vectojs-performancelisted

# VectoJS Performance Use this skill when VectoJS feels slow or when designing workloads that may exceed DOM or Canvas 2D limits. ## Diagnosis workflow 1. Separate render cost, layout/text cost, application compute, event/hit-test cost, and DOM semantic-sync cost. 2. Reproduce with a fixed workload and record entity count, text length, backend, viewport, DPR, hardware, and browser. 3. Check whether CPU compute dominates before changing renderer backends. 4. Reduce unnecessary work: on-demand rendering, viewport culling, virtualization, prepared text, and dirty-region discipline. 5. Choose GPU paths only for matching workloads: WebGL point batching for large points/rects, WebGPU particles for compute-driven simulations. 6. Verify with the same benchmark after each change. Read `references/performance-checklist.md` for concrete probes and fixes. For token streams / chat / log tails, read `references/streaming-recipes.md` — the per-frame batching pattern there is the single highest-leverage streaming fix and is NOT optional for LLM-speed streams. ## Decision matrix | Symptom | Likely area | First fix