You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@datafusion.apache.org by "alamb (via GitHub)" <gi...@apache.org> on 2026/08/17 11:08:56 UTC
[PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
alamb opened a new pull request, #24426:
URL: https://github.com/apache/datafusion/pull/24426
## Which issue does this PR close?
- Part of https://github.com/apache/datafusion/issues/24393
- Part of https://github.com/apache/datafusion/issues/20324
**Note: this is a benchmarking prototype and is NOT intended for merge or review.**
Stacked on https://github.com/apache/datafusion/pull/24395 (the `progressive_io` option).
## Rationale for this change
Measure whether enabling `pushdown_filters` by default can now match or beat today's ClickBench performance when the I/O pattern is kept the same as today (one request per row group), i.e. with `progressive_io = false` (the default introduced in #24395).
## What changes are included in this PR?
- The `progressive_io` commit from #24395
- Flips the default of `datafusion.execution.parquet.pushdown_filters` to `true` (`progressive_io` stays `false`, so scans still fetch each row group's data in a single request)
## Are these changes tested?
Benchmark runs only (`clickbench_partitioned`, with and without simulated object store latency); results to be posted below. Other tests that encode the pushdown-off default are expected to fail and are not being fixed in this prototype.
## Are there any user-facing changes?
Not intended for merge as-is.
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5331649129
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5331459712)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (026fb7d25c01c00144dbf3082a2370c1f6b56b10) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..026fb7d25c01c00144dbf3082a2370c1f6b56b10)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.23 ms β 1.23 ms β no change β
β QQuery 1 β 11.85 ms β 12.11 ms β no change β
β QQuery 2 β 36.88 ms β 36.39 ms β no change β
β QQuery 3 β 30.68 ms β 30.53 ms β no change β
β QQuery 4 β 227.74 ms β 222.32 ms β no change β
β QQuery 5 β 274.18 ms β 274.38 ms β no change β
β QQuery 6 β 1.28 ms β 1.26 ms β no change β
β QQuery 7 β 13.14 ms β 13.25 ms β no change β
β QQuery 8 β 326.10 ms β 327.62 ms β no change β
β QQuery 9 β 458.58 ms β 448.49 ms β no change β
β QQuery 10 β 70.14 ms β 69.68 ms β no change β
β QQuery 11 β 81.96 ms β 80.88 ms β no change β
β QQuery 12 β 270.35 ms β 272.82 ms β no change β
β QQuery 13 β 368.18 ms β 366.52 ms β no change β
β QQuery 14 β 285.55 ms β 282.33 ms β no change β
β QQuery 15 β 271.55 ms β 274.71 ms β no change β
β QQuery 16 β 620.35 ms β 619.38 ms β no change β
β QQuery 17 β 616.99 ms β 621.46 ms β no change β
β QQuery 18 β 1255.03 ms β 1260.62 ms β no change β
β QQuery 19 β 27.65 ms β 27.98 ms β no change β
β QQuery 20 β 511.77 ms β 519.82 ms β no change β
β QQuery 21 β 514.72 ms β 517.34 ms β no change β
β QQuery 22 β 979.32 ms β 989.09 ms β no change β
β QQuery 23 β 3059.93 ms β 151.89 ms β +20.15x faster β
β QQuery 24 β 40.93 ms β 55.16 ms β 1.35x slower β
β QQuery 25 β 109.30 ms β 144.42 ms β 1.32x slower β
β QQuery 26 β 41.61 ms β 58.02 ms β 1.39x slower β
β QQuery 27 β 509.43 ms β 512.16 ms β no change β
β QQuery 28 β 2909.28 ms β 2912.02 ms β no change β
β QQuery 29 β 41.38 ms β 41.74 ms β no change β
β QQuery 30 β 306.69 ms β 306.15 ms β no change β
β QQuery 31 β 277.70 ms β 286.65 ms β no change β
β QQuery 32 β 917.63 ms β 942.89 ms β no change β
β QQuery 33 β 1448.36 ms β 1463.73 ms β no change β
β QQuery 34 β 1479.07 ms β 1481.22 ms β no change β
β QQuery 35 β 280.36 ms β 283.60 ms β no change β
β QQuery 36 β 69.45 ms β 66.90 ms β no change β
β QQuery 37 β 35.89 ms β 35.65 ms β no change β
β QQuery 38 β 41.82 ms β 41.90 ms β no change β
β QQuery 39 β 136.46 ms β 131.18 ms β no change β
β QQuery 40 β 14.62 ms β 14.63 ms β no change β
β QQuery 41 β 14.31 ms β 14.20 ms β no change β
β QQuery 42 β 13.39 ms β 13.35 ms β no change β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 19002.81ms β
β Total Time (alamb_pushdown_by_default) β 16227.67ms β
β Average Time (HEAD) β 441.93ms β
β Average Time (alamb_pushdown_by_default) β 377.39ms β
β Queries Faster β 1 β
β Queries Slower β 3 β
β Queries with No Change β 39 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.23 / 4.01 Β±5.41 / 14.83 ms β 1.23 / 3.95 Β±5.33 / 14.60 ms β no change β
β QQuery 1 β 11.85 / 12.28 Β±0.25 / 12.64 ms β 12.11 / 12.36 Β±0.20 / 12.64 ms β no change β
β QQuery 2 β 36.88 / 37.08 Β±0.22 / 37.49 ms β 36.39 / 36.62 Β±0.15 / 36.82 ms β no change β
β QQuery 3 β 30.68 / 31.30 Β±0.68 / 32.51 ms β 30.53 / 30.92 Β±0.24 / 31.24 ms β no change β
β QQuery 4 β 227.74 / 230.79 Β±2.56 / 235.53 ms β 222.32 / 229.91 Β±4.72 / 235.97 ms β no change β
β QQuery 5 β 274.18 / 276.73 Β±1.85 / 279.39 ms β 274.38 / 276.28 Β±1.80 / 279.31 ms β no change β
β QQuery 6 β 1.28 / 1.44 Β±0.22 / 1.88 ms β 1.26 / 1.41 Β±0.23 / 1.86 ms β no change β
β QQuery 7 β 13.14 / 13.53 Β±0.34 / 14.16 ms β 13.25 / 13.55 Β±0.23 / 13.92 ms β no change β
β QQuery 8 β 326.10 / 332.64 Β±4.07 / 337.86 ms β 327.62 / 331.24 Β±2.45 / 334.29 ms β no change β
β QQuery 9 β 458.58 / 467.66 Β±6.15 / 475.11 ms β 448.49 / 468.37 Β±12.44 / 487.10 ms β no change β
β QQuery 10 β 70.14 / 70.91 Β±0.58 / 71.67 ms β 69.68 / 73.19 Β±6.30 / 85.78 ms β no change β
β QQuery 11 β 81.96 / 84.53 Β±4.27 / 93.05 ms β 80.88 / 84.01 Β±5.29 / 94.50 ms β no change β
β QQuery 12 β 270.35 / 277.77 Β±4.36 / 283.91 ms β 272.82 / 275.58 Β±3.34 / 282.11 ms β no change β
β QQuery 13 β 368.18 / 376.24 Β±9.16 / 393.77 ms β 366.52 / 377.81 Β±7.40 / 389.43 ms β no change β
β QQuery 14 β 285.55 / 290.35 Β±3.61 / 295.40 ms β 282.33 / 289.77 Β±5.06 / 297.95 ms β no change β
β QQuery 15 β 271.55 / 280.66 Β±6.47 / 291.83 ms β 274.71 / 284.78 Β±9.09 / 299.91 ms β no change β
β QQuery 16 β 620.35 / 629.53 Β±7.65 / 640.71 ms β 619.38 / 627.83 Β±5.63 / 636.76 ms β no change β
β QQuery 17 β 616.99 / 633.16 Β±10.74 / 648.25 ms β 621.46 / 634.35 Β±12.83 / 657.01 ms β no change β
β QQuery 18 β 1255.03 / 1279.95 Β±16.54 / 1306.73 ms β 1260.62 / 1294.29 Β±25.94 / 1334.77 ms β no change β
β QQuery 19 β 27.65 / 34.21 Β±12.60 / 59.41 ms β 27.98 / 30.49 Β±4.65 / 39.77 ms β +1.12x faster β
β QQuery 20 β 511.77 / 520.48 Β±8.73 / 537.10 ms β 519.82 / 528.90 Β±11.18 / 550.84 ms β no change β
β QQuery 21 β 514.72 / 519.95 Β±3.51 / 524.89 ms β 517.34 / 525.49 Β±8.92 / 539.64 ms β no change β
β QQuery 22 β 979.32 / 993.36 Β±7.64 / 1001.22 ms β 989.09 / 995.57 Β±4.28 / 1002.01 ms β no change β
β QQuery 23 β 3059.93 / 3091.39 Β±24.84 / 3114.55 ms β 151.89 / 158.58 Β±3.73 / 163.28 ms β +19.49x faster β
β QQuery 24 β 40.93 / 44.87 Β±3.73 / 50.09 ms β 55.16 / 61.38 Β±7.55 / 74.44 ms β 1.37x slower β
β QQuery 25 β 109.30 / 110.87 Β±1.55 / 112.94 ms β 144.42 / 145.43 Β±0.91 / 146.98 ms β 1.31x slower β
β QQuery 26 β 41.61 / 43.99 Β±2.40 / 47.19 ms β 58.02 / 61.70 Β±4.54 / 70.59 ms β 1.40x slower β
β QQuery 27 β 509.43 / 517.87 Β±6.95 / 530.25 ms β 512.16 / 521.27 Β±7.57 / 534.76 ms β no change β
β QQuery 28 β 2909.28 / 2926.58 Β±15.48 / 2952.34 ms β 2912.02 / 2929.57 Β±18.63 / 2963.52 ms β no change β
β QQuery 29 β 41.38 / 49.90 Β±10.26 / 64.91 ms β 41.74 / 52.43 Β±12.97 / 74.25 ms β 1.05x slower β
β QQuery 30 β 306.69 / 319.07 Β±13.87 / 346.10 ms β 306.15 / 317.53 Β±9.63 / 333.37 ms β no change β
β QQuery 31 β 277.70 / 293.67 Β±10.54 / 308.09 ms β 286.65 / 298.31 Β±9.84 / 309.88 ms β no change β
β QQuery 32 β 917.63 / 959.59 Β±28.07 / 995.17 ms β 942.89 / 966.10 Β±22.69 / 999.56 ms β no change β
β QQuery 33 β 1448.36 / 1476.94 Β±22.66 / 1504.19 ms β 1463.73 / 1480.66 Β±12.64 / 1498.75 ms β no change β
β QQuery 34 β 1479.07 / 1511.63 Β±17.85 / 1529.69 ms β 1481.22 / 1522.32 Β±31.19 / 1575.79 ms β no change β
β QQuery 35 β 280.36 / 309.17 Β±38.79 / 384.89 ms β 283.60 / 318.17 Β±38.02 / 372.76 ms β no change β
β QQuery 36 β 69.45 / 71.24 Β±1.75 / 74.40 ms β 66.90 / 73.28 Β±5.38 / 83.12 ms β no change β
β QQuery 37 β 35.89 / 39.18 Β±2.95 / 43.56 ms β 35.65 / 37.82 Β±3.30 / 44.27 ms β no change β
β QQuery 38 β 41.82 / 45.55 Β±2.94 / 50.35 ms β 41.90 / 47.26 Β±4.78 / 55.58 ms β no change β
β QQuery 39 β 136.46 / 145.48 Β±6.56 / 153.02 ms β 131.18 / 138.29 Β±4.82 / 145.23 ms β no change β
β QQuery 40 β 14.62 / 18.87 Β±4.44 / 25.24 ms β 14.63 / 19.02 Β±4.51 / 26.19 ms β no change β
β QQuery 41 β 14.31 / 16.67 Β±3.59 / 23.81 ms β 14.20 / 18.37 Β±7.52 / 33.39 ms β 1.10x slower β
β QQuery 42 β 13.39 / 15.98 Β±3.00 / 20.17 ms β 13.35 / 13.56 Β±0.13 / 13.72 ms β +1.18x faster β
βββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 19407.09ms β
β Total Time (alamb_pushdown_by_default) β 16607.70ms β
β Average Time (HEAD) β 451.33ms β
β Average Time (alamb_pushdown_by_default) β 386.23ms β
β Queries Faster β 3 β
β Queries Slower β 5 β
β Queries with No Change β 35 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 11.1 GiB |
| Avg memory | 4.0 GiB |
| CPU user | 996.8s |
| CPU sys | 68.1s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 85.0s |
| Peak memory | 10.3 GiB |
| Avg memory | 4.2 GiB |
| CPU user | 845.0s |
| CPU sys | 57.5s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5332422781
run benchmark clickbench_partitioned
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317648348
run benchmark clickbench_partitioned
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "zhuqi-lucas (via GitHub)" <gi...@apache.org>.
zhuqi-lucas commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5370788440
Thanks @alamb, great work β following the Q25 dig here, a precise DataFusion-side gate that stacks with the arrow-rs work, in case it's useful:
The row-level RowFilter is pure overhead in exactly one shape: when the (dynamic) filter column **is the sole projected column**. Then there's no "other column" for the selection to save a decode on; the filter column must be fully decoded to evaluate the threshold regardless; and for a high-cardinality key like `SearchPhrase`, RG pruning drops nothing β so the selection/mask construction is paid for zero payback.
That's a sharper predicate than "< 3 non-filter columns": decline only the *row-level* RowFilter (keep RG-level dynamic pruning) when `dynamic filter column == the sole projected column`. It hits Q25 (`ORDER BY SearchPhrase`, projection `{SearchPhrase}`) and leaves Q24 (`ORDER BY EventTime`, `EventTime` β projection β `SearchPhrase` is a savable "other column") and any multi-column projection untouched β those always have a β -filter column for the selection to earn its keep.
Orthogonal to the selection-construction speedups in #10776 β if those land this becomes unnecessary; until then it's a zero-risk way to drop the Q25 regression specifically.
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317648545
run benchmark clickbench_partitioned
```
env:
SIMULATE_LATENCY: "true"
```
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "github-actions[bot] (via GitHub)" <gi...@apache.org>.
github-actions[bot] commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5439472166
<!-- semver-check-comment -->
Thank you for opening this pull request!
Reviewer note: [cargo-semver-checks](https://github.com/obi1kenobi/cargo-semver-checks) reported the current version number is not SemVer-compatible with the changes in this pull request (compared against the base branch).
