Broaden the string-to-Arrow benchmarks - #9503
Conversation
Rework `vortex-arrow`'s string export benchmarks into a matrix over encoding (offset, view, FSST, OnPair, Zstd), structure (flat, dict, chunked), operator (identity, filter, take, slice, mask, zip), and nullability, exported to both Arrow offset and view layouts, plus the same cases appended directly into `VarBinBuilder` and `VarBinViewBuilder`. Add a `string_to_arrow` benchmark in `vortex` that measures the end-to-end file scan producing Arrow offset arrays, so the export is also covered with data laid out by the default compressor rather than by hand. Signed-off-by: Robert Kruszewski <github@robertk.io>
Merging this PR will degrade performance by 10.07%
|
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ❌ | Simulation | compress_fsst[(500, 64, 8)] |
526.2 µs | 585.1 µs | -10.07% |
Tip
Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.
Comparing rk/string-arrow-benches (90c6924) with develop (604cd33)
Footnotes
-
517 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩
1 participant
Add string arrow export to microbenchmarks