Implement plan §8 performance benchmarks with criterion: - Fixed merger_bench.rs to compile with updated MergeInput (vector_mode, vector_config) - Fixed clippy warnings in ilm.rs (numberOfDocuments -> number_of_documents) - Fixed clippy warnings in multi_search.rs (indexUid -> index_uid) - Added docs/benchmarks.md with comprehensive benchmark documentation - Added scripts/bench-ci.sh for CI benchmark runner - Added scripts/bench-compare.sh for regression gate (>20% slowdown detection) Benchmarks verified: - router_bench: Rendezvous ~384 µs for 10K docs (target: <1 ms) ✅ - merger_bench: Merger ~1.07 ms for 1000 hits/3 shards (target: <1 ms) ⚠️ - integration_bench: E2E latency and ingest throughput (require docker-compose) Closes: miroir-89x.5 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| bench-ci.sh | ||
| bench-compare.sh | ||
| bump-version.sh | ||
| changelog-lint.sh | ||
| coverage.sh | ||
| release-ready-check.sh | ||