#Unified Benchmark Matrix

One place to look at every architecture on the same host and ladder — IFR styles, VDOM vs Vapor, and React vs Vue — without mixing metric scales.

How to read storms: Select = point update (selection / a few classes per tick). Update = batch update (many rows touched each pass). Don’t compare those two curves as if they were the same workload.

#What this page is

The three older campaigns answered different questions:

They shared 1k→30k labels but not a metric scale. This report puts every cell in one schema and leads with decision-first conclusions (takeaway → so what → verify).

Sibling pages keep only what this matrix does not cover:

#How to read it

1. Conclusions — digestible product guidance; numbers are for checking.
2. Storm table / charts — point (select) vs batch (update) latency.
3. FCP table / charts — content-probe first-frame ladder (separate instrument!).
4. Optimization-flag matrix — every legal combination in V4 notation render [+b[:d|c|e]] [+ifr[:c|e]] (baseline = per-node; +b = block templates; +b:e = engine staging, N/A on web and omitted; +ifr:e = engine-painted first frame only), measured for create/update with per-flag marginal factor deltas and a full flag legend on the page. Formalization: GRAPH-ENG-REPORT.md §1.3c.

#Environment rules

• lynx-web — primary product scale (this report).
• bare-dom / node-jitless — different instruments; never ratio across them.

#Reproduce

pnpm --filter vue-lynx-benchmark bench:unified
pnpm --filter vue-lynx-benchmark bench:synthesize
pnpm --filter vue-lynx-benchmark bench:report

Design notes: packages/benchmark/UNIFIED.md · raw analysis: ANALYSIS.md.