compilation overhead slows us down

Author: codegithubka

docs/kimon_prompts.md

@@ -51,3 +51,4 @@
 
 1. I am considering using numba for optimization and faster runs in the HPC. Outline an implementation plan, practical considerations, and feasibility within a logical timeframe.
 
+2. Walk me through modifying tg

scripts/macro_benchmark.py

@@ -0,0 +1,62 @@
+import time
+import statistics
+import logging
+from pathlib import Path
+import sys
+import numpy as np
+
+# Ensure we can find our modules
+sys.path.insert(0, str(Path(__file__).parent.parent))
+
+import scripts.pp_analysis as ppa
+from scripts.pp_analysis import Config
+
+def run_stress_benchmark():
+    cfg = Config()
+    # INCREASE WORKLOAD: 150x150 grid and 1000 total steps
+    # This ensures the 'race' is long enough to overcome the 'warmup'
+    cfg.default_grid = 150       
+    cfg.n_prey_birth = 2        # 2x2 sweep is enough to see scaling
+    cfg.n_prey_death = 2
+    cfg.n_replicates = 3        
+    cfg.warmup_steps = 400
+    cfg.measurement_steps = 600
+    cfg.n_jobs = 4              
+    
+    output_dir = Path("results_stress")
+    output_dir.mkdir(exist_ok=True)
+    
+    logging.getLogger('scripts.pp_analysis').setLevel(logging.WARNING)
+    
+    print("="*60)
+    print("STRESS BENCHMARK: SCALING TO PRODUCTION SIZES")
+    print("="*60)
+
+    # --- MODE A: NUMBA ENABLED ---
+    print("[*] Running WITH Numba (Compiling + High-Speed Execution)...")
+    ppa.USE_NUMBA = True
+    t0 = time.perf_counter()
+    ppa.run_2d_sweep(cfg, output_dir, logging.getLogger("test"))
+    numba_total = time.perf_counter() - t0
+    print(f"    Total Time (Numba): {numba_total:.2f}s")
+
+    # --- MODE B: NUMBA DISABLED ---
+    print("[*] Running WITHOUT Numba (Pure Python)...")
+    ppa.USE_NUMBA = False
+    t0 = time.perf_counter()
+    ppa.run_2d_sweep(cfg, output_dir, logging.getLogger("test"))
+    python_total = time.perf_counter() - t0
+    print(f"    Total Time (Python): {python_total:.2f}s")
+
+    # --- FINAL REPORT ---
+    speedup = python_total / numba_total
+    print("\n" + "="*60)
+    print("STRESS BENCHMARK SUMMARY")
+    print("="*60)
+    print(f"Real-World Workflow Speedup: {speedup:.2f}x")
+    print(f"Estimated time for 1000 sims in Python: {(python_total/12)*1000/3600:.1f} hours")
+    print(f"Estimated time for 1000 sims in Numba:  {(numba_total/12)*1000/3600:.1f} hours")
+    print("="*60)
+
+if __name__ == "__main__":
+    run_stress_benchmark()