Performance

Benchmark reports describe their recorded workloads, hardware, and software builds. Query latency and end-to-end throughput measure different parts of a run; compare matching conditions.

Benchmark reports

Report	Scope
Database comparison	Latest recorded Northwind sweep across NornicDB, Neo4j, FalkorDB, Memgraph, and LadybugDB.
Shortest-path traversal	500,000 nodes and 3,969,418 edges; seed cost and latency across hop depths.
Parser scaling	Input-size growth, allocations, bytes, latency, and GC for the native and ANTLR parsers.
Northwind parser modes	Native and ANTLR query latency over the same seeded graph.
Neo4j comparison	Query-level results, ingestion, storage, power, and measurement definitions.
Retrieval quality	BEIR SciFact recall, ranking scores, and reproduction protocol.
Hybrid queries	Combined retrieval and graph-query workloads.

Report directories use stable topic names. Build identifiers, versions, and dates remain inside the reports and raw artifacts so the measurements can be reproduced.

Profiling and tuning

• Search methodology
• Hot-path query cookbook
• HTTP optimization
• Profiling with pprof
• Single-request measurements