You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "haohuaijin (via GitHub)" <gi...@apache.org> on 2026/10/17 04:16:08 UTC
[PR] perf(parquet): trim bitmap-backed row selections using 64-bit words [arrow-rs]
haohuaijin opened a new pull request, #11639:
URL: https://github.com/apache/arrow-rs/pull/11639
# Which issue does this PR close?
N/A
# Rationale for this change
Trimming a bitmap-backed row selection scans trailing unset bits byte by byte. Long tails of skipped rows add unnecessary selection setup overhead.
# What changes are included in this PR?
Scan backwards through 64-bit words using `BitChunks`, checking the masked remainder first. Use `leading_zeros` to locate the final selected row.
This reuses existing handling of bit offsets and padding while preserving the fast path for selections that need no trimming.
The implementation and tests were prepared with AI assistance.
# Are these changes tested?
Add deterministic tests covering word boundaries, long zero tails, non-zero bit offsets, and set bits outside the logical slice. Existing randomized trim equivalence tests also pass.
# Are there any significant user-facing changes?
No.
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): trim bitmap-backed row selections using 64-bit words [arrow-rs]
Posted by "Jefffrey (via GitHub)" <gi...@apache.org>.
Jefffrey commented on PR #11639:
URL: https://github.com/apache/arrow-rs/pull/11639#issuecomment-6231933745
thanks @haohuaijin
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): trim bitmap-backed row selections using 64-bit words [arrow-rs]
Posted by "haohuaijin (via GitHub)" <gi...@apache.org>.
haohuaijin commented on PR #11639:
URL: https://github.com/apache/arrow-rs/pull/11639#issuecomment-6232206051
Thanks @Jefffrey
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): trim bitmap-backed row selections using 64-bit words [arrow-rs]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #11639:
URL: https://github.com/apache/arrow-rs/pull/11639#issuecomment-6229841532
🤖 Benchmark completed (GKE) | [trigger](https://github.com/apache/arrow-rs/pull/11639#issuecomment-6229581773)
**Target:** row_selector_boolean_buffer
Comparing hj/parquet-trim-word-scan (d76db22eb4a8d2429b06ac70233f472d53ca8464) to 6b8df9c (merge-base) [diff](https://github.com/apache/arrow-rs/compare/6b8df9cd27a6b65a8027c83cf73c37a1015ae0c8..d76db22eb4a8d2429b06ac70233f472d53ca8464)
<details><summary>Run configuration</summary>
```yaml
run benchmark row_selector_boolean_buffer
shards: 1
```
</details>
<details><summary>Details</summary>
```
group base changed
----- ---- -------
row_selector_boolean_buffer/construction/from_boolean_buffer/clustered/select01 1.00 23.4±0.21ns ? ?/sec 1.00 23.6±0.25ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/clustered/select10 1.00 23.2±0.13ns ? ?/sec 1.01 23.5±0.10ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/clustered/select33 1.01 23.5±0.19ns ? ?/sec 1.00 23.2±0.09ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/clustered/select80 1.01 23.6±0.12ns ? ?/sec 1.00 23.5±0.10ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/fragmented/select01 1.01 23.5±0.19ns ? ?/sec 1.00 23.2±0.14ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/fragmented/select10 1.00 23.2±0.13ns ? ?/sec 1.01 23.5±0.10ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/fragmented/select33 1.00 23.2±0.09ns ? ?/sec 1.02 23.6±0.09ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/fragmented/select80 1.02 23.6±0.11ns ? ?/sec 1.00 23.2±0.13ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/random/select01 1.00 23.3±0.14ns ? ?/sec 1.01 23.5±0.21ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/random/select10 1.00 23.2±0.13ns ? ?/sec 1.01 23.6±0.22ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/random/select33 1.00 23.5±0.12ns ? ?/sec 1.00 23.6±0.25ns ? ?/sec