<details>
<summary>Details</summary>
```
Cloning apache/main
Building datafusion v55.0.0 (current)
Built [ 60.408s] (current)
Parsing datafusion v55.0.0 (current)
Parsed [ 0.036s] (current)
Building datafusion v55.0.0 (baseline)
Built [ 60.345s] (baseline)
Parsing datafusion v55.0.0 (baseline)
Parsed [ 0.036s] (baseline)
Checking datafusion v55.0.0 -> v55.0.0 (no change; assume patch)
Checked [ 0.628s] 223 checks: 223 pass, 31 skip
Summary no semver update required
Finished [ 124.008s] datafusion
Building datafusion-common v55.0.0 (current)
Built [ 35.643s] (current)
Parsing datafusion-common v55.0.0 (current)
Parsed [ 0.063s] (current)
Building datafusion-common v55.0.0 (baseline)
Built [ 35.171s] (baseline)
Parsing datafusion-common v55.0.0 (baseline)
Parsed [ 0.063s] (baseline)
Checking datafusion-common v55.0.0 -> v55.0.0 (no change; assume patch)
Checked [ 0.688s] 223 checks: 222 pass, 1 fail, 0 warn, 31 skip
--- failure constructible_struct_adds_field: struct exhaustively constructible through public API adds field ---
Description:
A pub struct that could be exhaustively constructed with a literal using only public API has a new pub field, breaking existing exhaustive literals.
ref: https://doc.rust-lang.org/reference/expressions/struct-expr.html
impl: https://github.com/obi1kenobi/cargo-semver-checks/tree/v0.50.0/src/lints/constructible_struct_adds_field.ron
Failed in:
field ParquetOptions.pushdown_filter_mode in /home/runner/work/datafusion/datafusion/datafusion/common/src/config.rs:1348
field ParquetOptions.progressive_io in /home/runner/work/datafusion/datafusion/datafusion/common/src/config.rs:1348
Summary semver requires new major version: 1 major and 0 minor checks failed
Finished [ 72.783s] datafusion-common
Building datafusion-datasource-parquet v55.0.0 (current)
Built [ 50.888s] (current)
Parsing datafusion-datasource-parquet v55.0.0 (current)
Parsed [ 0.033s] (current)
Building datafusion-datasource-parquet v55.0.0 (baseline)
Built [ 50.412s] (baseline)
Parsing datafusion-datasource-parquet v55.0.0 (baseline)
Parsed [ 0.036s] (baseline)
Checking datafusion-datasource-parquet v55.0.0 -> v55.0.0 (no change; assume patch)
Checked [ 0.155s] 223 checks: 223 pass, 31 skip
Summary no semver update required
Finished [ 102.608s] datafusion-datasource-parquet
Building datafusion-proto-common v55.0.0 (current)
Built [ 23.252s] (current)
Parsing datafusion-proto-common v55.0.0 (current)
Parsed [ 0.046s] (current)
Building datafusion-proto-common v55.0.0 (baseline)
Built [ 23.288s] (baseline)
Parsing datafusion-proto-common v55.0.0 (baseline)
Parsed [ 0.051s] (baseline)
Checking datafusion-proto-common v55.0.0 -> v55.0.0 (no change; assume patch)
Checked [ 1.071s] 223 checks: 222 pass, 1 fail, 0 warn, 31 skip
--- failure constructible_struct_adds_field: struct exhaustively constructible through public API adds field ---
Description:
A pub struct that could be exhaustively constructed with a literal using only public API has a new pub field, breaking existing exhaustive literals.
ref: https://doc.rust-lang.org/reference/expressions/struct-expr.html
impl: https://github.com/obi1kenobi/cargo-semver-checks/tree/v0.50.0/src/lints/constructible_struct_adds_field.ron
Failed in:
field ParquetOptions.pushdown_filter_mode in /home/runner/work/datafusion/datafusion/datafusion/proto-common/src/generated/prost.rs:819
field ParquetOptions.progressive_io in /home/runner/work/datafusion/datafusion/datafusion/proto-common/src/generated/prost.rs:828
field ParquetOptions.pushdown_filter_mode in /home/runner/work/datafusion/datafusion/datafusion/proto-common/src/generated/prost.rs:819
field ParquetOptions.progressive_io in /home/runner/work/datafusion/datafusion/datafusion/proto-common/src/generated/prost.rs:828
field ParquetOptions.pushdown_filter_mode in /home/runner/work/datafusion/datafusion/datafusion/proto-common/src/generated/prost.rs:819
field ParquetOptions.progressive_io in /home/runner/work/datafusion/datafusion/datafusion/proto-common/src/generated/prost.rs:828
Summary semver requires new major version: 1 major and 0 minor checks failed
Finished [ 48.782s] datafusion-proto-common
Building datafusion-proto-models v55.0.0 (current)
Built [ 26.330s] (current)
Parsing datafusion-proto-models v55.0.0 (current)
Parsed [ 0.130s] (current)
Building datafusion-proto-models v55.0.0 (baseline)
Built [ 25.779s] (baseline)
Parsing datafusion-proto-models v55.0.0 (baseline)
Parsed [ 0.131s] (baseline)
Checking datafusion-proto-models v55.0.0 -> v55.0.0 (no change; assume patch)
Checked [ 1.827s] 223 checks: 222 pass, 1 fail, 0 warn, 31 skip
--- failure constructible_struct_adds_field: struct exhaustively constructible through public API adds field ---
Description:
A pub struct that could be exhaustively constructed with a literal using only public API has a new pub field, breaking existing exhaustive literals.
ref: https://doc.rust-lang.org/reference/expressions/struct-expr.html
impl: https://github.com/obi1kenobi/cargo-semver-checks/tree/v0.50.0/src/lints/constructible_struct_adds_field.ron
Failed in:
field ParquetOptions.pushdown_filter_mode in /home/runner/work/datafusion/datafusion/datafusion/proto-models/src/generated/datafusion_proto_common.rs:819
field ParquetOptions.progressive_io in /home/runner/work/datafusion/datafusion/datafusion/proto-models/src/generated/datafusion_proto_common.rs:828
field ParquetOptions.pushdown_filter_mode in /home/runner/work/datafusion/datafusion/datafusion/proto-models/src/generated/datafusion_proto_common.rs:819
field ParquetOptions.progressive_io in /home/runner/work/datafusion/datafusion/datafusion/proto-models/src/generated/datafusion_proto_common.rs:828
Summary semver requires new major version: 1 major and 0 minor checks failed
Finished [ 55.286s] datafusion-proto-models
Building datafusion-sqllogictest v55.0.0 (current)
Built [ 103.047s] (current)
Parsing datafusion-sqllogictest v55.0.0 (current)
Parsed [ 0.021s] (current)
Building datafusion-sqllogictest v55.0.0 (baseline)
Built [ 103.276s] (baseline)
Parsing datafusion-sqllogictest v55.0.0 (baseline)
Parsed [ 0.022s] (baseline)
Checking datafusion-sqllogictest v55.0.0 -> v55.0.0 (no change; assume patch)
Checked [ 0.095s] 223 checks: 223 pass, 31 skip
Summary no semver update required
Finished [ 209.280s] datafusion-sqllogictest
```
</details>
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5438898177
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5438849398)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5438849398-2016-jd9zs 6.12.85+ #1 SMP Sat Jun 27 09:31:30 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (1eb7f67b4c8822c26dfab75fcc303cb2f2f0702d) to a38bb10 (merge-base) [diff](https://github.com/apache/datafusion/compare/a38bb10c33cff4e99a2707059df99f6971c96f33..1eb7f67b4c8822c26dfab75fcc303cb2f2f0702d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5315251974
run benchmark clickbench_partitioned
```
env:
SIMULATE_LATENCY: "true"
```
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317681312
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5317648545)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5317648545-1623-2mknm 6.12.85+ #1 SMP Wed Jun 17 20:31:55 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (026fb7d25c01c00144dbf3082a2370c1f6b56b10) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..026fb7d25c01c00144dbf3082a2370c1f6b56b10)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317954296
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5317648545)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (026fb7d25c01c00144dbf3082a2370c1f6b56b10) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..026fb7d25c01c00144dbf3082a2370c1f6b56b10)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.23 ms β 1.25 ms β no change β
β QQuery 1 β 1874.07 ms β 1876.58 ms β no change β
β QQuery 2 β 2180.09 ms β 2164.87 ms β no change β
β QQuery 3 β 2140.30 ms β 2144.53 ms β no change β
β QQuery 4 β 2315.64 ms β 2260.11 ms β no change β
β QQuery 5 β 2392.10 ms β 2344.03 ms β no change β
β QQuery 6 β 1.29 ms β 1.31 ms β no change β
β QQuery 7 β 1976.07 ms β 1940.32 ms β no change β
β QQuery 8 β 2390.81 ms β 2381.63 ms β no change β
β QQuery 9 β 2586.81 ms β 2638.75 ms β no change β
β QQuery 10 β 2216.17 ms β 2199.45 ms β no change β
β QQuery 11 β 2173.21 ms β 2189.64 ms β no change β
β QQuery 12 β 2330.21 ms β 2352.98 ms β no change β
β QQuery 13 β 2437.53 ms β 2494.81 ms β no change β
β QQuery 14 β 2385.64 ms β 2380.96 ms β no change β
β QQuery 15 β 2335.43 ms β 2360.66 ms β no change β
β QQuery 16 β 2652.69 ms β 2621.34 ms β no change β
β QQuery 17 β 2654.27 ms β 2629.95 ms β no change β
β QQuery 18 β 3126.72 ms β 3155.93 ms β no change β
β QQuery 19 β 1745.51 ms β 1770.19 ms β no change β
β QQuery 20 β 2616.70 ms β 2661.13 ms β no change β
β QQuery 21 β 2665.04 ms β 2609.38 ms β no change β
β QQuery 22 β 3034.14 ms β 3139.97 ms β no change β
β QQuery 23 β 4954.98 ms β 759.78 ms β +6.52x faster β
β QQuery 24 β 447.77 ms β 584.13 ms β 1.30x slower β
β QQuery 25 β 2260.50 ms β 2358.47 ms β no change β
β QQuery 26 β 448.81 ms β 627.72 ms β 1.40x slower β
β QQuery 27 β 2634.57 ms β 2631.15 ms β no change β
β QQuery 28 β 4947.99 ms β 4912.29 ms β no change β
β QQuery 29 β 2198.39 ms β 2177.74 ms β no change β
β QQuery 30 β 2435.75 ms β 2437.69 ms β no change β
β QQuery 31 β 2362.92 ms β 2394.56 ms β no change β
β QQuery 32 β 2502.89 ms β 2509.51 ms β no change β
β QQuery 33 β 3332.49 ms β 3408.86 ms β no change β
β QQuery 34 β 3424.15 ms β 3419.42 ms β no change β
β QQuery 35 β 2408.92 ms β 2352.88 ms β no change β
β QQuery 36 β 137.88 ms β 195.12 ms β 1.42x slower β
β QQuery 37 β 134.88 ms β 148.97 ms β 1.10x slower β
β QQuery 38 β 128.25 ms β 174.88 ms β 1.36x slower β
β QQuery 39 β 180.86 ms β 275.47 ms β 1.52x slower β
β QQuery 40 β 117.56 ms β 144.88 ms β 1.23x slower β
β QQuery 41 β 114.25 ms β 143.36 ms β 1.25x slower β
β QQuery 42 β 114.48 ms β 143.35 ms β 1.25x slower β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 85519.98ms β
β Total Time (alamb_pushdown_by_default) β 82119.99ms β
β Average Time (HEAD) β 1988.84ms β
β Average Time (alamb_pushdown_by_default) β 1909.77ms β
β Queries Faster β 1 β
β Queries Slower β 9 β
β Queries with No Change β 33 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³βββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.23 / 3.89 Β±5.24 / 14.37 ms β 1.25 / 3.95 Β±5.37 / 14.69 ms β no change β
β QQuery 1 β 1874.07 / 1918.13 Β±24.62 / 1940.54 ms β 1876.58 / 1934.18 Β±42.96 / 2009.75 ms β no change β
β QQuery 2 β 2180.09 / 2238.33 Β±54.17 / 2303.60 ms β 2164.87 / 2219.81 Β±39.74 / 2278.53 ms β no change β
β QQuery 3 β 2140.30 / 2206.13 Β±33.71 / 2231.81 ms β 2144.53 / 2230.72 Β±71.22 / 2332.83 ms β no change β
β QQuery 4 β 2315.64 / 2357.87 Β±47.68 / 2448.93 ms β 2260.11 / 2332.16 Β±61.62 / 2445.64 ms β no change β
β QQuery 5 β 2392.10 / 2420.90 Β±29.48 / 2469.07 ms β 2344.03 / 2403.98 Β±42.38 / 2459.70 ms β no change β
β QQuery 6 β 1.29 / 1.44 Β±0.21 / 1.85 ms β 1.31 / 1.46 Β±0.25 / 1.96 ms β no change β
β QQuery 7 β 1976.07 / 2005.48 Β±25.76 / 2053.00 ms β 1940.32 / 1983.20 Β±52.12 / 2081.08 ms β no change β
β QQuery 8 β 2390.81 / 2435.29 Β±32.30 / 2474.35 ms β 2381.63 / 2424.25 Β±56.15 / 2534.25 ms β no change β
β QQuery 9 β 2586.81 / 2657.05 Β±79.73 / 2807.33 ms β 2638.75 / 2703.63 Β±53.21 / 2800.53 ms β no change β
β QQuery 10 β 2216.17 / 2271.67 Β±39.84 / 2320.81 ms β 2199.45 / 2273.42 Β±64.23 / 2366.73 ms β no change β
β QQuery 11 β 2173.21 / 2263.68 Β±68.76 / 2340.39 ms β 2189.64 / 2273.48 Β±49.70 / 2345.43 ms β no change β
β QQuery 12 β 2330.21 / 2404.62 Β±49.24 / 2462.25 ms β 2352.98 / 2428.26 Β±58.86 / 2518.53 ms β no change β
β QQuery 13 β 2437.53 / 2503.10 Β±53.55 / 2591.48 ms β 2494.81 / 2517.78 Β±25.17 / 2565.94 ms β no change β
β QQuery 14 β 2385.64 / 2423.11 Β±40.78 / 2498.23 ms β 2380.96 / 2434.76 Β±46.19 / 2514.03 ms β no change β
β QQuery 15 β 2335.43 / 2459.11 Β±65.93 / 2511.54 ms β 2360.66 / 2411.39 Β±42.47 / 2474.95 ms β no change β
β QQuery 16 β 2652.69 / 2697.11 Β±43.38 / 2759.97 ms β 2621.34 / 2699.60 Β±44.79 / 2760.64 ms β no change β
β QQuery 17 β 2654.27 / 2702.91 Β±49.78 / 2797.22 ms β 2629.95 / 2692.79 Β±50.56 / 2777.12 ms β no change β
β QQuery 18 β 3126.72 / 3190.92 Β±38.52 / 3232.90 ms β 3155.93 / 3181.61 Β±27.01 / 3227.39 ms β no change β
β QQuery 19 β 1745.51 / 1810.68 Β±64.62 / 1930.16 ms β 1770.19 / 1859.68 Β±73.30 / 1962.76 ms β no change β
β QQuery 20 β 2616.70 / 2680.66 Β±41.05 / 2737.98 ms β 2661.13 / 2715.75 Β±48.36 / 2781.24 ms β no change β
β QQuery 21 β 2665.04 / 2720.23 Β±31.71 / 2759.25 ms β 2609.38 / 2670.45 Β±54.52 / 2738.12 ms β no change β
β QQuery 22 β 3034.14 / 3125.14 Β±52.51 / 3189.99 ms β 3139.97 / 3190.47 Β±32.28 / 3227.32 ms β no change β
β QQuery 23 β 4954.98 / 5044.51 Β±49.93 / 5090.24 ms β 759.78 / 861.21 Β±85.11 / 968.38 ms β +5.86x faster β
β QQuery 24 β 447.77 / 490.17 Β±22.47 / 511.47 ms β 584.13 / 661.37 Β±56.26 / 728.41 ms β 1.35x slower β
β QQuery 25 β 2260.50 / 2298.14 Β±25.00 / 2335.34 ms β 2358.47 / 2393.03 Β±31.75 / 2440.80 ms β no change β
β QQuery 26 β 448.81 / 510.92 Β±42.44 / 573.91 ms β 627.72 / 705.78 Β±46.55 / 760.46 ms β 1.38x slower β
β QQuery 27 β 2634.57 / 2733.77 Β±76.57 / 2862.94 ms β 2631.15 / 2757.48 Β±95.44 / 2894.10 ms β no change β
β QQuery 28 β 4947.99 / 5029.38 Β±68.27 / 5121.56 ms β 4912.29 / 5046.07 Β±110.84 / 5188.86 ms β no change β
β QQuery 29 β 2198.39 / 2260.22 Β±39.25 / 2305.98 ms β 2177.74 / 2296.70 Β±72.69 / 2370.69 ms β no change β
β QQuery 30 β 2435.75 / 2486.75 Β±39.73 / 2551.32 ms β 2437.69 / 2518.04 Β±77.30 / 2624.17 ms β no change β
β QQuery 31 β 2362.92 / 2436.59 Β±75.42 / 2579.46 ms β 2394.56 / 2467.98 Β±59.81 / 2532.39 ms β no change β
β QQuery 32 β 2502.89 / 2596.28 Β±59.50 / 2679.58 ms β 2509.51 / 2583.78 Β±63.22 / 2692.49 ms β no change β
β QQuery 33 β 3332.49 / 3444.51 Β±79.07 / 3563.65 ms β 3408.86 / 3459.71 Β±47.62 / 3538.58 ms β no change β
β QQuery 34 β 3424.15 / 3456.10 Β±24.54 / 3494.30 ms β 3419.42 / 3455.53 Β±31.42 / 3504.63 ms β no change β
β QQuery 35 β 2408.92 / 2482.63 Β±68.24 / 2602.29 ms β 2352.88 / 2488.46 Β±93.47 / 2628.35 ms β no change β
β QQuery 36 β 137.88 / 294.24 Β±94.40 / 393.81 ms β 195.12 / 274.26 Β±65.58 / 350.23 ms β +1.07x faster β
β QQuery 37 β 134.88 / 251.36 Β±76.26 / 350.48 ms β 148.97 / 241.38 Β±54.55 / 308.62 ms β no change β
β QQuery 38 β 128.25 / 268.06 Β±86.79 / 375.32 ms β 174.88 / 241.99 Β±57.66 / 339.18 ms β +1.11x faster β
β QQuery 39 β 180.86 / 313.18 Β±74.67 / 405.28 ms β 275.47 / 308.42 Β±37.46 / 366.78 ms β no change β
β QQuery 40 β 117.56 / 246.78 Β±84.29 / 346.77 ms β 144.88 / 226.65 Β±45.51 / 274.38 ms β +1.09x faster β
β QQuery 41 β 114.25 / 214.36 Β±62.52 / 308.29 ms β 143.36 / 231.09 Β±54.15 / 300.19 ms β 1.08x slower β
β QQuery 42 β 114.48 / 240.57 Β±82.10 / 346.68 ms β 143.35 / 231.42 Β±54.43 / 301.43 ms β no change β
βββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 88595.97ms β
β Total Time (alamb_pushdown_by_default) β 85037.13ms β
β Average Time (HEAD) β 2060.37ms β
β Average Time (alamb_pushdown_by_default) β 1977.61ms β
β Queries Faster β 4 β
β Queries Slower β 3 β
β Queries with No Change β 36 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 445.1s |
| Peak memory | 9.6 GiB |
| Avg memory | 1.8 GiB |
| CPU user | 973.9s |
| CPU sys | 109.6s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 430.1s |
| Peak memory | 9.3 GiB |
| Avg memory | 1.8 GiB |
| CPU user | 831.4s |
| CPU sys | 98.4s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5438901284
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5438849038)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5438849038-2017-d22pn 6.12.85+ #1 SMP Sat Jun 27 09:31:30 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (1eb7f67b4c8822c26dfab75fcc303cb2f2f0702d) to a38bb10 (merge-base) [diff](https://github.com/apache/datafusion/compare/a38bb10c33cff4e99a2707059df99f6971c96f33..1eb7f67b4c8822c26dfab75fcc303cb2f2f0702d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "zhuqi-lucas (via GitHub)" <gi...@apache.org>.
