100%

Macro-Benchmark Validation

Goal

Measurement section (no grind). Validates the grind sections’ compounded effect against the macro anchor (01.4) + parser-perf-v0. A micro-win that does not show up macro is surfaced here.

Work Items

  • Re-run the macro benchmark vs parser-perf-v0; confirm the micro-tuning wins from 03/04/05 COMPOUND into end-to-end frontend throughput improvement.
  • Reconcile micro vs macro deltas; catch any macro regression the micro-benchmarks missed (a micro-win that does not compound, or regresses end-to-end).