add-benchmarks
Guidelines for designing or extending DataFusion benchmarks.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-15
- Stars
- ★ 9.6k
- Forks
- 2.6k
01 前言 Front Matter
- name
- add-benchmarks
- description
- Guidelines for designing or extending DataFusion benchmarks.
02 资源文件 82
- 参考文档canvas-fonts/ArsenalSC-OFL.txt4.3 KB
- 资产文件canvas-fonts/ArsenalSC-Regular.ttf162.0 KB
- 资产文件canvas-fonts/BigShoulders-Bold.ttf92.3 KB
- 参考文档canvas-fonts/BigShoulders-OFL.txt4.3 KB
- 资产文件canvas-fonts/BigShoulders-Regular.ttf92.2 KB
- 参考文档canvas-fonts/Boldonse-OFL.txt4.3 KB
- 资产文件canvas-fonts/Boldonse-Regular.ttf75.4 KB
- 资产文件canvas-fonts/BricolageGrotesque-Bold.ttf88.8 KB
- 参考文档canvas-fonts/BricolageGrotesque-OFL.txt4.3 KB
- 资产文件canvas-fonts/BricolageGrotesque-Regular.ttf88.8 KB
- 资产文件canvas-fonts/CrimsonPro-Bold.ttf104.8 KB
- 资产文件canvas-fonts/CrimsonPro-Italic.ttf106.3 KB
- 参考文档canvas-fonts/CrimsonPro-OFL.txt4.3 KB
- 资产文件canvas-fonts/CrimsonPro-Regular.ttf104.2 KB
- 参考文档canvas-fonts/DMMono-OFL.txt4.3 KB
- 资产文件canvas-fonts/DMMono-Regular.ttf47.7 KB
- 参考文档canvas-fonts/EricaOne-OFL.txt4.3 KB
- 资产文件canvas-fonts/EricaOne-Regular.ttf24.3 KB
- 资产文件canvas-fonts/GeistMono-Bold.ttf76.5 KB
- 参考文档canvas-fonts/GeistMono-OFL.txt4.3 KB
- 资产文件canvas-fonts/GeistMono-Regular.ttf76.4 KB
- 参考文档canvas-fonts/Gloock-OFL.txt4.3 KB
- 资产文件canvas-fonts/Gloock-Regular.ttf92.9 KB
- 资产文件canvas-fonts/IBMPlexMono-Bold.ttf132.8 KB
- 参考文档canvas-fonts/IBMPlexMono-OFL.txt4.3 KB
- 资产文件canvas-fonts/IBMPlexMono-Regular.ttf130.7 KB
- 资产文件canvas-fonts/IBMPlexSerif-Bold.ttf157.2 KB
- 资产文件canvas-fonts/IBMPlexSerif-BoldItalic.ttf165.9 KB
- 资产文件canvas-fonts/IBMPlexSerif-Italic.ttf166.0 KB
- 资产文件canvas-fonts/IBMPlexSerif-Regular.ttf156.6 KB
- 资产文件canvas-fonts/InstrumentSans-Bold.ttf66.5 KB
- 资产文件canvas-fonts/InstrumentSans-BoldItalic.ttf68.4 KB
- 资产文件canvas-fonts/InstrumentSans-Italic.ttf68.3 KB
- 参考文档canvas-fonts/InstrumentSans-OFL.txt4.3 KB
- 资产文件canvas-fonts/InstrumentSans-Regular.ttf66.4 KB
- 资产文件canvas-fonts/InstrumentSerif-Italic.ttf69.2 KB
- 资产文件canvas-fonts/InstrumentSerif-Regular.ttf67.7 KB
- 参考文档canvas-fonts/Italiana-OFL.txt4.3 KB
- 资产文件canvas-fonts/Italiana-Regular.ttf26.5 KB
- 资产文件canvas-fonts/JetBrainsMono-Bold.ttf112.1 KB
- 参考文档canvas-fonts/JetBrainsMono-OFL.txt4.3 KB
- 资产文件canvas-fonts/JetBrainsMono-Regular.ttf112.2 KB
- 资产文件canvas-fonts/Jura-Light.ttf150.7 KB
- 资产文件canvas-fonts/Jura-Medium.ttf150.9 KB
- 参考文档canvas-fonts/Jura-OFL.txt4.3 KB
- 参考文档canvas-fonts/LibreBaskerville-OFL.txt4.3 KB
- 资产文件canvas-fonts/LibreBaskerville-Regular.ttf144.1 KB
- 资产文件canvas-fonts/Lora-Bold.ttf130.7 KB
- 资产文件canvas-fonts/Lora-BoldItalic.ttf137.0 KB
- 资产文件canvas-fonts/Lora-Italic.ttf136.1 KB
- 参考文档canvas-fonts/Lora-OFL.txt4.3 KB
- 资产文件canvas-fonts/Lora-Regular.ttf130.8 KB
- 资产文件canvas-fonts/NationalPark-Bold.ttf77.4 KB
- 参考文档canvas-fonts/NationalPark-OFL.txt4.3 KB
- 资产文件canvas-fonts/NationalPark-Regular.ttf74.6 KB
- 参考文档canvas-fonts/NothingYouCouldDo-OFL.txt4.3 KB
- 资产文件canvas-fonts/NothingYouCouldDo-Regular.ttf31.3 KB
- 资产文件canvas-fonts/Outfit-Bold.ttf54.1 KB
- 参考文档canvas-fonts/Outfit-OFL.txt4.3 KB
- 资产文件canvas-fonts/Outfit-Regular.ttf53.6 KB
- 资产文件canvas-fonts/PixelifySans-Medium.ttf49.9 KB
- 参考文档canvas-fonts/PixelifySans-OFL.txt4.3 KB
- 参考文档canvas-fonts/PoiretOne-OFL.txt4.3 KB
- 资产文件canvas-fonts/PoiretOne-Regular.ttf44.2 KB
- 资产文件canvas-fonts/RedHatMono-Bold.ttf33.6 KB
- 参考文档canvas-fonts/RedHatMono-OFL.txt4.3 KB
- 资产文件canvas-fonts/RedHatMono-Regular.ttf33.7 KB
- 参考文档canvas-fonts/Silkscreen-OFL.txt4.3 KB
- 资产文件canvas-fonts/Silkscreen-Regular.ttf31.2 KB
- 资产文件canvas-fonts/SmoochSans-Medium.ttf58.3 KB