zhuqi-lucas commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5406601538
> > the filter column must be fully decoded to evaluate the threshold regardless; and for a high-cardinality key like SearchPhrase, RG pruning drops nothing β so the selection/mask construction is paid for zero payback.
>
> Thanks @zhuqi-lucas - I think I had claude try this - it is what is listed as
>
> > and a refinement ([106653a](https://github.com/apache/datafusion/commit/106653a393b8815e09c09db239afa966f0f627b3)) that gates only static conjuncts when a dynamic (TopK/aggregate) filter is present
>
> And it didn't seem to work (though perhaps it wasn't implemented correctly)
Thank you @alamb for the info , i can try to benchmark it locally first.
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317672617
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5317648348)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5317648348-1624-mt4rj 6.12.85+ #1 SMP Wed Jun 17 20:31:55 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (026fb7d25c01c00144dbf3082a2370c1f6b56b10) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..026fb7d25c01c00144dbf3082a2370c1f6b56b10)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317313196
run benchmark clickbench_partitioned
```
env:
SIMULATE_LATENCY: "true"
```
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317848101
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5317648348)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (026fb7d25c01c00144dbf3082a2370c1f6b56b10) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..026fb7d25c01c00144dbf3082a2370c1f6b56b10)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.22 ms β 1.25 ms β no change β
β QQuery 1 β 12.15 ms β 11.98 ms β no change β
β QQuery 2 β 37.21 ms β 36.36 ms β no change β
β QQuery 3 β 31.37 ms β 30.79 ms β no change β
β QQuery 4 β 225.77 ms β 226.67 ms β no change β
β QQuery 5 β 273.47 ms β 271.78 ms β no change β
β QQuery 6 β 1.29 ms β 1.26 ms β no change β
β QQuery 7 β 13.54 ms β 13.06 ms β no change β
β QQuery 8 β 329.43 ms β 330.86 ms β no change β
β QQuery 9 β 452.98 ms β 465.81 ms β no change β
β QQuery 10 β 69.27 ms β 68.82 ms β no change β
β QQuery 11 β 81.35 ms β 79.76 ms β no change β
β QQuery 12 β 269.24 ms β 272.96 ms β no change β
β QQuery 13 β 370.12 ms β 362.15 ms β no change β
β QQuery 14 β 283.00 ms β 283.14 ms β no change β
β QQuery 15 β 276.07 ms β 270.51 ms β no change β
β QQuery 16 β 610.13 ms β 618.37 ms β no change β
β QQuery 17 β 617.02 ms β 622.08 ms β no change β
β QQuery 18 β 1252.91 ms β 1272.72 ms β no change β
β QQuery 19 β 27.91 ms β 27.26 ms β no change β
β QQuery 20 β 516.13 ms β 520.09 ms β no change β
β QQuery 21 β 513.77 ms β 514.33 ms β no change β
β QQuery 22 β 985.68 ms β 988.83 ms β no change β
β QQuery 23 β 3063.67 ms β 150.47 ms β +20.36x faster β
β QQuery 24 β 41.36 ms β 54.90 ms β 1.33x slower β
β QQuery 25 β 111.83 ms β 142.45 ms β 1.27x slower β
β QQuery 26 β 41.29 ms β 57.21 ms β 1.39x slower β
β QQuery 27 β 514.48 ms β 511.47 ms β no change β
β QQuery 28 β 2897.36 ms β 2900.94 ms β no change β
β QQuery 29 β 41.12 ms β 41.02 ms β no change β
β QQuery 30 β 307.57 ms β 299.05 ms β no change β
β QQuery 31 β 283.26 ms β 282.00 ms β no change β
β QQuery 32 β 939.63 ms β 929.59 ms β no change β
β QQuery 33 β 1503.38 ms β 1443.24 ms β no change β
β QQuery 34 β 1497.18 ms β 1468.03 ms β no change β
β QQuery 35 β 283.30 ms β 281.92 ms β no change β
β QQuery 36 β 70.96 ms β 67.69 ms β no change β
β QQuery 37 β 35.30 ms β 36.04 ms β no change β
β QQuery 38 β 40.39 ms β 44.51 ms β 1.10x slower β
β QQuery 39 β 153.47 ms β 142.18 ms β +1.08x faster β
β QQuery 40 β 14.30 ms β 14.65 ms β no change β
β QQuery 41 β 14.31 ms β 13.74 ms β no change β
β QQuery 42 β 13.78 ms β 13.59 ms β no change β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 19118.99ms β
β Total Time (alamb_pushdown_by_default) β 16185.53ms β
β Average Time (HEAD) β 444.63ms β
β Average Time (alamb_pushdown_by_default) β 376.41ms β
β Queries Faster β 2 β
β Queries Slower β 4 β
β Queries with No Change β 37 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.22 / 4.03 Β±5.47 / 14.98 ms β 1.25 / 4.08 Β±5.53 / 15.14 ms β no change β
β QQuery 1 β 12.15 / 12.43 Β±0.21 / 12.73 ms β 11.98 / 12.28 Β±0.17 / 12.50 ms β no change β
β QQuery 2 β 37.21 / 37.81 Β±0.43 / 38.52 ms β 36.36 / 36.62 Β±0.26 / 36.99 ms β no change β
β QQuery 3 β 31.37 / 32.00 Β±0.62 / 33.09 ms β 30.79 / 31.21 Β±0.30 / 31.52 ms β no change β
β QQuery 4 β 225.77 / 231.97 Β±3.46 / 236.05 ms β 226.67 / 230.23 Β±2.06 / 232.98 ms β no change β
β QQuery 5 β 273.47 / 276.10 Β±1.37 / 277.20 ms β 271.78 / 275.33 Β±2.88 / 278.61 ms β no change β
β QQuery 6 β 1.29 / 1.44 Β±0.22 / 1.87 ms β 1.26 / 1.42 Β±0.23 / 1.87 ms β no change β
β QQuery 7 β 13.54 / 13.73 Β±0.11 / 13.85 ms β 13.06 / 13.32 Β±0.17 / 13.56 ms β no change β
β QQuery 8 β 329.43 / 331.60 Β±1.91 / 334.62 ms β 330.86 / 333.41 Β±2.38 / 337.25 ms β no change β
β QQuery 9 β 452.98 / 468.06 Β±8.65 / 476.64 ms β 465.81 / 469.38 Β±4.87 / 478.98 ms β no change β
β QQuery 10 β 69.27 / 72.62 Β±4.44 / 81.36 ms β 68.82 / 72.95 Β±7.10 / 87.13 ms β no change β
β QQuery 11 β 81.35 / 84.93 Β±5.98 / 96.88 ms β 79.76 / 84.26 Β±6.88 / 97.93 ms β no change β
β QQuery 12 β 269.24 / 273.62 Β±3.43 / 279.60 ms β 272.96 / 277.56 Β±4.17 / 285.08 ms β no change β
β QQuery 13 β 370.12 / 384.02 Β±7.72 / 391.35 ms β 362.15 / 373.87 Β±6.67 / 381.10 ms β no change β
β QQuery 14 β 283.00 / 288.02 Β±3.73 / 292.36 ms β 283.14 / 294.09 Β±8.50 / 309.01 ms β no change β
β QQuery 15 β 276.07 / 280.21 Β±3.85 / 286.03 ms β 270.51 / 278.82 Β±6.47 / 284.40 ms β no change β
β QQuery 16 β 610.13 / 618.19 Β±5.10 / 625.76 ms β 618.37 / 635.55 Β±16.27 / 664.93 ms β no change β
β QQuery 17 β 617.02 / 632.26 Β±9.69 / 643.43 ms β 622.08 / 642.44 Β±16.56 / 671.68 ms β no change β
β QQuery 18 β 1252.91 / 1280.70 Β±14.86 / 1297.58 ms β 1272.72 / 1297.24 Β±34.36 / 1364.24 ms β no change β
β QQuery 19 β 27.91 / 31.75 Β±5.16 / 41.02 ms β 27.26 / 27.67 Β±0.24 / 27.92 ms β +1.15x faster β
β QQuery 20 β 516.13 / 524.16 Β±9.26 / 542.27 ms β 520.09 / 525.94 Β±6.82 / 534.95 ms β no change β
β QQuery 21 β 513.77 / 520.89 Β±5.71 / 530.46 ms β 514.33 / 523.50 Β±8.30 / 533.82 ms β no change β
β QQuery 22 β 985.68 / 992.98 Β±7.30 / 1004.38 ms β 988.83 / 992.49 Β±4.02 / 999.26 ms β no change β
β QQuery 23 β 3063.67 / 3091.19 Β±20.46 / 3125.30 ms β 150.47 / 156.57 Β±4.14 / 160.22 ms β +19.74x faster β
β QQuery 24 β 41.36 / 47.80 Β±9.05 / 64.76 ms β 54.90 / 63.17 Β±15.82 / 94.82 ms β 1.32x slower β
β QQuery 25 β 111.83 / 115.64 Β±6.42 / 128.47 ms β 142.45 / 145.63 Β±2.59 / 149.14 ms β 1.26x slower β
β QQuery 26 β 41.29 / 44.38 Β±5.52 / 55.42 ms β 57.21 / 58.83 Β±1.58 / 61.35 ms β 1.33x slower β
β QQuery 27 β 514.48 / 517.85 Β±2.87 / 522.93 ms β 511.47 / 515.79 Β±3.25 / 521.14 ms β no change β
β QQuery 28 β 2897.36 / 2918.66 Β±22.97 / 2955.45 ms β 2900.94 / 2918.35 Β±13.44 / 2939.83 ms β no change β
β QQuery 29 β 41.12 / 44.05 Β±4.78 / 53.56 ms β 41.02 / 41.34 Β±0.34 / 41.99 ms β +1.07x faster β
β QQuery 30 β 307.57 / 311.62 Β±3.86 / 318.36 ms β 299.05 / 311.50 Β±6.38 / 316.57 ms β no change β
β QQuery 31 β 283.26 / 293.93 Β±10.29 / 308.02 ms β 282.00 / 293.35 Β±9.72 / 311.38 ms β no change β
β QQuery 32 β 939.63 / 979.56 Β±23.37 / 1006.26 ms β 929.59 / 956.87 Β±25.14 / 993.89 ms β no change β
β QQuery 33 β 1503.38 / 1511.99 Β±11.33 / 1534.34 ms β 1443.24 / 1490.47 Β±29.31 / 1529.73 ms β no change β
β QQuery 34 β 1497.18 / 1525.63 Β±23.96 / 1552.41 ms β 1468.03 / 1544.19 Β±43.79 / 1602.20 ms β no change β
β QQuery 35 β 283.30 / 315.11 Β±35.66 / 372.69 ms β 281.92 / 318.65 Β±26.57 / 351.55 ms β no change β
β QQuery 36 β 70.96 / 79.52 Β±6.50 / 87.96 ms β 67.69 / 77.82 Β±9.76 / 93.34 ms β no change β
β QQuery 37 β 35.30 / 40.43 Β±6.53 / 53.32 ms β 36.04 / 39.78 Β±4.07 / 46.99 ms β no change β
β QQuery 38 β 40.39 / 46.85 Β±6.12 / 55.59 ms β 44.51 / 49.41 Β±5.33 / 59.14 ms β 1.05x slower β
β QQuery 39 β 153.47 / 158.91 Β±5.20 / 168.67 ms β 142.18 / 145.98 Β±3.20 / 151.87 ms β +1.09x faster β
β QQuery 40 β 14.30 / 15.89 Β±2.31 / 20.47 ms β 14.65 / 18.31 Β±5.75 / 29.64 ms β 1.15x slower β
β QQuery 41 β 14.31 / 14.43 Β±0.13 / 14.66 ms β 13.74 / 14.01 Β±0.20 / 14.34 ms β no change β
β QQuery 42 β 13.78 / 15.02 Β±2.20 / 19.42 ms β 13.59 / 16.33 Β±5.24 / 26.81 ms β 1.09x slower β
βββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 19482.04ms β
β Total Time (alamb_pushdown_by_default) β 16610.00ms β
β Average Time (HEAD) β 453.07ms β
β Average Time (alamb_pushdown_by_default) β 386.28ms β
β Queries Faster β 4 β
β Queries Slower β 6 β
β Queries with No Change β 33 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 11.5 GiB |
| Avg memory | 4.3 GiB |
| CPU user | 996.5s |
| CPU sys | 71.5s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 85.0s |
| Peak memory | 10.1 GiB |
| Avg memory | 4.0 GiB |
| CPU user | 842.5s |
| CPU sys | 60.1s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5331549149
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5331512757)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5331512757-1658-2hvnr 6.12.85+ #1 SMP Wed Jun 17 20:31:55 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (026fb7d25c01c00144dbf3082a2370c1f6b56b10) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..026fb7d25c01c00144dbf3082a2370c1f6b56b10)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5315465149
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5315251680)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (79dd91dc7c0119253e85e64cf0521617cd513680) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..79dd91dc7c0119253e85e64cf0521617cd513680)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.36 ms β 1.26 ms β +1.08x faster β
β QQuery 1 β 12.45 ms β 12.47 ms β no change β
β QQuery 2 β 36.98 ms β 37.03 ms β no change β
