Complete experimental infrastructure supporting:Empirical Validation of Shadow-Price-Guided Inlining in MIR DOI: 10.5281/zenodo.18828679 Source code and analysis scripts for the empirical validation of the compiler-as-NUM (Network Utility Maximization) hypothesis using the MIR lightweight JIT compiler. The repository implements a three-phase experimental pipeline: 1. Synthetic profiling — generates 750 function-call scenarios across 50 functions × 5 sizes × 3 optimization levels 2. Shadow-price-guided mutation — applies dual decomposition to adjust inlining thresholds via shadow prices (λ) 3. Wall-clock benchmarking — measures outcomes across four conditions (baseline, uniform, skewed, perturbed) Key results: Spearman ρ = 0.71 (p < 10⁻¹¹⁴), Cohen's d = 2.00 for hot vs cold functions, 98.4% perturbation robustness. Shadow-price optimization matches inline-all performance with half the mutations. Language: C (MIR patches), Python (analysis scripts).
compiler optimization · JIT compilation · network utility maximization · dual decomposition · empirical validation · statistical hypothesis testing · shadow pricing