevance: Optional setup
• GPU status: GPU-enabled
• Procedures/modules found: trees_setup_ls, sdistfcn_tree_array, dist_plate, dist_circle, dist_circle_b, dist_circle_c, dist_circle_bc, dist_circle_nobc, dist_square, dist_square_b, dist_square_c, dist_square_bc, dist_square_nobc, sdistfcn_branch_square_gpu, fill_tree_array, read_trees_conf, case_n_tree, case_tree, init_tree, cartesian_correction, heightwidth_bbox_br, set_zone_branch
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

tridag_array.f90

• Area: Pressure
• Runtime relevance: Runtime hot path
• GPU status: GPU-enabled
• Procedures/modules found: tridag_transpose_gpu_mod, tridag_array, tridag_array_transpose_thomas_cuda, tridag_array_spike2_cuda, tridag_array_replicated_cuda, tridag_array
• Retained switches: CUDA_VISIBLE_DEVICES, LESGO_PRESS_DIRECT_THOMAS_OUT, LESGO_PRESS_TRANSPOSE_GENERIC, LESGO_PRESS_TRANSPOSE_TIMING, MPICH_GPU_SUPPORT_ENABLED
• Developer note: Pressure math and zero-mode handling are release-sensitive; validate divergence and wall stress after edits.

turbine_indicator.f90

• Area: Turbines
• Runtime relevance: Optional setup/runtime
• GPU status: GPU-enabled
• Procedures/modules found: turbine_indicator, val, init
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

turbines.f90

• Area: Turbines
• Runtime relevance: Optional runtime
• GPU status: GPU-enabled
• Procedures/modules found: turbines, turbines_init, turbines_nodes, turbines_forcing, turbines_finalize, turbines_checkpoint, turbine_vel_init, place_turbines, read_control_files
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

turbines_gpu.f90

• Area: Turbines
• Runtime relevance: Optional runtime
• GPU status: GPU-enabled
• Procedures/modules found: turbines_gpu, turbines_interp_w_to_uv_gpu
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.

types.f90

• Area: Shared types
• Runtime relevance: Shared
• GPU status: CPU acceptable: setup/I/O
• Procedures/modules found: types
• Retained switches: -
• Developer note: CPU is acceptable unless this path becomes repeated inside the timestep.

wallstress.f90

• Area: Wall model
• Runtime relevance: Runtime
• GPU status: GPU-enabled
• Procedures/modules found: wallstress, wallstress_cuda_sync, ws_free_lbc, ws_free_ubc, ws_dns_lbc, ws_dns_ubc, ws_equilibrium_lbc, ws_equilibrium_ubc
• Retained switches: -
• Developer note: Runtime loops have GPU coverage; preserve device-resident data flow.