β QQuery 3 β 32.66 ms β 30.89 ms β +1.06x faster β
β QQuery 4 β 236.02 ms β 222.75 ms β +1.06x faster β
β QQuery 5 β 270.92 ms β 271.40 ms β no change β
β QQuery 6 β 1.28 ms β 1.27 ms β no change β
β QQuery 7 β 13.10 ms β 15.73 ms β 1.20x slower β
β QQuery 8 β 331.19 ms β 321.39 ms β no change β
β QQuery 9 β 447.50 ms β 447.00 ms β no change β
β QQuery 10 β 69.49 ms β 92.55 ms β 1.33x slower β
β QQuery 11 β 81.42 ms β 103.86 ms β 1.28x slower β
β QQuery 12 β 281.09 ms β 293.47 ms β no change β
β QQuery 13 β 399.79 ms β 400.89 ms β no change β
β QQuery 14 β 279.58 ms β 312.53 ms β 1.12x slower β
β QQuery 15 β 265.82 ms β 265.67 ms β no change β
β QQuery 16 β 607.81 ms β 620.94 ms β no change β
β QQuery 17 β 611.41 ms β 709.03 ms β 1.16x slower β
β QQuery 18 β 1243.83 ms β 1287.79 ms β no change β
β QQuery 19 β 27.42 ms β 28.70 ms β no change β
β QQuery 20 β 514.15 ms β 513.01 ms β no change β
β QQuery 21 β 510.35 ms β 567.14 ms β 1.11x slower β
β QQuery 22 β 976.92 ms β 792.99 ms β +1.23x faster β
β QQuery 23 β 3021.65 ms β 148.70 ms β +20.32x faster β
β QQuery 24 β 43.96 ms β 44.76 ms β no change β
β QQuery 25 β 116.41 ms β 140.46 ms β 1.21x slower β
β QQuery 26 β 43.65 ms β 48.20 ms β 1.10x slower β
β QQuery 27 β 538.11 ms β 563.97 ms β no change β
β QQuery 28 β 2918.55 ms β 2901.35 ms β no change β
β QQuery 29 β 40.75 ms β 42.00 ms β no change β
β QQuery 30 β 295.36 ms β 303.85 ms β no change β
β QQuery 31 β 276.96 ms β 283.40 ms β no change β
β QQuery 32 β 923.13 ms β 952.44 ms β no change β
β QQuery 33 β 1431.63 ms β 1437.87 ms β no change β
β QQuery 34 β 1422.36 ms β 1562.41 ms β 1.10x slower β
β QQuery 35 β 272.20 ms β 276.78 ms β no change β
β QQuery 36 β 66.97 ms β 66.16 ms β no change β
β QQuery 37 β 35.59 ms β 35.68 ms β no change β
β QQuery 38 β 42.04 ms β 34.45 ms β +1.22x faster β
β QQuery 39 β 153.05 ms β 134.28 ms β +1.14x faster β
β QQuery 40 β 15.62 ms β 17.11 ms β 1.10x slower β
β QQuery 41 β 15.48 ms β 16.11 ms β no change β
β QQuery 42 β 14.63 ms β 13.68 ms β +1.07x faster β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 18940.64ms β
β Total Time (alamb_pushdown_by_default) β 16373.43ms β
β Average Time (HEAD) β 440.48ms β
β Average Time (alamb_pushdown_by_default) β 380.78ms β
β Queries Faster β 8 β
β Queries Slower β 10 β
β Queries with No Change β 25 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³βββββββββββββββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.36 / 4.33 Β±5.74 / 15.81 ms β 1.26 / 4.22 Β±5.75 / 15.71 ms β no change β
β QQuery 1 β 12.45 / 12.83 Β±0.27 / 13.19 ms β 12.47 / 12.69 Β±0.12 / 12.82 ms β no change β
β QQuery 2 β 36.98 / 37.45 Β±0.42 / 38.22 ms β 37.03 / 37.63 Β±0.42 / 38.26 ms β no change β
β QQuery 3 β 32.66 / 34.37 Β±1.19 / 35.97 ms β 30.89 / 31.14 Β±0.20 / 31.47 ms β +1.10x faster β
β QQuery 4 β 236.02 / 262.29 Β±13.45 / 272.50 ms β 222.75 / 228.57 Β±3.06 / 230.94 ms β +1.15x faster β
β QQuery 5 β 270.92 / 276.89 Β±4.79 / 283.91 ms β 271.40 / 273.73 Β±1.95 / 276.29 ms β no change β
β QQuery 6 β 1.28 / 1.43 Β±0.22 / 1.85 ms β 1.27 / 1.42 Β±0.23 / 1.87 ms β no change β
β QQuery 7 β 13.10 / 13.22 Β±0.14 / 13.46 ms β 15.73 / 15.89 Β±0.11 / 16.04 ms β 1.20x slower β
β QQuery 8 β 331.19 / 346.22 Β±11.21 / 357.60 ms β 321.39 / 325.47 Β±2.60 / 328.46 ms β +1.06x faster β
β QQuery 9 β 447.50 / 461.45 Β±7.24 / 468.32 ms β 447.00 / 457.81 Β±7.70 / 467.93 ms β no change β
β QQuery 10 β 69.49 / 73.34 Β±5.49 / 84.22 ms β 92.55 / 96.04 Β±6.18 / 108.38 ms β 1.31x slower β
β QQuery 11 β 81.42 / 82.06 Β±0.51 / 82.71 ms β 103.86 / 105.06 Β±0.92 / 106.28 ms β 1.28x slower β
β QQuery 12 β 281.09 / 299.85 Β±9.49 / 306.26 ms β 293.47 / 304.44 Β±11.11 / 324.69 ms β no change β
β QQuery 13 β 399.79 / 410.01 Β±8.86 / 425.12 ms β 400.89 / 414.11 Β±16.11 / 444.86 ms β no change β
β QQuery 14 β 279.58 / 297.39 Β±11.23 / 310.94 ms β 312.53 / 324.69 Β±9.40 / 336.35 ms β 1.09x slower β
β QQuery 15 β 265.82 / 274.65 Β±5.58 / 280.85 ms β 265.67 / 275.66 Β±6.22 / 282.81 ms β no change β
β QQuery 16 β 607.81 / 616.71 Β±6.81 / 628.31 ms β 620.94 / 690.46 Β±37.38 / 726.20 ms β 1.12x slower β
β QQuery 17 β 611.41 / 624.37 Β±11.77 / 645.51 ms β 709.03 / 723.94 Β±14.76 / 747.40 ms β 1.16x slower β
β QQuery 18 β 1243.83 / 1261.06 Β±16.57 / 1292.07 ms β 1287.79 / 1397.23 Β±78.96 / 1479.38 ms β 1.11x slower β
β QQuery 19 β 27.42 / 29.58 Β±3.46 / 36.46 ms β 28.70 / 35.84 Β±14.06 / 63.95 ms β 1.21x slower β
β QQuery 20 β 514.15 / 532.53 Β±14.67 / 556.77 ms β 513.01 / 519.41 Β±6.28 / 530.02 ms β no change β
β QQuery 21 β 510.35 / 518.51 Β±9.63 / 536.45 ms β 567.14 / 574.79 Β±5.70 / 584.51 ms β 1.11x slower β
β QQuery 22 β 976.92 / 992.77 Β±8.73 / 1000.90 ms β 792.99 / 811.20 Β±21.22 / 852.79 ms β +1.22x faster β
β QQuery 23 β 3021.65 / 3101.22 Β±66.02 / 3195.90 ms β 148.70 / 160.31 Β±8.14 / 170.60 ms β +19.34x faster β
β QQuery 24 β 43.96 / 49.11 Β±4.18 / 55.17 ms β 44.76 / 51.54 Β±12.76 / 77.05 ms β no change β
β QQuery 25 β 116.41 / 118.23 Β±1.11 / 119.40 ms β 140.46 / 145.23 Β±4.65 / 152.75 ms β 1.23x slower β
β QQuery 26 β 43.65 / 47.97 Β±5.50 / 57.53 ms β 48.20 / 48.78 Β±0.39 / 49.28 ms β no change β
β QQuery 27 β 538.11 / 542.75 Β±4.11 / 548.93 ms β 563.97 / 575.28 Β±9.08 / 587.97 ms β 1.06x slower β
β QQuery 28 β 2918.55 / 2969.61 Β±37.88 / 3030.16 ms β 2901.35 / 2958.81 Β±34.02 / 2997.36 ms β no change β
β QQuery 29 β 40.75 / 40.95 Β±0.21 / 41.33 ms β 42.00 / 63.69 Β±41.99 / 147.66 ms β 1.56x slower β
β QQuery 30 β 295.36 / 313.38 Β±12.64 / 330.77 ms β 303.85 / 318.19 Β±12.10 / 337.74 ms β no change β
β QQuery 31 β 276.96 / 296.42 Β±21.49 / 338.29 ms β 283.40 / 293.56 Β±6.85 / 300.71 ms β no change β
β QQuery 32 β 923.13 / 938.07 Β±13.60 / 961.54 ms β 952.44 / 984.60 Β±24.47 / 1007.72 ms β no change β
β QQuery 33 β 1431.63 / 1504.88 Β±68.82 / 1608.74 ms β 1437.87 / 1486.73 Β±54.20 / 1589.92 ms β no change β
β QQuery 34 β 1422.36 / 1482.86 Β±48.25 / 1543.57 ms β 1562.41 / 1743.52 Β±100.86 / 1857.86 ms β 1.18x slower β
β QQuery 35 β 272.20 / 316.56 Β±80.19 / 476.56 ms β 276.78 / 318.90 Β±39.42 / 385.75 ms β no change β
β QQuery 36 β 66.97 / 71.25 Β±3.43 / 75.92 ms β 66.16 / 77.53 Β±12.90 / 102.69 ms β 1.09x slower β
β QQuery 37 β 35.59 / 42.75 Β±12.20 / 67.07 ms β 35.68 / 39.02 Β±3.00 / 42.63 ms β +1.10x faster β
β QQuery 38 β 42.04 / 43.92 Β±1.10 / 45.15 ms β 34.45 / 37.14 Β±2.07 / 39.25 ms β +1.18x faster β
β QQuery 39 β 153.05 / 162.85 Β±7.41 / 173.07 ms β 134.28 / 141.76 Β±4.97 / 149.92 ms β +1.15x faster β
β QQuery 40 β 15.62 / 16.05 Β±0.28 / 16.37 ms β 17.11 / 17.49 Β±0.28 / 17.99 ms β 1.09x slower β
β QQuery 41 β 15.48 / 17.70 Β±4.10 / 25.90 ms β 16.11 / 19.37 Β±3.66 / 25.32 ms β 1.09x slower β
β QQuery 42 β 14.63 / 14.88 Β±0.17 / 15.14 ms β 13.68 / 15.23 Β±2.39 / 19.99 ms β no change β
βββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 19554.70ms β
β Total Time (alamb_pushdown_by_default) β 17158.08ms β
β Average Time (HEAD) β 454.76ms β
β Average Time (alamb_pushdown_by_default) β 399.03ms β
β Queries Faster β 8 β
β Queries Slower β 16 β
β Queries with No Change β 19 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 12.4 GiB |
| Avg memory | 4.8 GiB |
| CPU user | 1005.1s |
| CPU sys | 70.4s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 90.0s |
| Peak memory | 12.3 GiB |
| Avg memory | 4.6 GiB |
| CPU user | 863.2s |
| CPU sys | 63.8s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5315576540
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5315251974)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (79dd91dc7c0119253e85e64cf0521617cd513680) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..79dd91dc7c0119253e85e64cf0521617cd513680)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.26 ms β 1.25 ms β no change β
β QQuery 1 β 1848.35 ms β 1876.73 ms β no change β
β QQuery 2 β 2189.42 ms β 2137.36 ms β no change β
β QQuery 3 β 2190.85 ms β 2136.69 ms β no change β
β QQuery 4 β 2299.50 ms β 2339.71 ms β no change β
β QQuery 5 β 2385.86 ms β 2391.03 ms β no change β
β QQuery 6 β 1.31 ms β 1.29 ms β no change β
β QQuery 7 β 2000.35 ms β 2001.56 ms β no change β
β QQuery 8 β 2395.68 ms β 2413.79 ms β no change β
β QQuery 9 β 2693.23 ms β 2669.13 ms β no change β
β QQuery 10 β 2212.87 ms β 2215.75 ms β no change β
β QQuery 11 β 2189.61 ms β 2204.18 ms β no change β
β QQuery 12 β 2367.53 ms β 2403.37 ms β no change β
β QQuery 13 β 2475.70 ms β 2522.12 ms β no change β
β QQuery 14 β 2357.78 ms β 2413.93 ms β no change β
β QQuery 15 β 2333.21 ms β 2366.86 ms β no change β
β QQuery 16 β 2644.50 ms β 2662.76 ms β no change β
β QQuery 17 β 2677.59 ms β 2662.68 ms β no change β
β QQuery 18 β 3208.64 ms β 3202.14 ms β no change β
β QQuery 19 β 1763.04 ms β 1757.59 ms β no change β
β QQuery 20 β 2677.89 ms β 2634.88 ms β no change β
β QQuery 21 β 2630.22 ms β 3586.81 ms β 1.36x slower β
β QQuery 22 β 3076.05 ms β 3664.77 ms β 1.19x slower β
β QQuery 23 β 4984.91 ms β 836.95 ms β +5.96x faster β
β QQuery 24 β 483.42 ms β 464.97 ms β no change β
β QQuery 25 β 2279.05 ms β 2300.48 ms β no change β
β QQuery 26 β 486.54 ms β 475.11 ms β no change β
β QQuery 27 β 2620.21 ms β 2716.04 ms β no change β
β QQuery 28 β 4984.21 ms β 5131.77 ms β no change β
β QQuery 29 β 2213.55 ms β 2268.36 ms β no change β
β QQuery 30 β 2400.21 ms β 2479.30 ms β no change β
β QQuery 31 β 2343.45 ms β 2376.46 ms β no change β
β QQuery 32 β 2505.26 ms β 2613.50 ms β no change β
β QQuery 33 β 3465.86 ms β 3447.20 ms β no change β
β QQuery 34 β 3469.37 ms β 3443.42 ms β no change β
β QQuery 35 β 2413.31 ms β 2398.03 ms β no change β
β QQuery 36 β 182.12 ms β 186.89 ms β no change β
β QQuery 37 β 143.71 ms β 149.96 ms β no change β
β QQuery 38 β 141.98 ms β 162.34 ms β 1.14x slower β
β QQuery 39 β 238.26 ms β 252.39 ms β 1.06x slower β
β QQuery 40 β 118.05 ms β 147.19 ms β 1.25x slower β
β QQuery 41 β 118.75 ms β 145.85 ms β 1.23x slower β