row_selector_boolean_buffer/construction/from_boolean_buffer/random/select80 1.02 23.7±1.40ns ? ?/sec 1.00 23.2±0.14ns ? ?/sec
row_selector_boolean_buffer/construction/from_filters/clustered/select01 1.08 51.2±0.54µs ? ?/sec 1.00 47.3±0.05µs ? ?/sec
row_selector_boolean_buffer/construction/from_filters/clustered/select10 1.00 47.9±0.02µs ? ?/sec 1.00 47.8±0.02µs ? ?/sec
row_selector_boolean_buffer/construction/from_filters/clustered/select33 1.00 49.1±0.03µs ? ?/sec 1.00 49.1±0.03µs ? ?/sec
row_selector_boolean_buffer/construction/from_filters/clustered/select80 1.00 48.3±0.01µs ? ?/sec 1.00 48.4±0.02µs ? ?/sec
row_selector_boolean_buffer/construction/from_filters/fragmented/select01 1.00 146.1±1.56µs ? ?/sec 1.03 150.0±5.27µs ? ?/sec
row_selector_boolean_buffer/construction/from_filters/fragmented/select10 1.00 1514.6±18.31µs ? ?/sec 1.01 1523.5±21.50µs ? ?/sec
row_selector_boolean_buffer/construction/from_filters/fragmented/select33 1.01 14.7±0.23ms ? ?/sec 1.00 14.5±0.18ms ? ?/sec
row_selector_boolean_buffer/construction/from_filters/fragmented/select80 1.00 8.7±0.44ms ? ?/sec 1.00 8.7±0.27ms ? ?/sec
row_selector_boolean_buffer/construction/from_filters/random/select01 1.00 303.6±3.05µs ? ?/sec 1.02 310.3±2.80µs ? ?/sec
row_selector_boolean_buffer/construction/from_filters/random/select10 1.00 1771.9±25.41µs ? ?/sec 1.01 1781.5±30.04µs ? ?/sec
row_selector_boolean_buffer/construction/from_filters/random/select33 1.01 10.3±0.23ms ? ?/sec 1.00 10.1±0.23ms ? ?/sec
row_selector_boolean_buffer/construction/from_filters/random/select80 1.00 2.9±0.05ms ? ?/sec 1.00 2.9±0.06ms ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/clustered/select01 1.36 244.9±1.31µs ? ?/sec 1.00 179.8±1.51µs ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/clustered/select10 1.02 513.4±7.00µs ? ?/sec 1.00 502.7±5.35µs ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/clustered/select33 1.03 1093.4±16.16µs ? ?/sec 1.00 1059.2±13.87µs ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/clustered/select80 1.06 2.3±0.06ms ? ?/sec 1.00 2.1±0.04ms ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/fragmented/select01 1.02 2.0±0.04ms ? ?/sec 1.00 1979.2±26.23µs ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/fragmented/select10 1.01 1770.7±31.62µs ? ?/sec 1.00 1746.2±35.77µs ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/fragmented/select33 1.01 3.1±0.06ms ? ?/sec 1.00 3.0±0.08ms ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/fragmented/select80 1.00 5.3±0.09ms ? ?/sec 1.00 5.4±0.09ms ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/random/select01 1.02 2.4±0.05ms ? ?/sec 1.00 2.4±0.03ms ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/random/select10 1.01 2.2±0.05ms ? ?/sec 1.00 2.2±0.05ms ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/random/select33 1.02 3.5±0.10ms ? ?/sec 1.00 3.4±0.12ms ? ?/sec
row_selector_boolean_buffer/reader/from_boolean_buffer/random/select80 1.00 6.0±0.13ms ? ?/sec 1.00 6.0±0.13ms ? ?/sec
row_selector_boolean_buffer/reader/from_filters/clustered/select01 1.01 138.8±1.75µs ? ?/sec 1.00 137.5±0.94µs ? ?/sec
row_selector_boolean_buffer/reader/from_filters/clustered/select10 1.02 461.5±7.13µs ? ?/sec 1.00 454.6±7.68µs ? ?/sec
row_selector_boolean_buffer/reader/from_filters/clustered/select33 1.04 1042.7±13.62µs ? ?/sec 1.00 999.7±15.02µs ? ?/sec
row_selector_boolean_buffer/reader/from_filters/clustered/select80 1.06 2.2±0.06ms ? ?/sec 1.00 2.0±0.03ms ? ?/sec
row_selector_boolean_buffer/reader/from_filters/fragmented/select01 1.01 1970.2±24.68µs ? ?/sec 1.00 1945.9±13.77µs ? ?/sec
row_selector_boolean_buffer/reader/from_filters/fragmented/select10 1.00 4.6±0.23ms ? ?/sec 1.04 4.8±0.20ms ? ?/sec
row_selector_boolean_buffer/reader/from_filters/fragmented/select33 1.00 22.1±0.56ms ? ?/sec 1.00 22.2±0.46ms ? ?/sec
row_selector_boolean_buffer/reader/from_filters/fragmented/select80 1.01 17.1±0.35ms ? ?/sec 1.00 16.9±0.56ms ? ?/sec
