Module Reference

Class Hierarchy

Simulator (simulator/simulator.py)
  └── GpuSimulator (gpu_core/gpu_simulator/gpu_simulator.py)

Layer (core/layer/layer.py)
  └── GpuLayer (gpu_core/gpu_layer/gpu_layer.py)

PhaseScreen (core/phase_screens/infinitephasescreen.py)
  ├── PhaseScreenVonKarman
  └── PhaseScreenKolmogorov

File Map

Entry Points

File	Description
gpu_main.py	Primary CLI entry point
main.py	Legacy CPU entry point
simulator_api.py	Programmatic API (run_simulation())
benchmark_runner.py	Standalone benchmark tool

Core Modules

File	Key Classes/Functions
core/utils.py	load_config(), solve_for_A(), calc_r0_profile(), create_output_directory(), FFT utilities
core/geometry/geometry.py	Geometry dataclass -- layer placement, r0/wind profiles
core/layer/layer.py	Layer -- wraps phase screen, manages extrusion and MCF
core/propagation/propagator.py	Propagator -- angular spectrum Fresnel propagation
core/propagation/basic_funcs.py	ft2(), ift2(), circ(), rect(), TF equivalents
core/phase_screens/infinitephasescreen.py	PhaseScreenVonKarman, PhaseScreenKolmogorov -- autoregressive infinite screens
core/phase_screens/PhaseScreenGen.py	PhaseScreenGen -- TF-based Von Karman screen generator (used for Monte Carlo regeneration)

GPU Modules

File	Key Classes/Functions
gpu_core/gpu_simulator/gpu_simulator.py	GpuSimulator -- TF-accelerated simulator with streaming support
gpu_core/gpu_layer/gpu_layer.py	GpuLayer -- TF variable-backed layer
gpu_core/monitoring/system_monitor.py	SimulationMonitor -- system metrics + TensorBoard integration

Analysis Modules

File	Key Classes/Functions
analysis/performance.py	PerformanceTester, PerformanceConfig -- systematic benchmarking
analysis/monte_carlo.py	MonteCarloAnalyzer -- Monte Carlo statistical analysis

Key Methods

GpuSimulator

Method	Decorator	Description
simulate_turb(steps)	@tf.function	Single-shot simulation, all frames in GPU memory
simulate_turb_chunk(steps)	@tf.function(input_signature)	Run a fixed chunk of steps, return tensor. Uses input_signature to avoid retracing
simulate_turb_chunk_complex(steps)	@tf.function(input_signature)	Like simulate_turb_chunk but returns complex Uout directly
simulate_turb_streaming(steps, chunk_size)	Python	Streaming mode with double-buffered I/O. Passes tf.constant to chunk functions
simulate_turb_streaming_complex(steps, chunk_size)	Python	Streaming mode returning complex fields
simulate_turb_no_save(steps)	Python	Throughput test, no frame storage
animate_turb(steps)	Python	Full pipeline: simulate + save + animate + validate
return_first_wvfrnt()	@tf.function	Single propagation step (no time evolution)
MCF_calc(real_num)	@tf.function	Compute Mutual Coherence Function
tf_move_one_tick_all_layers()	@tf.function	Advance all layers by one time step. Uses tf.cast instead of int() to stay in-graph

Propagator

Method	Description
propagate(Uin, PS_arr)	Propagate input field through phase screen array. Returns (Uout, xn, yn)

Geometry

Method	Description
move_one_tick_all_layers()	Advance all layers, return phase screen array
recreate_ps_arr()	Regenerate all phase screens (for Monte Carlo)
show_object_info(logger)	Log geometry parameters

GpuLayer

Method	Decorator	Description
get_new_row()	@tf.function	Generate new row via autoregressive model
add_row()	@tf.function	Prepend new row and trim screen

Utility Functions

core/utils.py

Function	Description
load_config(file_path)	Load JSON config file
solve_for_A(k, w, delta_z, r0)	Solve for Hufnagel-Valley A parameter
calc_r0_profile(...)	Compute layer heights, r0 array, wind profile
calculate_number_of_extrusions(...)	Pixels to extrude per time step
create_output_directory(...)	Create timestamped output directory
ft2(data, delta) / ift2(G, delta_f)	NumPy 2D FFT/IFFT
tf_ft2(g, delta) / tf_ift2(g, delta_f)	TensorFlow 2D FFT/IFFT
corr2_ft(u1, u2, mask, delta)	Fourier-domain cross-correlation