β QQuery 42 β 115.18 ms β 144.74 ms β 1.26x slower β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 86327.83ms β
β Total Time (alamb_pushdown_by_default) β 84407.35ms β
β Average Time (HEAD) β 2007.62ms β
β Average Time (alamb_pushdown_by_default) β 1962.96ms β
β Queries Faster β 1 β
β Queries Slower β 7 β
β Queries with No Change β 35 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.26 / 4.11 Β±5.56 / 15.23 ms β 1.25 / 4.13 Β±5.61 / 15.33 ms β no change β
β QQuery 1 β 1848.35 / 1914.40 Β±60.47 / 2018.82 ms β 1876.73 / 1947.40 Β±37.94 / 1983.46 ms β no change β
β QQuery 2 β 2189.42 / 2241.41 Β±31.68 / 2284.30 ms β 2137.36 / 2237.74 Β±69.34 / 2326.16 ms β no change β
β QQuery 3 β 2190.85 / 2245.02 Β±43.49 / 2297.16 ms β 2136.69 / 2190.18 Β±35.77 / 2235.74 ms β no change β
β QQuery 4 β 2299.50 / 2366.58 Β±42.61 / 2427.65 ms β 2339.71 / 2372.27 Β±36.30 / 2438.07 ms β no change β
β QQuery 5 β 2385.86 / 2479.22 Β±50.46 / 2526.91 ms β 2391.03 / 2427.39 Β±33.37 / 2486.94 ms β no change β
β QQuery 6 β 1.31 / 1.46 Β±0.21 / 1.88 ms β 1.29 / 1.45 Β±0.24 / 1.91 ms β no change β
β QQuery 7 β 2000.35 / 2043.00 Β±40.10 / 2110.82 ms β 2001.56 / 2046.45 Β±50.05 / 2108.01 ms β no change β
β QQuery 8 β 2395.68 / 2451.42 Β±33.18 / 2489.12 ms β 2413.79 / 2450.64 Β±20.86 / 2473.63 ms β no change β
β QQuery 9 β 2693.23 / 2710.81 Β±19.74 / 2748.86 ms β 2669.13 / 2756.41 Β±51.70 / 2808.61 ms β no change β
β QQuery 10 β 2212.87 / 2285.71 Β±58.40 / 2360.74 ms β 2215.75 / 2261.15 Β±38.06 / 2322.83 ms β no change β
β QQuery 11 β 2189.61 / 2245.50 Β±52.26 / 2335.60 ms β 2204.18 / 2274.91 Β±51.48 / 2341.57 ms β no change β
β QQuery 12 β 2367.53 / 2402.31 Β±28.95 / 2452.92 ms β 2403.37 / 2437.71 Β±25.40 / 2462.72 ms β no change β
β QQuery 13 β 2475.70 / 2530.72 Β±50.80 / 2602.82 ms β 2522.12 / 2560.69 Β±31.44 / 2613.47 ms β no change β
β QQuery 14 β 2357.78 / 2423.57 Β±75.77 / 2556.98 ms β 2413.93 / 2461.96 Β±48.57 / 2539.46 ms β no change β
β QQuery 15 β 2333.21 / 2427.64 Β±65.13 / 2525.92 ms β 2366.86 / 2445.35 Β±45.01 / 2496.45 ms β no change β
β QQuery 16 β 2644.50 / 2723.64 Β±68.49 / 2829.84 ms β 2662.76 / 2676.84 Β±10.76 / 2689.26 ms β no change β
β QQuery 17 β 2677.59 / 2715.85 Β±53.66 / 2818.73 ms β 2662.68 / 2694.45 Β±20.30 / 2715.31 ms β no change β
β QQuery 18 β 3208.64 / 3232.65 Β±19.80 / 3264.22 ms β 3202.14 / 3236.82 Β±23.92 / 3276.65 ms β no change β
β QQuery 19 β 1763.04 / 1810.45 Β±79.37 / 1967.81 ms β 1757.59 / 1813.03 Β±41.00 / 1878.18 ms β no change β
β QQuery 20 β 2677.89 / 2755.53 Β±62.43 / 2834.16 ms β 2634.88 / 2700.88 Β±56.01 / 2776.05 ms β no change β
β QQuery 21 β 2630.22 / 2685.09 Β±42.43 / 2732.52 ms β 3586.81 / 3666.40 Β±49.87 / 3728.71 ms β 1.37x slower β
β QQuery 22 β 3076.05 / 3209.35 Β±107.71 / 3375.40 ms β 3664.77 / 3810.82 Β±86.89 / 3930.12 ms β 1.19x slower β
β QQuery 23 β 4984.91 / 5096.60 Β±82.33 / 5189.53 ms β 836.95 / 899.33 Β±68.48 / 1020.75 ms β +5.67x faster β
β QQuery 24 β 483.42 / 498.83 Β±13.05 / 515.87 ms β 464.97 / 567.20 Β±61.71 / 653.84 ms β 1.14x slower β
β QQuery 25 β 2279.05 / 2340.66 Β±40.19 / 2386.15 ms β 2300.48 / 2382.51 Β±57.16 / 2449.83 ms β no change β
β QQuery 26 β 486.54 / 539.20 Β±38.57 / 598.87 ms β 475.11 / 553.67 Β±51.91 / 606.74 ms β no change β
β QQuery 27 β 2620.21 / 2709.57 Β±66.07 / 2786.24 ms β 2716.04 / 2811.02 Β±56.10 / 2890.88 ms β no change β
β QQuery 28 β 4984.21 / 5131.57 Β±89.49 / 5265.07 ms β 5131.77 / 5196.11 Β±48.06 / 5264.78 ms β no change β
β QQuery 29 β 2213.55 / 2280.38 Β±46.81 / 2334.42 ms β 2268.36 / 2345.16 Β±53.22 / 2431.37 ms β no change β
β QQuery 30 β 2400.21 / 2480.67 Β±47.32 / 2547.25 ms β 2479.30 / 2577.55 Β±60.11 / 2642.76 ms β no change β
β QQuery 31 β 2343.45 / 2432.20 Β±72.91 / 2531.38 ms β 2376.46 / 2455.27 Β±48.55 / 2516.65 ms β no change β
β QQuery 32 β 2505.26 / 2572.94 Β±46.34 / 2649.99 ms β 2613.50 / 2680.35 Β±51.49 / 2728.30 ms β no change β
β QQuery 33 β 3465.86 / 3509.40 Β±31.64 / 3558.82 ms β 3447.20 / 3544.98 Β±62.03 / 3635.64 ms β no change β
β QQuery 34 β 3469.37 / 3550.53 Β±84.17 / 3673.26 ms β 3443.42 / 3560.98 Β±60.79 / 3608.10 ms β no change β
β QQuery 35 β 2413.31 / 2490.05 Β±49.19 / 2543.80 ms β 2398.03 / 2500.53 Β±77.28 / 2589.96 ms β no change β
β QQuery 36 β 182.12 / 212.84 Β±34.27 / 276.64 ms β 186.89 / 265.06 Β±49.60 / 322.79 ms β 1.25x slower β
β QQuery 37 β 143.71 / 189.86 Β±39.40 / 246.20 ms β 149.96 / 232.63 Β±64.12 / 319.77 ms β 1.23x slower β
β QQuery 38 β 141.98 / 192.21 Β±35.63 / 248.76 ms β 162.34 / 245.10 Β±54.92 / 319.82 ms β 1.28x slower β
β QQuery 39 β 238.26 / 272.94 Β±38.26 / 346.23 ms β 252.39 / 311.00 Β±44.93 / 372.98 ms β 1.14x slower β
β QQuery 40 β 118.05 / 174.89 Β±37.14 / 216.52 ms β 147.19 / 245.50 Β±61.76 / 327.73 ms β 1.40x slower β
β QQuery 41 β 118.75 / 168.82 Β±34.91 / 217.62 ms β 145.85 / 233.44 Β±55.28 / 306.06 ms β 1.38x slower β
β QQuery 42 β 115.18 / 167.98 Β±36.13 / 216.26 ms β 144.74 / 232.38 Β±54.65 / 302.92 ms β 1.38x slower β
βββββββββββββ΄βββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 88917.58ms β
β Total Time (alamb_pushdown_by_default) β 87314.82ms β
β Average Time (HEAD) β 2067.85ms β
β Average Time (alamb_pushdown_by_default) β 2030.58ms β
β Queries Faster β 1 β
β Queries Slower β 10 β
β Queries with No Change β 32 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 450.1s |
| Peak memory | 9.4 GiB |
| Avg memory | 1.8 GiB |
| CPU user | 1003.1s |
| CPU sys | 114.6s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 440.1s |
| Peak memory | 9.3 GiB |
| Avg memory | 1.9 GiB |
| CPU user | 852.5s |
| CPU sys | 112.1s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5315279061
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5315251680)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5315251680-1621-8hdcg 6.12.85+ #1 SMP Wed Jun 17 20:31:55 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (79dd91dc7c0119253e85e64cf0521617cd513680) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..79dd91dc7c0119253e85e64cf0521617cd513680)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5331480824
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5331459712)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5331459712-1656-kxrbk 6.12.85+ #1 SMP Wed Jun 17 20:31:55 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (026fb7d25c01c00144dbf3082a2370c1f6b56b10) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..026fb7d25c01c00144dbf3082a2370c1f6b56b10)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5438849398
run benchmark clickbench_partitioned
```
env:
SIMULATE_LATENCY: "true"
```
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5370315929
Update:
I did some profiling of q25 and I found a few things that could be improved (getting closer)
- https://github.com/apache/arrow-rs/issues/10774
- https://github.com/apache/arrow-rs/issues/10776
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb closed pull request #24426: WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O
URL: https://github.com/apache/datafusion/pull/24426
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-6201494827
Testing after we have done more work upstream
- https://github.com/apache/datafusion/pull/26384
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317646804
FYI: I have pushed 026fb7d which ports the narrow-projection pushdown gate (`pushdown_filter_mode`, default `auto`) from https://github.com/apache/datafusion/pull/23420 (thanks @zhuqi-lucas) onto this branch, to see if it eliminates the remaining regressions.
The two benchmark requests above from 15:21/15:23 UTC were still queued at push time, so check the `Comparing ... (<sha>)` line in their results to see which commit they measured: `79dd91d` is pushdown-by-default only, `026fb7d` adds the gate. Posting fresh requests below for the gated head.
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5333084040
## Status update
This PR now carries three logical changes (see description): `progressive_io` one-shot row-group I/O (#24395), the narrow-projection pushdown gate ported from #23420 (`pushdown_filter_mode`, thanks @zhuqi-lucas), and a refinement (`106653a`) that gates only *static* conjuncts when a dynamic (TopK/aggregate) filter is present, so the dynamic threshold still reaches the scan while e.g. `SearchPhrase <> ''` stays in the `FilterExec` above.
### What the benchmark runs so far show (comparing repeated runs to separate signal from noise)
- The gate eliminates the broad pushdown regressions (Q7/Q10/Q11/Q14/Q17/Q21/Q22/Q34 in earlier runs), and Q23 keeps its ~20x (plain) / ~6x (simulated latency) improvement throughout.
- With `106653a` (dynamic-only gating): **Q24 flips to +1.43x faster** and Q26 drops into the noise band β the TopK threshold on `EventTime` (a cheap int64 compare) prunes rows before decoding `SearchPhrase`.
- **Q25 got worse** (~1.45x slower): it sorts on `SearchPhrase` itself, so the dynamic threshold is a per-row UTF-8 comparison on a column that must be decoded anyway β no decode savings, pure compare overhead. Theory: a dynamic threshold is only worth pushing as a *row* filter when its column is cheap to compare and/or pruning it avoids decoding other columns; string-keyed thresholds on narrow projections should rely on row-group-level pruning only.
- **Q41/Q42 under simulated latency** show a stubbornly consistent ~1.25x (~+28ms) across all three latency runs, unaffected by either gate variant. The near-constant absolute delta looks like one extra round trip somewhere in the pushdown path (aggregate-TopK dynamic filter is a suspect). Not yet explained.
### Next steps
Building `--profile=profiling` binaries at the merge-base, `026fb7d`, and `106653a` to reproduce these locally and profile: `EXPLAIN ANALYZE` on q24/q25/q26/q41/q42, object-store request counts, and one-variable-at-a-time toggles of `pushdown_filter_mode`, `progressive_io`, and `enable_aggregate_dynamic_filter_pushdown`. Results to follow here.