row_selector_boolean_buffer/reader/from_filters/random/select01 1.00 2.3±0.03ms ? ?/sec 1.00 2.3±0.04ms ? ?/sec
row_selector_boolean_buffer/reader/from_filters/random/select10 1.00 5.3±0.16ms ? ?/sec 1.00 5.3±0.11ms ? ?/sec
row_selector_boolean_buffer/reader/from_filters/random/select33 1.03 17.2±0.21ms ? ?/sec 1.00 16.6±0.48ms ? ?/sec
row_selector_boolean_buffer/reader/from_filters/random/select80 1.01 12.6±0.11ms ? ?/sec 1.00 12.5±0.32ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/clustered/select01 1.50 439.1±7.19µs ? ?/sec 1.00 293.6±5.56µs ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/clustered/select10 1.01 782.9±16.32µs ? ?/sec 1.00 778.9±10.48µs ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/clustered/select33 1.00 1653.8±48.94µs ? ?/sec 1.01 1670.9±44.29µs ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/clustered/select80 1.01 2.8±0.05ms ? ?/sec 1.00 2.8±0.08ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/fragmented/select01 1.00 2.6±0.05ms ? ?/sec 1.02 2.6±0.06ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/fragmented/select10 1.00 2.3±0.07ms ? ?/sec 1.01 2.3±0.07ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/fragmented/select33 1.00 3.5±0.08ms ? ?/sec 1.01 3.5±0.09ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/fragmented/select80 1.00 5.8±0.06ms ? ?/sec 1.00 5.8±0.06ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/random/select01 1.00 3.0±0.05ms ? ?/sec 1.01 3.0±0.05ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/random/select10 1.00 2.7±0.05ms ? ?/sec 1.02 2.8±0.04ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/random/select33 1.02 4.0±0.07ms ? ?/sec 1.00 3.9±0.08ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/random/select80 1.00 6.5±0.09ms ? ?/sec 1.01 6.5±0.06ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/clustered/select01 1.00 207.2±3.65µs ? ?/sec 1.01 210.2±4.70µs ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/clustered/select10 1.00 691.8±8.62µs ? ?/sec 1.00 689.6±11.05µs ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/clustered/select33 1.01 1590.1±45.19µs ? ?/sec 1.00 1582.0±48.12µs ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/clustered/select80 1.02 2.8±0.05ms ? ?/sec 1.00 2.7±0.07ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/fragmented/select01 1.03 2.8±0.06ms ? ?/sec 1.00 2.8±0.07ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/fragmented/select10 1.00 9.2±0.19ms ? ?/sec 1.01 9.3±0.24ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/fragmented/select33 1.00 36.5±0.95ms ? ?/sec 1.01 36.9±0.88ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/fragmented/select80 1.03 25.7±0.52ms ? ?/sec 1.00 24.8±0.59ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/random/select01 1.00 3.1±0.07ms ? ?/sec 1.01 3.1±0.08ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/random/select10 1.00 9.2±0.13ms ? ?/sec 1.01 9.3±0.23ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/random/select33 1.00 26.6±0.63ms ? ?/sec 1.00 26.7±0.61ms ? ?/sec
row_selector_boolean_buffer/reader_async/from_filters/random/select80 1.00 19.2±0.39ms ? ?/sec 1.00 19.2±0.37ms ? ?/sec
```
</details>
<details><summary>Per-runner information</summary>
<details><summary>row_selector_boolean_buffer — shard 1/1</summary>
**Node:** gk3-benchmark-cluster-nap-lfpgw1us-1a51caa6-854q
**Instance:** c4a-highmem-16 (12 vCPU / 65 GiB)
**uname:**
```text
Linux bench-c6229581773-3531-rn4pf 6.12.94+ #1 SMP Fri Aug 21 08:00:16 UTC 2026 aarch64 GNU/Linux
```
**BENCH_COMMAND:**
```sh
cargo bench --features=arrow,async,test_common,experimental,object_store --bench row_selector_boolean_buffer
```
<details><summary>CPU Details (lscpu)</summary>
```text
Architecture: aarch64
CPU op-mode(s): 64-bit
Byte Order: Little Endian
CPU(s): 16
On-line CPU(s) list: 0-15
Vendor ID: ARM
Model name: Neoverse-V2
Model: 1
Thread(s) per core: 1
Core(s) per cluster: 16
