iceberg-kitten opened a new pull request, #1637:
URL: https://github.com/apache/arrow-go/pull/1637

   ### Why
   An all-true filter still builds take indices and copies every column.
   
   ### What
   Return the original record batch or table when the mask selects every row.
   
   ### Implementation
   Check the logical bitmap range, including sliced and chunked masks. Retain 
the original input when every slot is valid and true.
   
   ### Benchmark
   
   Filter Int64 record batches and tables with an all-true mask and one 
execution worker.
   
   | Input | Time/op before → after | Change | B/op | Allocs/op |
   | --- | ---: | ---: | ---: | ---: |
   | Record batch, 8 cols × 65,536 rows | 1.03 ms → 201 ns | -99.980% | 
4,545,896 → 0 | 240 → 0 |
   | Record batch, 64 cols × 65,536 rows | 7.57 ms → 200 ns | -99.997% | 
34,472,202 → 0 | 1,865 → 0 |
   | Table, 8 cols × 65,536 rows | 1.07 ms → 198 ns | -99.981% | 4,552,060 → 0 
| 325 → 0 |
   | Table, 64 cols × 65,536 rows | 8.59 ms → 198 ns | -99.998% | 34,514,206 → 
0 | 2,442 → 0 |
   
   Medians of 8 runs per revision on Linux amd64, Intel Core i9-12900HK, Go 
1.26.11, pinned to one efficiency core. Baseline is `main` at `aba35416`. Use 
the same benchmark on both revisions. Copy the PR's benchmark file into the 
baseline checkout before running.
   
   ```sh
   go test ./arrow/compute -run '^$' -bench 
'^BenchmarkFilterAllSelected(RecordBatch|Table)$/^columns=(8|64)$/^rows=65536$' 
-benchmem -benchtime=120ms -count=8 -cpu=1
   ```
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to