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5332422047
run benchmark clickbench_partitioned
```
env:
SIMULATE_LATENCY: "true"
```
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5438849038
run benchmark clickbench_partitioned
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317451259
> > Oh wait this is just mostly filter pushdown itself I guess
>
> Yes, exactly. However, the results are very close to not regressing with filter pushdown
>
> I think a few more tweaks and we'll have it
>
Amazing
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5332626480
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5332422781)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (106653a393b8815e09c09db239afa966f0f627b3) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..106653a393b8815e09c09db239afa966f0f627b3)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.23 ms β 1.25 ms β no change β
β QQuery 1 β 12.03 ms β 12.15 ms β no change β
β QQuery 2 β 37.21 ms β 37.14 ms β no change β
β QQuery 3 β 31.35 ms β 31.36 ms β no change β
β QQuery 4 β 227.84 ms β 227.39 ms β no change β
β QQuery 5 β 278.15 ms β 279.94 ms β no change β
β QQuery 6 β 1.30 ms β 1.31 ms β no change β
β QQuery 7 β 13.58 ms β 13.48 ms β no change β
β QQuery 8 β 335.97 ms β 332.89 ms β no change β
β QQuery 9 β 449.47 ms β 468.19 ms β no change β
β QQuery 10 β 69.84 ms β 70.45 ms β no change β
β QQuery 11 β 81.61 ms β 82.18 ms β no change β
β QQuery 12 β 273.74 ms β 273.74 ms β no change β
β QQuery 13 β 367.77 ms β 370.43 ms β no change β
β QQuery 14 β 284.28 ms β 289.49 ms β no change β
β QQuery 15 β 279.76 ms β 277.15 ms β no change β
β QQuery 16 β 622.03 ms β 626.26 ms β no change β
β QQuery 17 β 618.01 ms β 624.37 ms β no change β
β QQuery 18 β 1271.05 ms β 1269.58 ms β no change β
β QQuery 19 β 27.63 ms β 27.48 ms β no change β
β QQuery 20 β 519.71 ms β 517.93 ms β no change β
β QQuery 21 β 514.97 ms β 519.67 ms β no change β
β QQuery 22 β 989.31 ms β 993.50 ms β no change β
β QQuery 23 β 3068.66 ms β 150.91 ms β +20.33x faster β
β QQuery 24 β 41.32 ms β 30.32 ms β +1.36x faster β
β QQuery 25 β 110.72 ms β 164.95 ms β 1.49x slower β
β QQuery 26 β 41.81 ms β 48.34 ms β 1.16x slower β
β QQuery 27 β 519.17 ms β 518.27 ms β no change β
β QQuery 28 β 2920.13 ms β 2915.95 ms β no change β
β QQuery 29 β 41.02 ms β 41.88 ms β no change β
β QQuery 30 β 310.42 ms β 313.80 ms β no change β
β QQuery 31 β 283.64 ms β 289.68 ms β no change β
β QQuery 32 β 943.54 ms β 947.54 ms β no change β
β QQuery 33 β 1470.70 ms β 1483.00 ms β no change β
β QQuery 34 β 1476.32 ms β 1500.05 ms β no change β
β QQuery 35 β 289.92 ms β 289.42 ms β no change β
β QQuery 36 β 68.82 ms β 70.05 ms β no change β
β QQuery 37 β 35.76 ms β 36.01 ms β no change β
β QQuery 38 β 41.42 ms β 41.59 ms β no change β
β QQuery 39 β 143.52 ms β 144.19 ms β no change β
β QQuery 40 β 14.58 ms β 14.78 ms β no change β
β QQuery 41 β 13.81 ms β 13.96 ms β no change β
β QQuery 42 β 13.91 ms β 13.63 ms β no change β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 19157.05ms β
β Total Time (alamb_pushdown_by_default) β 16375.65ms β
β Average Time (HEAD) β 445.51ms β
β Average Time (alamb_pushdown_by_default) β 380.83ms β
β Queries Faster β 2 β
β Queries Slower β 2 β
β Queries with No Change β 39 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.23 / 4.05 Β±5.50 / 15.05 ms β 1.25 / 4.02 Β±5.43 / 14.88 ms β no change β
β QQuery 1 β 12.03 / 12.41 Β±0.24 / 12.69 ms β 12.15 / 12.31 Β±0.20 / 12.66 ms β no change β
β QQuery 2 β 37.21 / 37.63 Β±0.25 / 37.97 ms β 37.14 / 37.47 Β±0.35 / 38.15 ms β no change β
β QQuery 3 β 31.35 / 32.29 Β±0.77 / 33.44 ms β 31.36 / 32.11 Β±1.05 / 34.18 ms β no change β
β QQuery 4 β 227.84 / 230.50 Β±1.56 / 232.23 ms β 227.39 / 235.27 Β±4.06 / 238.57 ms β no change β
β QQuery 5 β 278.15 / 280.82 Β±2.58 / 285.30 ms β 279.94 / 281.62 Β±1.74 / 284.95 ms β no change β
β QQuery 6 β 1.30 / 1.45 Β±0.23 / 1.91 ms β 1.31 / 1.46 Β±0.24 / 1.93 ms β no change β
β QQuery 7 β 13.58 / 13.72 Β±0.09 / 13.82 ms β 13.48 / 13.68 Β±0.15 / 13.87 ms β no change β
β QQuery 8 β 335.97 / 341.57 Β±3.88 / 347.85 ms β 332.89 / 338.41 Β±3.42 / 342.84 ms β no change β
β QQuery 9 β 449.47 / 468.96 Β±15.60 / 494.83 ms β 468.19 / 477.94 Β±6.00 / 483.53 ms β no change β
β QQuery 10 β 69.84 / 75.00 Β±5.43 / 82.05 ms β 70.45 / 72.59 Β±1.80 / 75.70 ms β no change β
β QQuery 11 β 81.61 / 82.66 Β±0.69 / 83.54 ms β 82.18 / 82.86 Β±0.73 / 84.07 ms β no change β
β QQuery 12 β 273.74 / 280.07 Β±5.58 / 289.82 ms β 273.74 / 282.05 Β±6.74 / 291.27 ms β no change β
β QQuery 13 β 367.77 / 384.13 Β±10.70 / 396.54 ms β 370.43 / 394.78 Β±21.92 / 430.18 ms β no change β
β QQuery 14 β 284.28 / 290.11 Β±7.12 / 303.67 ms β 289.49 / 301.74 Β±12.23 / 323.56 ms β no change β
β QQuery 15 β 279.76 / 287.96 Β±6.67 / 297.50 ms β 277.15 / 286.18 Β±9.06 / 298.67 ms β no change β
β QQuery 16 β 622.03 / 629.11 Β±6.26 / 639.45 ms β 626.26 / 641.80 Β±11.58 / 657.23 ms β no change β
β QQuery 17 β 618.01 / 634.21 Β±10.25 / 647.90 ms β 624.37 / 633.67 Β±6.31 / 639.07 ms β no change β
β QQuery 18 β 1271.05 / 1294.26 Β±14.88 / 1310.57 ms β 1269.58 / 1297.21 Β±23.10 / 1336.66 ms β no change β
β QQuery 19 β 27.63 / 27.99 Β±0.38 / 28.58 ms β 27.48 / 36.03 Β±10.62 / 53.85 ms β 1.29x slower β
β QQuery 20 β 519.71 / 528.41 Β±9.66 / 545.01 ms β 517.93 / 526.29 Β±6.70 / 536.40 ms β no change β
β QQuery 21 β 514.97 / 524.87 Β±9.77 / 541.62 ms β 519.67 / 524.23 Β±5.23 / 531.42 ms β no change β
β QQuery 22 β 989.31 / 1000.52 Β±8.69 / 1011.21 ms β 993.50 / 1003.23 Β±6.20 / 1012.80 ms β no change β
β QQuery 23 β 3068.66 / 3114.82 Β±32.78 / 3171.41 ms β 150.91 / 163.08 Β±14.26 / 190.70 ms β +19.10x faster β
β QQuery 24 β 41.32 / 43.51 Β±1.89 / 46.29 ms β 30.32 / 30.43 Β±0.08 / 30.58 ms β +1.43x faster β
β QQuery 25 β 110.72 / 116.19 Β±8.96 / 134.06 ms β 164.95 / 168.74 Β±4.75 / 178.07 ms β 1.45x slower β
β QQuery 26 β 41.81 / 46.89 Β±9.72 / 66.32 ms β 48.34 / 50.37 Β±2.97 / 56.24 ms β 1.07x slower β
β QQuery 27 β 519.17 / 524.70 Β±6.68 / 537.43 ms β 518.27 / 527.30 Β±8.77 / 542.99 ms β no change β
β QQuery 28 β 2920.13 / 2937.28 Β±9.38 / 2947.03 ms β 2915.95 / 2935.42 Β±14.77 / 2961.62 ms β no change β
β QQuery 29 β 41.02 / 45.95 Β±9.17 / 64.28 ms β 41.88 / 46.95 Β±9.70 / 66.35 ms β no change β
β QQuery 30 β 310.42 / 317.85 Β±8.19 / 333.82 ms β 313.80 / 317.67 Β±4.86 / 326.64 ms β no change β
β QQuery 31 β 283.64 / 295.01 Β±6.64 / 302.71 ms β 289.68 / 298.46 Β±6.12 / 307.57 ms β no change β
β QQuery 32 β 943.54 / 955.48 Β±11.49 / 974.90 ms β 947.54 / 973.43 Β±17.19 / 1000.39 ms β no change β
β QQuery 33 β 1470.70 / 1489.13 Β±15.84 / 1513.46 ms β 1483.00 / 1493.59 Β±5.56 / 1498.38 ms β no change β
β QQuery 34 β 1476.32 / 1526.34 Β±27.06 / 1550.87 ms β 1500.05 / 1532.24 Β±23.09 / 1570.68 ms β no change β
β QQuery 35 β 289.92 / 308.55 Β±24.67 / 355.35 ms β 289.42 / 315.74 Β±45.08 / 405.82 ms β no change β
β QQuery 36 β 68.82 / 74.07 Β±3.98 / 81.04 ms β 70.05 / 74.10 Β±5.47 / 84.45 ms β no change β
β QQuery 37 β 35.76 / 42.42 Β±5.97 / 49.97 ms β 36.01 / 42.45 Β±6.05 / 50.60 ms β no change β
β QQuery 38 β 41.42 / 44.20 Β±1.87 / 47.19 ms β 41.59 / 43.88 Β±2.58 / 48.74 ms β no change β
β QQuery 39 β 143.52 / 151.51 Β±6.87 / 160.13 ms β 144.19 / 148.11 Β±4.93 / 157.58 ms β no change β
β QQuery 40 β 14.58 / 14.78 Β±0.28 / 15.33 ms β 14.78 / 16.45 Β±2.15 / 20.60 ms β 1.11x slower β
β QQuery 41 β 13.81 / 14.02 Β±0.19 / 14.29 ms β 13.96 / 14.18 Β±0.19 / 14.50 ms β no change β
β QQuery 42 β 13.91 / 15.19 Β±2.27 / 19.73 ms β 13.63 / 13.74 Β±0.15 / 14.04 ms β +1.11x faster β
βββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 19540.61ms β
β Total Time (alamb_pushdown_by_default) β 16723.30ms β
β Average Time (HEAD) β 454.43ms β
β Average Time (alamb_pushdown_by_default) β 388.91ms β
β Queries Faster β 3 β
β Queries Slower β 4 β
β Queries with No Change β 36 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 11.5 GiB |
| Avg memory | 4.5 GiB |
| CPU user | 1007.7s |
| CPU sys | 68.6s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 85.0s |
| Peak memory | 11.3 GiB |
| Avg memory | 4.4 GiB |
| CPU user | 850.9s |
| CPU sys | 57.9s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5342167688
## Local Q25 experiment: the regression is real, and the dynamic-only refinement makes it worse
To take benchmark-runner noise out of the picture, I built `--profile=profiling` binaries at three commits β the merge-base (`00eba79`), the gate (`026fb7d`), and the dynamic-only refinement (`106653a`) β and ran ClickBench Q25 locally against `hits_partitioned` on a quiet machine, replicating the harness setup (`binary_as_string=true`, the `hits` view with the `EventDate` cast, `reorder_filters` left at default). Each cell below is 10 runs Γ 3 interleaved rounds (30 runs) after 2 warmups; round-to-round averages agreed within ~1ms, stdev β€3ms.
Q25: `SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;`
| Binary | pushdown ON avg (min) ms | pushdown OFF avg (min) ms | ON/OFF |
|---|---|---|---|
| merge-base `00eba79` | 107.8 (104) | 86.4 (84) | 1.25x |
| gate `026fb7d` | 110.8 (108) | 88.1 (86) | 1.26x |
| dynamic-only `106653a` | 115.9 (113) | 85.2 (83) | **1.36x** |
Conclusions:
1. **The Q25 pushdown penalty pre-exists this PR**: enabling pushdown costs ~25% even on the merge-base. Flipping the default merely exposes it.
2. **The dynamic-only refinement (`106653a`) is a real loss on Q25** (+5ms over `026fb7d`, +8ms over base, reproduced every round) β moving `SearchPhrase <> ''` out of the RowFilter while keeping the string-keyed TopK threshold in it is the wrong trade for this shape. Notably `026fb7d`-ON β base-ON, so the gate's all-or-nothing dynamic carve-out was not the main cost for Q25 either β the RowFilter machinery on this query shape (filter column = sort column = only projected column) is.
3. Pushdown OFF is identical across all three binaries (~85β88ms): the changes are clean no-ops when pushdown is disabled.
**Decision: reverting `106653a` from this branch.** The Q24 improvement it showed in the bot run needs re-validation after understanding the real mechanism; the next step is profiling the ~20β30ms ON-vs-OFF delta that exists on *every* binary for Q25 (filter-phase decode, selection building, predicate cache round-trip, and the per-row threshold compare are the candidates), plus the same local matrix on Q24 (int sort key) for contrast.
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5331459712
run benchmark clickbench_partitioned
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5331512757
run benchmark clickbench_partitioned
```
env:
SIMULATE_LATENCY: "true"
```
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5439166568
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5438849038)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (1eb7f67b4c8822c26dfab75fcc303cb2f2f0702d) to a38bb10 (merge-base) [diff](https://github.com/apache/datafusion/compare/a38bb10c33cff4e99a2707059df99f6971c96f33..1eb7f67b4c8822c26dfab75fcc303cb2f2f0702d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.20 ms β 1.20 ms β no change β
β QQuery 1 β 11.89 ms β 11.77 ms β no change β
β QQuery 2 β 36.47 ms β 35.78 ms β no change β
β QQuery 3 β 30.92 ms β 31.87 ms β no change β
β QQuery 4 β 222.85 ms β 250.59 ms β 1.12x slower β
β QQuery 5 β 272.39 ms β 274.59 ms β no change β
β QQuery 6 β 1.24 ms β 1.27 ms β no change β
β QQuery 7 β 13.44 ms β 13.50 ms β no change β
β QQuery 8 β 323.00 ms β 331.59 ms β no change β
β QQuery 9 β 455.26 ms β 454.14 ms β no change β
β QQuery 10 β 69.67 ms β 70.64 ms β no change β
β QQuery 11 β 80.46 ms β 81.19 ms β no change β
β QQuery 12 β 266.04 ms β 269.83 ms β no change β
β QQuery 13 β 362.98 ms β 370.57 ms β no change β
β QQuery 14 β 281.67 ms β 284.15 ms β no change β
β QQuery 15 β 265.71 ms β 267.06 ms β no change β
β QQuery 16 β 650.94 ms β 610.11 ms β +1.07x faster β
