Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Benchmark Harness

The committed benchmark reports are the current reference for performance comparisons.

Current Reference Artifacts

  • perf_baselines/current_max/2026-04-05/solver_benchmarks.txt
  • perf_baselines/current_max/2026-04-05/runtime_benchmarks.txt
  • perf_baselines/current_max/2026-04-05/manifest.json
  • perf_baselines/current_max/2026-04-05/README.md

Current Coverage

The committed baseline wave still covers both algorithm-heavy and runtime-heavy paths:

  • solver scenarios such as dense rows, wrapped cards, and grid dashboards
  • runtime scenarios such as root capsules, idle-after-settle, render-only churn, text measurement, picking traversal, widget-heavy panels, and World3d panels

How To Use It Now

  • read the committed .txt reports for the raw scenario output
  • read manifest.json for the structured metadata
  • use scripts/render_benchmark_report.py if you want a rendered report from the stored data

Status Of run_perf_baselines.sh

./scripts/run_perf_baselines.sh currently points maintainers to the committed baseline data used by this branch.