- 参考文档canvas-fonts/SmoochSans-OFL.txt4.3 KB
- 资产文件canvas-fonts/Tektur-Medium.ttf74.5 KB
- 参考文档canvas-fonts/Tektur-OFL.txt4.3 KB
- 资产文件canvas-fonts/Tektur-Regular.ttf73.8 KB
- 资产文件canvas-fonts/WorkSans-Bold.ttf186.8 KB
- 资产文件canvas-fonts/WorkSans-BoldItalic.ttf171.7 KB
- 资产文件canvas-fonts/WorkSans-Italic.ttf170.2 KB
- 参考文档canvas-fonts/WorkSans-OFL.txt4.3 KB
- 资产文件canvas-fonts/WorkSans-Regular.ttf184.5 KB
- 参考文档canvas-fonts/YoungSerif-OFL.txt4.3 KB
- 资产文件canvas-fonts/YoungSerif-Regular.ttf102.7 KB
- 许可证LICENSE.txt11.1 KB
03 SKILL.md 正文
Benchmark Design
Follow these guidelines when designing or adding a new benchmark.
Design principles
Use the highest-level interface
Use the highest-level interface possible, preferably SQL, while keeping the work around the operator being measured as cheap as possible. When benchmarking a function, exercise its evaluation path instead of benchmarking internal utility functions in isolation.
This makes benchmarks easier to maintain and helps assess how much an optimization matters to end-to-end runtime. It also helps avoid spending time optimizing code that accounts for only a small fraction of the total runtime.
Practical criteria for choosing SQL or Rust benchmarks: Try implementing the benchmark in SQL first. If a Rust microbenchmark still seems like a better fit, use Criterion.
Vary the key workload axes
First identify the key axes to vary. For example, for a join benchmark:
- Input size on each side.
- Join-filter selectivity.
- ...
Then choose benchmark cases that exercise representative variations. Full combinatorial coverage is unnecessary; focus on typical workloads that reflect real use cases.
When adding benchmark queries, simply tag each query with the decision made for every axis. For example, for a join benchmark with input sizes/filter selectivity to tune:
-- Q1: Small left input, large right input; 0.1% of pairs match.
SELECT *
FROM generate_series(1, 100) AS l
JOIN generate_series(1, 100000) AS r
ON (l.value + r.value) % 1000 = 0;
-- Q2: Medium inputs on both sides; no filter.
SELECT *
FROM generate_series(1, 1000) AS l
CROSS JOIN generate_series(1, 1000) AS r;
SQL benchmarks
For implementation details, see the SQL benchmark README.
-
Keep other operators cheap.
When a SQL benchmark targets a specific operator, keep the work done by other operators as lightweight as possible. For example, use a data source such as
generate_series()instead of a Parquet scan so scan overhead does not dominate the measurement. See thenljbenchmark for examples. -
Integrate with the top-level benchmark script.
Ensure the benchmark can be prepared and run through
bench.sh:# Run from the benchmarks directory. # Generate any required dataset. ./bench.sh data new_bench # Run the benchmark. ./bench.sh run new_bench -
Keep query runtimes practical.
Tune the workload so each query takes roughly a few seconds per execution. This helps reduce the relative impact of timing noise while keeping the suite practical to run.
04 同仓库更多 Skills
pr_review
Review Apache DataFusion pull requests following the project's PR review guide. Use whenever asked to review a DataFusion PR or PR URL, and whenever creating a PR, to check the changes against the same criteria before submitting.
datafusion-ffi
Patterns and review checklist for the `datafusion-ffi` crate. Use whenever the user adds, edits, or reviews code under `datafusion/ffi/` — new `FFI_X` wrappers, `ForeignX` implementations, codec changes, or expanding an existing wrapper to cover more of a trait's surface. Also use when reviewing PRs that touch this crate.