Benchmarks

Real measurements, per benchmark type — honestly labelled where a competitor baseline does not yet exist.

commit a40cf2a1·latest reported 7/3/2026·source: benchmark-data branch (dev/bench/data.js)

Each card is a benchmark type. The per-commit series (every metric smaller is better) is recorded by the CI benchmark workflow on a GitHub-hosted runner. Where a real same-box competitor measurement exists, the type page shows a live-computed speedup vs the next-best engine; where it does not, it says so plainly and shows sparq’s absolute numbers. Cross-machine figures are kept separate and clearly labelled — never folded into a speedup.

SPARQL query suites

Recognised public SPARQL benchmarks — LUBM, WatDiv, SP2Bench, BSBM, DBPSB — plus operator micro-benchmarks and the synthetic qlever-style suite.

View benchmarks

SHACL validation

SHACL Core + SHACL-SPARQL validation latency over the LUBM(1) ABox × five hand-authored shape graphs, with same-box pySHACL + Apache Jena baselines.

View benchmarks

geof: spatial functions + R-tree GeoIndex over a fixed ~100k CRS84 point corpus — within / nearest-k / topology compliance.

View benchmarks

Full-Text search

BM25 full-text search via magic predicates over a synthetic 100k-literal corpus, plus index footprint.

View benchmarks

Approximate-nearest-neighbour recall deficits (HNSW / Vamana / PQ) vs exact kNN.

View benchmarks

Reasoning (N3 · RDFS · OWL-RL)

Forward-closure materialization — the Deep Taxonomy depth series — with an EYE cross-engine reference.

View benchmarks

Core (load · dict · memory)

End-to-end load (parse + index build), dictionary footprint, and per-triple store memory.

View benchmarks

Zero-knowledge (commit · trace · circuit)

Commitment, trace-capture and circuit micro-benchmarks for the research-grade ZK estate (commit · canonicalisation · trace · circuit gates). Research-grade only — the v1 verifier is not externally audited, so these are indicative engineering numbers, not an audited cryptographic guarantee.

View benchmarks

Solid / access control (WAC · ACP)

Solid access-control evaluation latency — Web Access Control (WAC) and Access Control Policy (ACP) decision micro-benchmarks.

View benchmarks

HDT archive ingest

HDT (Header-Dictionary-Triples) archive ingest — decode + id-translation throughput when loading .hdt archives into a sparq Graph.

View benchmarks

RDF Stream Processing (continuous queries)

RDF Stream Processing — continuous-query and windowed-evaluation micro-benchmarks over streaming RDF.

View benchmarks

GenAI (similarity · introspection)

Generative-AI-adjacent micro-benchmarks — embedding similarity and graph introspection helpers.

View benchmarks

GPU kernels (experimental)

Experimental GPU-kernel micro-benchmarks — an exploratory surface, reported as-is.

View benchmarks