Socket(s): -
Cluster(s): 1
Stepping: r0p1
BogoMIPS: 2000.00
Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache: 1 MiB (16 instances)
L1i cache: 1 MiB (16 instances)
L2 cache: 32 MiB (16 instances)
L3 cache: 80 MiB (1 instance)
NUMA node(s): 1
NUMA node0 CPU(s): 0-15
Vulnerability Gather data sampling: Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit: Not affected
Vulnerability L1tf: Not affected
Vulnerability Mds: Not affected
Vulnerability Meltdown: Not affected
Vulnerability Mmio stale data: Not affected
Vulnerability Reg file data sampling: Not affected
Vulnerability Retbleed: Not affected
Vulnerability Spec rstack overflow: Not affected
Vulnerability Spec store bypass: Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1: Mitigation; __user pointer sanitization
Vulnerability Spectre v2: Mitigation; CSV2, BHB
Vulnerability Srbds: Not affected
Vulnerability Tsa: Not affected
Vulnerability Tsx async abort: Not affected
Vulnerability Vmscape: Not affected
```
</details>
<details><summary>Resource Usage</summary>
**base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 710.2s |
| Peak memory | 108.6 MiB |
| Avg memory | 55.0 MiB |
| CPU user | 670.8s |
| CPU sys | 34.8s |
| Peak spill | 0 B |
**branch**
| Metric | Value |
|--------|-------|
| Wall time | 720.2s |
| Peak memory | 108.6 MiB |
| Avg memory | 55.0 MiB |
| CPU user | 681.8s |
| CPU sys | 34.1s |
| Peak spill | 0 B |
</details>
</details>
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
<!-- benchmark-request:6229581773:finish:4bdc59bc19d7e814b1a281d05ba31ace -->
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): trim bitmap-backed row selections using 64-bit words [arrow-rs]
Posted by "Jefffrey (via GitHub)" <gi...@apache.org>.
Jefffrey merged PR #11639:
URL: https://github.com/apache/arrow-rs/pull/11639
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): trim bitmap-backed row selections using 64-bit words [arrow-rs]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #11639:
URL: https://github.com/apache/arrow-rs/pull/11639#issuecomment-6229582588
🤖 Benchmark starting (GKE) | [trigger](https://github.com/apache/arrow-rs/pull/11639#issuecomment-6229581773)
**Target:** row_selector_boolean_buffer
**Sharding factor: 1** (1 workers).
Comparing hj/parquet-trim-word-scan (d76db22eb4a8d2429b06ac70233f472d53ca8464) to 6b8df9c (merge-base) [diff](https://github.com/apache/arrow-rs/compare/6b8df9cd27a6b65a8027c83cf73c37a1015ae0c8..d76db22eb4a8d2429b06ac70233f472d53ca8464)
<details><summary>Run configuration</summary>
```yaml
run benchmark row_selector_boolean_buffer
shards: 1
```
</details>
Results will be posted when all workers finish.
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
<!-- benchmark-request:6229581773:start:2a35cdb1b55d28006abca52bdafaa1cd -->
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): trim bitmap-backed row selections using 64-bit words [arrow-rs]
Posted by "haohuaijin (via GitHub)" <gi...@apache.org>.
haohuaijin commented on PR #11639:
URL: https://github.com/apache/arrow-rs/pull/11639#issuecomment-6229581773
run benchmark row_selector_boolean_buffer
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] perf(parquet): trim bitmap-backed row selections using 64-bit words [arrow-rs]
Posted by "haohuaijin (via GitHub)" <gi...@apache.org>.
haohuaijin commented on PR #11639:
URL: https://github.com/apache/arrow-rs/pull/11639#issuecomment-6230047937
the performance only improve when the tail have many rows, but this change make the code much easy to understand
```
group base changed
----- ---- -------
row_selector_boolean_buffer/reader/from_boolean_buffer/clustered/select01 1.36 244.9±1.31µs ? ?/sec 1.00 179.8±1.51µs ? ?/sec
row_selector_boolean_buffer/reader_async/from_boolean_buffer/clustered/select01 1.50 439.1±7.19µs ? ?/sec 1.00 293.6±5.56µs ? ?/sec
```
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org