β QQuery 17 β 714.06 ms β 627.16 ms β +1.14x faster β
β QQuery 18 β 1298.43 ms β 1271.40 ms β no change β
β QQuery 19 β 27.82 ms β 27.72 ms β no change β
β QQuery 20 β 512.41 ms β 511.93 ms β no change β
β QQuery 21 β 523.00 ms β 512.74 ms β no change β
β QQuery 22 β 978.14 ms β 1008.15 ms β no change β
β QQuery 23 β 3031.55 ms β 161.66 ms β +18.75x faster β
β QQuery 24 β 40.88 ms β 57.92 ms β 1.42x slower β
β QQuery 25 β 109.72 ms β 155.53 ms β 1.42x slower β
β QQuery 26 β 41.33 ms β 62.95 ms β 1.52x slower β
β QQuery 27 β 514.73 ms β 541.28 ms β 1.05x slower β
β QQuery 28 β 2879.91 ms β 2917.88 ms β no change β
β QQuery 29 β 40.62 ms β 40.59 ms β no change β
β QQuery 30 β 302.94 ms β 301.62 ms β no change β
β QQuery 31 β 279.23 ms β 290.88 ms β no change β
β QQuery 32 β 1102.08 ms β 957.62 ms β +1.15x faster β
β QQuery 33 β 1489.37 ms β 1437.56 ms β no change β
β QQuery 34 β 1539.28 ms β 1461.10 ms β +1.05x faster β
β QQuery 35 β 276.40 ms β 281.40 ms β no change β
β QQuery 36 β 69.61 ms β 67.10 ms β no change β
β QQuery 37 β 35.45 ms β 35.25 ms β no change β
β QQuery 38 β 41.53 ms β 45.07 ms β 1.09x slower β
β QQuery 39 β 132.27 ms β 130.69 ms β no change β
β QQuery 40 β 14.03 ms β 14.02 ms β no change β
β QQuery 41 β 13.56 ms β 13.53 ms β no change β
β QQuery 42 β 13.14 ms β 13.05 ms β no change β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 19367.64ms β
β Total Time (alamb_pushdown_by_default) β 16305.67ms β
β Average Time (HEAD) β 450.41ms β
β Average Time (alamb_pushdown_by_default) β 379.20ms β
β Queries Faster β 5 β
β Queries Slower β 6 β
β Queries with No Change β 32 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³βββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.20 / 3.89 Β±5.31 / 14.51 ms β 1.20 / 3.85 Β±5.24 / 14.32 ms β no change β
β QQuery 1 β 11.89 / 12.16 Β±0.14 / 12.32 ms β 11.77 / 12.04 Β±0.19 / 12.24 ms β no change β
β QQuery 2 β 36.47 / 36.80 Β±0.30 / 37.32 ms β 35.78 / 36.11 Β±0.31 / 36.63 ms β no change β
β QQuery 3 β 30.92 / 31.86 Β±1.02 / 33.81 ms β 31.87 / 32.32 Β±0.52 / 33.32 ms β no change β
β QQuery 4 β 222.85 / 233.18 Β±10.50 / 251.41 ms β 250.59 / 257.31 Β±5.75 / 265.50 ms β 1.10x slower β
β QQuery 5 β 272.39 / 276.48 Β±3.89 / 283.40 ms β 274.59 / 278.19 Β±1.95 / 279.98 ms β no change β
β QQuery 6 β 1.24 / 1.40 Β±0.23 / 1.85 ms β 1.27 / 1.42 Β±0.22 / 1.85 ms β no change β
β QQuery 7 β 13.44 / 13.58 Β±0.10 / 13.70 ms β 13.50 / 15.99 Β±4.85 / 25.69 ms β 1.18x slower β
β QQuery 8 β 323.00 / 330.43 Β±5.28 / 336.62 ms β 331.59 / 340.90 Β±9.05 / 356.14 ms β no change β
β QQuery 9 β 455.26 / 460.47 Β±6.56 / 472.95 ms β 454.14 / 469.21 Β±12.64 / 486.76 ms β no change β
β QQuery 10 β 69.67 / 70.22 Β±0.51 / 71.13 ms β 70.64 / 72.62 Β±2.66 / 77.78 ms β no change β
β QQuery 11 β 80.46 / 81.82 Β±0.92 / 83.22 ms β 81.19 / 81.83 Β±0.34 / 82.11 ms β no change β
β QQuery 12 β 266.04 / 270.50 Β±3.54 / 275.02 ms β 269.83 / 272.77 Β±2.05 / 275.80 ms β no change β
β QQuery 13 β 362.98 / 374.91 Β±13.97 / 400.89 ms β 370.57 / 380.76 Β±7.63 / 392.58 ms β no change β
β QQuery 14 β 281.67 / 287.53 Β±5.35 / 295.40 ms β 284.15 / 295.24 Β±11.83 / 317.89 ms β no change β
β QQuery 15 β 265.71 / 275.86 Β±7.32 / 285.05 ms β 267.06 / 276.26 Β±6.13 / 283.70 ms β no change β
β QQuery 16 β 650.94 / 695.78 Β±23.09 / 714.92 ms β 610.11 / 643.68 Β±32.88 / 697.73 ms β +1.08x faster β
β QQuery 17 β 714.06 / 721.73 Β±5.10 / 729.30 ms β 627.16 / 647.08 Β±16.53 / 672.12 ms β +1.12x faster β
β QQuery 18 β 1298.43 / 1337.57 Β±34.84 / 1394.69 ms β 1271.40 / 1289.69 Β±12.54 / 1308.89 ms β no change β
β QQuery 19 β 27.82 / 38.36 Β±12.39 / 55.48 ms β 27.72 / 28.15 Β±0.37 / 28.75 ms β +1.36x faster β
β QQuery 20 β 512.41 / 534.93 Β±12.64 / 546.48 ms β 511.93 / 523.75 Β±11.81 / 543.04 ms β no change β
β QQuery 21 β 523.00 / 529.30 Β±6.17 / 540.08 ms β 512.74 / 519.51 Β±4.89 / 525.41 ms β no change β
β QQuery 22 β 978.14 / 992.43 Β±9.48 / 1002.09 ms β 1008.15 / 1057.53 Β±25.24 / 1077.41 ms β 1.07x slower β
β QQuery 23 β 3031.55 / 3081.44 Β±45.89 / 3148.94 ms β 161.66 / 168.30 Β±5.37 / 176.72 ms β +18.31x faster β
β QQuery 24 β 40.88 / 48.61 Β±9.25 / 63.96 ms β 57.92 / 62.55 Β±7.81 / 78.10 ms β 1.29x slower β
β QQuery 25 β 109.72 / 111.58 Β±1.15 / 113.26 ms β 155.53 / 157.84 Β±1.96 / 160.58 ms β 1.41x slower β
β QQuery 26 β 41.33 / 42.36 Β±0.94 / 44.00 ms β 62.95 / 65.02 Β±1.39 / 66.89 ms β 1.53x slower β
β QQuery 27 β 514.73 / 531.04 Β±10.79 / 547.22 ms β 541.28 / 549.90 Β±4.42 / 553.88 ms β no change β
β QQuery 28 β 2879.91 / 2912.10 Β±22.35 / 2942.88 ms β 2917.88 / 2954.52 Β±27.14 / 2998.67 ms β no change β
β QQuery 29 β 40.62 / 43.84 Β±5.69 / 55.17 ms β 40.59 / 50.32 Β±12.93 / 74.09 ms β 1.15x slower β
β QQuery 30 β 302.94 / 307.31 Β±3.48 / 311.63 ms β 301.62 / 308.45 Β±6.29 / 318.44 ms β no change β
β QQuery 31 β 279.23 / 299.75 Β±25.77 / 347.42 ms β 290.88 / 296.17 Β±3.21 / 299.71 ms β no change β
β QQuery 32 β 1102.08 / 1120.54 Β±18.16 / 1149.36 ms β 957.62 / 1026.01 Β±58.53 / 1126.88 ms β +1.09x faster β
β QQuery 33 β 1489.37 / 1560.90 Β±81.84 / 1703.68 ms β 1437.56 / 1465.44 Β±19.37 / 1486.14 ms β +1.07x faster β
β QQuery 34 β 1539.28 / 1574.02 Β±34.85 / 1638.67 ms β 1461.10 / 1540.10 Β±84.37 / 1701.50 ms β no change β
β QQuery 35 β 276.40 / 316.81 Β±61.78 / 438.14 ms β 281.40 / 308.39 Β±36.04 / 377.94 ms β no change β
β QQuery 36 β 69.61 / 80.68 Β±6.51 / 86.56 ms β 67.10 / 68.96 Β±1.01 / 70.08 ms β +1.17x faster β
β QQuery 37 β 35.45 / 38.09 Β±4.28 / 46.63 ms β 35.25 / 38.14 Β±3.09 / 42.82 ms β no change β
β QQuery 38 β 41.53 / 45.43 Β±2.49 / 48.61 ms β 45.07 / 47.92 Β±3.00 / 53.07 ms β 1.05x slower β
β QQuery 39 β 132.27 / 147.52 Β±14.67 / 174.57 ms β 130.69 / 137.80 Β±5.14 / 143.80 ms β +1.07x faster β
β QQuery 40 β 14.03 / 14.42 Β±0.33 / 14.83 ms β 14.02 / 16.09 Β±3.49 / 23.06 ms β 1.12x slower β
β QQuery 41 β 13.56 / 14.37 Β±1.38 / 17.12 ms β 13.53 / 17.54 Β±4.74 / 26.09 ms β 1.22x slower β
β QQuery 42 β 13.14 / 13.26 Β±0.09 / 13.36 ms β 13.05 / 13.96 Β±1.54 / 17.04 ms β 1.05x slower β
βββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 19915.24ms β
β Total Time (alamb_pushdown_by_default) β 16829.65ms β
β Average Time (HEAD) β 463.15ms β
β Average Time (alamb_pushdown_by_default) β 391.39ms β
β Queries Faster β 8 β
β Queries Slower β 11 β
β Queries with No Change β 24 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 100.0s |
| Peak memory | 11.3 GiB |
| Avg memory | 4.7 GiB |
| CPU user | 1020.7s |
| CPU sys | 72.6s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 85.0s |
| Peak memory | 12.3 GiB |
| Avg memory | 4.5 GiB |
| CPU user | 852.7s |
| CPU sys | 60.5s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5438472525
I am now going to build and profile the following combinations
1. Merge base (aka main) with no changes
2. `pushdown_filters` on by default with `progressive_io`
2. `pushdown_filters` and `reorder_filters` on by default with `progressive_io`
3. `pushdown_filters` and `reorder_filters` on by default, and the gate from [#23420](https://github.com/apache/datafusion/pull/23420)
3. `pushdown_filters` and `reorder_filters` on by default, and the gate from [#23420](https://github.com/apache/datafusion/pull/23420) AND [#arrow/10852](https://github.com/apache/arrow-rs/pull/10852) from @hhhizzz
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5444557100
I reran the results and 10852 is looking good
```text
ββββββββββββββββββββββββββββββββ¬ββββββββ¬ββββββββ¬ββββββββββ
β Combination β min β avg β vs main β
ββββββββββββββββββββββββββββββββΌββββββββΌββββββββΌββββββββββ€
β 1. main β 81.0 β 86.1 β 1.00Γ β
ββββββββββββββββββββββββββββββββΌββββββββΌββββββββΌββββββββββ€
β 2. + progressive_io/pushdown β 104.0 β 108.5 β 1.26Γ β
ββββββββββββββββββββββββββββββββΌββββββββΌββββββββΌββββββββββ€
β 3. + reorder_filters β 104.0 β 106.7 β 1.24Γ β
ββββββββββββββββββββββββββββββββΌββββββββΌββββββββΌββββββββββ€
β 4. + gate β 109.0 β 112.2 β 1.30Γ β
ββββββββββββββββββββββββββββββββΌββββββββΌββββββββΌββββββββββ€
β 5. gate + #10852 β 97.0 β 104.6 β 1.21Γ β
ββββββββββββββββββββββββββββββββΌββββββββΌββββββββΌββββββββββ€
β 6. no gate + #10852 β 94.0 β 100.3 β 1.16Γ β
ββββββββββββββββββββββββββββββββ΄ββββββββ΄ββββββββ΄ββββββββββ
```
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317078983
Oh wait this is just mostly filter pushdown itself I guess
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5332748043
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5332422047)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (106653a393b8815e09c09db239afa966f0f627b3) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..106653a393b8815e09c09db239afa966f0f627b3)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.26 ms β 1.23 ms β no change β
β QQuery 1 β 1871.77 ms β 1843.75 ms β no change β
β QQuery 2 β 2160.56 ms β 2167.08 ms β no change β
β QQuery 3 β 2159.06 ms β 2150.34 ms β no change β
β QQuery 4 β 2295.58 ms β 2307.74 ms β no change β
β QQuery 5 β 2384.43 ms β 2367.97 ms β no change β
β QQuery 6 β 1.26 ms β 1.30 ms β no change β
β QQuery 7 β 1947.43 ms β 1930.49 ms β no change β
β QQuery 8 β 2428.16 ms β 2430.06 ms β no change β
β QQuery 9 β 2614.44 ms β 2621.92 ms β no change β
β QQuery 10 β 2237.50 ms β 2173.87 ms β no change β
β QQuery 11 β 2216.64 ms β 2226.80 ms β no change β
β QQuery 12 β 2329.77 ms β 2348.00 ms β no change β
β QQuery 13 β 2457.92 ms β 2498.13 ms β no change β
β QQuery 14 β 2348.71 ms β 2400.56 ms β no change β
β QQuery 15 β 2376.89 ms β 2385.19 ms β no change β
β QQuery 16 β 2666.58 ms β 2652.67 ms β no change β
β QQuery 17 β 2672.98 ms β 2690.51 ms β no change β
β QQuery 18 β 3124.09 ms β 3109.43 ms β no change β
β QQuery 19 β 1798.35 ms β 1770.31 ms β no change β
β QQuery 20 β 2682.31 ms β 2628.05 ms β no change β
β QQuery 21 β 2628.76 ms β 2606.78 ms β no change β
β QQuery 22 β 3083.78 ms β 3065.96 ms β no change β
β QQuery 23 β 4828.18 ms β 824.21 ms β +5.86x faster β
β QQuery 24 β 476.63 ms β 514.81 ms β 1.08x slower β
β QQuery 25 β 2248.04 ms β 2312.26 ms β no change β
β QQuery 26 β 476.49 ms β 507.45 ms β 1.06x slower β
β QQuery 27 β 2720.84 ms β 2675.52 ms β no change β
β QQuery 28 β 4937.12 ms β 5004.37 ms β no change β
β QQuery 29 β 2153.17 ms β 2217.85 ms β no change β
β QQuery 30 β 2483.32 ms β 2461.64 ms β no change β
β QQuery 31 β 2325.68 ms β 2427.11 ms β no change β
β QQuery 32 β 2528.16 ms β 2586.05 ms β no change β
β QQuery 33 β 3445.53 ms β 3442.57 ms β no change β
β QQuery 34 β 3418.43 ms β 3477.62 ms β no change β
β QQuery 35 β 2369.64 ms β 2332.33 ms β no change β
β QQuery 36 β 185.35 ms β 191.78 ms β no change β
β QQuery 37 β 142.68 ms β 146.56 ms β no change β
β QQuery 38 β 147.33 ms β 178.79 ms β 1.21x slower β
β QQuery 39 β 247.21 ms β 254.13 ms β no change β
β QQuery 40 β 151.08 ms β 148.48 ms β no change β
β QQuery 41 β 115.89 ms β 143.44 ms β 1.24x slower β
β QQuery 42 β 114.42 ms β 143.84 ms β 1.26x slower β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 86003.43ms β
β Total Time (alamb_pushdown_by_default) β 82368.96ms β
β Average Time (HEAD) β 2000.08ms β
β Average Time (alamb_pushdown_by_default) β 1915.56ms β
β Queries Faster β 1 β
β Queries Slower β 5 β
β Queries with No Change β 37 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.26 / 4.00 Β±5.43 / 14.86 ms β 1.23 / 4.00 Β±5.47 / 14.93 ms β no change β
β QQuery 1 β 1871.77 / 1921.47 Β±35.02 / 1966.01 ms β 1843.75 / 1900.52 Β±34.90 / 1953.65 ms β no change β
β QQuery 2 β 2160.56 / 2234.26 Β±45.17 / 2302.51 ms β 2167.08 / 2266.93 Β±58.74 / 2327.74 ms β no change β
β QQuery 3 β 2159.06 / 2221.57 Β±34.46 / 2249.09 ms β 2150.34 / 2188.43 Β±51.30 / 2289.60 ms β no change β
β QQuery 4 β 2295.58 / 2359.57 Β±51.05 / 2433.88 ms β 2307.74 / 2363.86 Β±28.30 / 2383.20 ms β no change β
β QQuery 5 β 2384.43 / 2464.84 Β±66.04 / 2544.27 ms β 2367.97 / 2429.85 Β±47.36 / 2481.49 ms β no change β
β QQuery 6 β 1.26 / 1.41 Β±0.22 / 1.84 ms β 1.30 / 1.44 Β±0.22 / 1.88 ms β no change β
