Historical MKPP/MICM diagnostic

This page documents a retired diagnostic path. It must not be used to claim an MKPP-versus-MICM speedup or a solver-comparison result.

scripts/compare_micm.py parses a mechanism and constructs an independent SciPy reference. It does not execute MICM, and it deliberately fails when a real named MKPP checkpoint result is unavailable. Its historical throughput and speedup examples were therefore not fair solver-comparison evidence.

Use the governed solver-comparison workflow instead:

python scripts/compare_solvers.py --list-mechanisms
python scripts/compare_solvers.py \
  --validate --campaign benchmarks/solver_comparison/campaigns/cpu-primary-v1.yaml

The governed workflow requires real MICM, KPP (where admitted), and MKPP runners; pinned canonical chemistry; matching resources; scientific admission; and retained JSONL evidence. TS1 remains a MICM/MKPP-only comparison until a KPP binding is explicitly admitted. See Benchmarking MICM, KPP, and MKPP for the publication methodology.

The legacy script may still be useful while developing its independent reference equations, but treat every result as reference-only diagnostics.