← Back to Homepage

TRANSPARENT BENCHMARK HARNESS

Browser/WASM edition (honest: real Oxigraph WASM from CDN, full source visible). Native dual-mode (full engine, Lazy SuperBlocks + LZ4 + WebRTC telemetry, canonical LLM JSON):
cargo run --release -p qualia-cli -- bench --suite full (or benchmark alias). Pairs with ../benchmark_visualizer.html and the root benchmark.html.

live in-browser · real performance.now() · full source visible · oxigraph wasm attempted

⚠️ Full Disclosure
Numbers are produced by real JavaScript executing now. No hardcoded values exist. The harness attempts to load Oxigraph WASM directly from the npm CDN (esm.sh). If it loads, it runs real SPARQL queries. If it fails (CORS, CDN downtime, module API mismatch), the harness clearly marks those cells as LOAD FAILED and falls back to a BTreeMap-equivalent proxy — labelled as such.

Native Criterion benchmarks (reproducible, machine-local) live in the repo at benches/qualia_benchmark.rs. Run them with cargo bench --package qualia-core-db.
QualiaDB JS Oxigraph WASM BTreeMap Proxy
📊 Rust-Native Criterion Results (reproducible)
The browser JS is not a substitute for native benchmarks. To reproduce on your own hardware:

git clone https://github.com/mediaprophet/qualiaDB.git
cargo bench --package qualia-core-db
# HTML report → target/criterion/report/index.html

The bench compares QualiaDB's FNV HashMap against BTreeMap (Oxigraph class) and std HashMap (SurrealDB class) across point lookup, 2-hop traversal, predicate filter, and cold ingestion — all on the same 10k deterministic dataset.

EXECUTION LOG

FULL BENCHMARK SOURCE (this file)

The script tag below is the literal function that produced the numbers above. 


        

    


    

        Limitations:

        • Oxigraph WASM is loaded from esm.sh/oxigraph. If the CDN changes its module
          format, loading will fail and cells will show LOAD FAILED.

        • performance.memory is Chromium-only. Firefox/Safari will show N/A.

        • Browser JIT warmup means first-run numbers differ from steady-state. We run 50 iterations
          and discard no warm-up — this is conservative (makes qualiaDB look worse, not better).

        • SurrealDB WASM target was experimental at time of writing and has no stable CDN bundle.
          BTreeMap proxy is used in its place and clearly labelled.