β QQuery 7 β 1947.43 / 2026.97 Β±51.14 / 2103.19 ms β 1930.49 / 1985.02 Β±43.41 / 2050.03 ms β no change β
β QQuery 8 β 2428.16 / 2476.90 Β±33.72 / 2527.32 ms β 2430.06 / 2508.86 Β±54.53 / 2591.27 ms β no change β
β QQuery 9 β 2614.44 / 2732.09 Β±66.16 / 2789.50 ms β 2621.92 / 2687.77 Β±51.16 / 2764.94 ms β no change β
β QQuery 10 β 2237.50 / 2297.99 Β±33.18 / 2331.06 ms β 2173.87 / 2234.02 Β±41.13 / 2289.85 ms β no change β
β QQuery 11 β 2216.64 / 2288.46 Β±55.76 / 2352.44 ms β 2226.80 / 2258.92 Β±37.47 / 2332.18 ms β no change β
β QQuery 12 β 2329.77 / 2381.81 Β±43.13 / 2452.21 ms β 2348.00 / 2441.03 Β±81.26 / 2548.92 ms β no change β
β QQuery 13 β 2457.92 / 2526.79 Β±52.73 / 2620.96 ms β 2498.13 / 2554.39 Β±46.14 / 2617.04 ms β no change β
β QQuery 14 β 2348.71 / 2407.00 Β±41.67 / 2478.67 ms β 2400.56 / 2445.12 Β±35.05 / 2491.32 ms β no change β
β QQuery 15 β 2376.89 / 2438.09 Β±41.15 / 2485.79 ms β 2385.19 / 2416.60 Β±24.35 / 2455.48 ms β no change β
β QQuery 16 β 2666.58 / 2719.47 Β±45.11 / 2771.12 ms β 2652.67 / 2729.79 Β±92.20 / 2906.59 ms β no change β
β QQuery 17 β 2672.98 / 2716.41 Β±46.91 / 2797.15 ms β 2690.51 / 2729.81 Β±40.64 / 2808.13 ms β no change β
β QQuery 18 β 3124.09 / 3243.00 Β±76.16 / 3353.98 ms β 3109.43 / 3182.96 Β±74.10 / 3314.73 ms β no change β
β QQuery 19 β 1798.35 / 1819.72 Β±21.87 / 1858.85 ms β 1770.31 / 1834.36 Β±51.94 / 1923.37 ms β no change β
β QQuery 20 β 2682.31 / 2731.75 Β±42.11 / 2802.38 ms β 2628.05 / 2685.56 Β±50.72 / 2773.78 ms β no change β
β QQuery 21 β 2628.76 / 2681.80 Β±28.81 / 2703.73 ms β 2606.78 / 2665.40 Β±35.11 / 2702.49 ms β no change β
β QQuery 22 β 3083.78 / 3183.94 Β±59.57 / 3262.62 ms β 3065.96 / 3151.24 Β±67.58 / 3254.06 ms β no change β
β QQuery 23 β 4828.18 / 5046.85 Β±133.92 / 5189.33 ms β 824.21 / 889.70 Β±38.24 / 939.47 ms β +5.67x faster β
β QQuery 24 β 476.63 / 501.36 Β±29.95 / 558.27 ms β 514.81 / 556.54 Β±45.81 / 645.34 ms β 1.11x slower β
β QQuery 25 β 2248.04 / 2321.99 Β±49.59 / 2404.04 ms β 2312.26 / 2362.42 Β±26.99 / 2391.93 ms β no change β
β QQuery 26 β 476.49 / 509.52 Β±40.79 / 589.90 ms β 507.45 / 555.63 Β±36.19 / 614.47 ms β 1.09x slower β
β QQuery 27 β 2720.84 / 2737.26 Β±14.13 / 2759.84 ms β 2675.52 / 2765.12 Β±54.47 / 2833.91 ms β no change β
β QQuery 28 β 4937.12 / 5018.74 Β±60.48 / 5118.95 ms β 5004.37 / 5033.66 Β±27.10 / 5084.24 ms β no change β
β QQuery 29 β 2153.17 / 2243.26 Β±45.99 / 2283.43 ms β 2217.85 / 2276.23 Β±35.40 / 2313.28 ms β no change β
β QQuery 30 β 2483.32 / 2513.11 Β±29.75 / 2568.44 ms β 2461.64 / 2546.85 Β±50.51 / 2620.48 ms β no change β
β QQuery 31 β 2325.68 / 2434.34 Β±61.42 / 2511.51 ms β 2427.11 / 2489.42 Β±51.52 / 2576.66 ms β no change β
β QQuery 32 β 2528.16 / 2629.23 Β±55.71 / 2688.46 ms β 2586.05 / 2651.18 Β±41.69 / 2699.57 ms β no change β
β QQuery 33 β 3445.53 / 3510.09 Β±75.79 / 3657.91 ms β 3442.57 / 3498.73 Β±45.08 / 3556.75 ms β no change β
β QQuery 34 β 3418.43 / 3574.91 Β±101.66 / 3701.56 ms β 3477.62 / 3545.14 Β±92.65 / 3728.72 ms β no change β
β QQuery 35 β 2369.64 / 2445.59 Β±63.32 / 2561.51 ms β 2332.33 / 2445.81 Β±63.19 / 2509.64 ms β no change β
β QQuery 36 β 185.35 / 220.87 Β±38.25 / 293.21 ms β 191.78 / 284.39 Β±61.95 / 377.43 ms β 1.29x slower β
β QQuery 37 β 142.68 / 191.18 Β±38.51 / 248.36 ms β 146.56 / 235.47 Β±55.33 / 307.21 ms β 1.23x slower β
β QQuery 38 β 147.33 / 194.52 Β±33.95 / 248.82 ms β 178.79 / 265.34 Β±55.83 / 338.10 ms β 1.36x slower β
β QQuery 39 β 247.21 / 286.51 Β±35.70 / 351.34 ms β 254.13 / 311.37 Β±49.25 / 372.83 ms β 1.09x slower β
β QQuery 40 β 151.08 / 180.24 Β±31.36 / 233.11 ms β 148.48 / 237.97 Β±61.27 / 318.82 ms β 1.32x slower β
β QQuery 41 β 115.89 / 175.35 Β±45.32 / 250.52 ms β 143.44 / 218.43 Β±65.16 / 304.51 ms β 1.25x slower β
β QQuery 42 β 114.42 / 167.19 Β±36.02 / 216.25 ms β 143.84 / 231.34 Β±54.49 / 301.17 ms β 1.38x slower β
βββββββββββββ΄βββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 88781.43ms β
β Total Time (alamb_pushdown_by_default) β 85066.61ms β
β Average Time (HEAD) β 2064.68ms β
β Average Time (alamb_pushdown_by_default) β 1978.29ms β
β Queries Faster β 1 β
β Queries Slower β 9 β
β Queries with No Change β 33 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 445.1s |
| Peak memory | 9.5 GiB |
| Avg memory | 1.8 GiB |
| CPU user | 1003.9s |
| CPU sys | 114.7s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 430.1s |
| Peak memory | 9.4 GiB |
| Avg memory | 1.8 GiB |
| CPU user | 852.8s |
| CPU sys | 100.8s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5331813959
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5331512757)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (026fb7d25c01c00144dbf3082a2370c1f6b56b10) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..026fb7d25c01c00144dbf3082a2370c1f6b56b10)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.25 ms β 1.22 ms β no change β
β QQuery 1 β 1854.32 ms β 1854.45 ms β no change β
β QQuery 2 β 2137.20 ms β 2151.07 ms β no change β
β QQuery 3 β 2166.47 ms β 2198.93 ms β no change β
β QQuery 4 β 2300.69 ms β 2302.18 ms β no change β
β QQuery 5 β 2349.06 ms β 2353.96 ms β no change β
β QQuery 6 β 1.26 ms β 1.29 ms β no change β
β QQuery 7 β 1894.01 ms β 1953.37 ms β no change β
β QQuery 8 β 2374.38 ms β 2447.38 ms β no change β
β QQuery 9 β 2617.56 ms β 2630.52 ms β no change β
β QQuery 10 β 2203.57 ms β 2235.82 ms β no change β
β QQuery 11 β 2194.41 ms β 2194.34 ms β no change β
β QQuery 12 β 2362.45 ms β 2382.09 ms β no change β
β QQuery 13 β 2439.86 ms β 2477.32 ms β no change β
β QQuery 14 β 2361.69 ms β 2409.51 ms β no change β
β QQuery 15 β 2416.88 ms β 2415.11 ms β no change β
β QQuery 16 β 2606.97 ms β 2710.07 ms β no change β
β QQuery 17 β 2634.06 ms β 2655.58 ms β no change β
β QQuery 18 β 3157.94 ms β 3193.49 ms β no change β
β QQuery 19 β 1798.49 ms β 1766.70 ms β no change β
β QQuery 20 β 2703.82 ms β 2598.89 ms β no change β
β QQuery 21 β 2669.62 ms β 2593.09 ms β no change β
β QQuery 22 β 3053.31 ms β 3122.97 ms β no change β
β QQuery 23 β 4832.07 ms β 762.05 ms β +6.34x faster β
β QQuery 24 β 437.05 ms β 592.89 ms β 1.36x slower β
β QQuery 25 β 2248.05 ms β 2291.17 ms β no change β
β QQuery 26 β 482.88 ms β 665.40 ms β 1.38x slower β
β QQuery 27 β 2656.75 ms β 2643.30 ms β no change β
β QQuery 28 β 4957.17 ms β 4935.80 ms β no change β
β QQuery 29 β 2190.14 ms β 2223.51 ms β no change β
β QQuery 30 β 2424.38 ms β 2439.81 ms β no change β
β QQuery 31 β 2381.38 ms β 2410.69 ms β no change β
β QQuery 32 β 2556.29 ms β 2496.58 ms β no change β
β QQuery 33 β 3413.96 ms β 3402.28 ms β no change β
β QQuery 34 β 3413.20 ms β 3408.91 ms β no change β
β QQuery 35 β 2342.13 ms β 2376.63 ms β no change β
β QQuery 36 β 187.12 ms β 195.72 ms β no change β
β QQuery 37 β 137.54 ms β 148.56 ms β 1.08x slower β
β QQuery 38 β 176.50 ms β 176.56 ms β no change β
β QQuery 39 β 235.64 ms β 249.80 ms β 1.06x slower β
β QQuery 40 β 116.03 ms β 218.71 ms β 1.88x slower β
β QQuery 41 β 118.75 ms β 147.01 ms β 1.24x slower β
β QQuery 42 β 115.00 ms β 144.23 ms β 1.25x slower β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 85721.31ms β
β Total Time (alamb_pushdown_by_default) β 82578.97ms β
β Average Time (HEAD) β 1993.52ms β
β Average Time (alamb_pushdown_by_default) β 1920.44ms β
β Queries Faster β 1 β
β Queries Slower β 7 β
β Queries with No Change β 35 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββββββββββββββββββββββββββββββ³βββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.25 / 4.06 Β±5.51 / 15.09 ms β 1.22 / 4.01 Β±5.46 / 14.94 ms β no change β
β QQuery 1 β 1854.32 / 1919.10 Β±40.36 / 1978.32 ms β 1854.45 / 1939.13 Β±55.28 / 2020.40 ms β no change β
β QQuery 2 β 2137.20 / 2224.15 Β±52.25 / 2301.30 ms β 2151.07 / 2235.15 Β±69.35 / 2321.54 ms β no change β
β QQuery 3 β 2166.47 / 2222.81 Β±42.13 / 2276.08 ms β 2198.93 / 2261.06 Β±38.57 / 2318.38 ms β no change β
β QQuery 4 β 2300.69 / 2365.40 Β±64.98 / 2485.61 ms β 2302.18 / 2339.10 Β±22.97 / 2365.18 ms β no change β
β QQuery 5 β 2349.06 / 2407.22 Β±51.87 / 2498.06 ms β 2353.96 / 2434.57 Β±66.06 / 2545.31 ms β no change β
β QQuery 6 β 1.26 / 1.40 Β±0.22 / 1.83 ms β 1.29 / 1.44 Β±0.24 / 1.91 ms β no change β
β QQuery 7 β 1894.01 / 1977.12 Β±75.32 / 2105.83 ms β 1953.37 / 2001.59 Β±36.93 / 2053.17 ms β no change β
β QQuery 8 β 2374.38 / 2452.72 Β±67.16 / 2570.33 ms β 2447.38 / 2495.16 Β±31.39 / 2534.54 ms β no change β
β QQuery 9 β 2617.56 / 2670.83 Β±59.18 / 2774.43 ms β 2630.52 / 2685.03 Β±46.05 / 2742.39 ms β no change β
β QQuery 10 β 2203.57 / 2247.70 Β±30.32 / 2285.12 ms β 2235.82 / 2287.82 Β±47.16 / 2364.17 ms β no change β
β QQuery 11 β 2194.41 / 2295.33 Β±71.04 / 2375.19 ms β 2194.34 / 2269.84 Β±60.04 / 2337.63 ms β no change β
β QQuery 12 β 2362.45 / 2457.92 Β±58.64 / 2533.70 ms β 2382.09 / 2437.39 Β±44.30 / 2503.47 ms β no change β
β QQuery 13 β 2439.86 / 2546.42 Β±65.31 / 2623.02 ms β 2477.32 / 2543.60 Β±35.41 / 2572.64 ms β no change β
β QQuery 14 β 2361.69 / 2440.85 Β±55.73 / 2516.33 ms β 2409.51 / 2448.89 Β±47.32 / 2531.08 ms β no change β
β QQuery 15 β 2416.88 / 2462.80 Β±31.54 / 2509.69 ms β 2415.11 / 2462.31 Β±36.40 / 2521.75 ms β no change β
β QQuery 16 β 2606.97 / 2682.08 Β±38.77 / 2713.83 ms β 2710.07 / 2750.88 Β±44.49 / 2826.66 ms β no change β
β QQuery 17 β 2634.06 / 2692.02 Β±61.05 / 2798.25 ms β 2655.58 / 2685.66 Β±15.32 / 2698.08 ms β no change β
β QQuery 18 β 3157.94 / 3178.41 Β±13.18 / 3197.29 ms β 3193.49 / 3241.69 Β±25.27 / 3262.32 ms β no change β
β QQuery 19 β 1798.49 / 1835.97 Β±27.07 / 1868.54 ms β 1766.70 / 1827.26 Β±40.45 / 1890.40 ms β no change β
β QQuery 20 β 2703.82 / 2726.20 Β±16.43 / 2746.04 ms β 2598.89 / 2643.67 Β±38.78 / 2690.76 ms β no change β
β QQuery 21 β 2669.62 / 2730.29 Β±49.27 / 2791.29 ms β 2593.09 / 2679.54 Β±45.43 / 2726.58 ms β no change β
β QQuery 22 β 3053.31 / 3126.10 Β±45.65 / 3185.08 ms β 3122.97 / 3149.68 Β±32.79 / 3213.38 ms β no change β
β QQuery 23 β 4832.07 / 4982.10 Β±136.13 / 5226.23 ms β 762.05 / 825.30 Β±46.69 / 900.37 ms β +6.04x faster β
β QQuery 24 β 437.05 / 471.38 Β±20.67 / 489.51 ms β 592.89 / 725.57 Β±87.50 / 865.75 ms β 1.54x slower β
β QQuery 25 β 2248.05 / 2297.27 Β±33.20 / 2348.69 ms β 2291.17 / 2382.13 Β±80.66 / 2500.45 ms β no change β
β QQuery 26 β 482.88 / 493.61 Β±11.94 / 514.92 ms β 665.40 / 702.70 Β±33.93 / 762.26 ms β 1.42x slower β
β QQuery 27 β 2656.75 / 2728.97 Β±64.80 / 2846.77 ms β 2643.30 / 2736.79 Β±58.87 / 2810.41 ms β no change β
β QQuery 28 β 4957.17 / 5001.97 Β±35.63 / 5041.45 ms β 4935.80 / 5000.47 Β±61.89 / 5117.63 ms β no change β
β QQuery 29 β 2190.14 / 2269.71 Β±70.88 / 2379.09 ms β 2223.51 / 2262.37 Β±27.91 / 2304.92 ms β no change β
β QQuery 30 β 2424.38 / 2524.13 Β±95.45 / 2689.51 ms β 2439.81 / 2575.79 Β±87.87 / 2710.56 ms β no change β
β QQuery 31 β 2381.38 / 2470.03 Β±80.08 / 2614.18 ms β 2410.69 / 2457.84 Β±35.48 / 2494.54 ms β no change β
β QQuery 32 β 2556.29 / 2589.37 Β±34.75 / 2651.76 ms β 2496.58 / 2585.92 Β±105.66 / 2781.38 ms β no change β
β QQuery 33 β 3413.96 / 3512.47 Β±91.06 / 3678.76 ms β 3402.28 / 3483.27 Β±69.45 / 3603.56 ms β no change β
β QQuery 34 β 3413.20 / 3529.14 Β±61.77 / 3593.18 ms β 3408.91 / 3490.92 Β±67.37 / 3605.85 ms β no change β
β QQuery 35 β 2342.13 / 2482.83 Β±80.87 / 2586.00 ms β 2376.63 / 2441.37 Β±45.78 / 2500.79 ms β no change β
β QQuery 36 β 187.12 / 213.76 Β±31.82 / 273.58 ms β 195.72 / 276.28 Β±46.71 / 340.72 ms β 1.29x slower β
β QQuery 37 β 137.54 / 186.33 Β±39.18 / 241.32 ms β 148.56 / 237.84 Β±57.92 / 318.59 ms β 1.28x slower β
β QQuery 38 β 176.50 / 201.18 Β±25.46 / 246.92 ms β 176.56 / 266.14 Β±55.83 / 335.57 ms β 1.32x slower β
β QQuery 39 β 235.64 / 273.33 Β±45.85 / 361.15 ms β 249.80 / 304.85 Β±42.18 / 356.60 ms β 1.12x slower β
β QQuery 40 β 116.03 / 169.50 Β±37.99 / 225.50 ms β 218.71 / 252.64 Β±32.08 / 301.97 ms β 1.49x slower β
β QQuery 41 β 118.75 / 168.95 Β±35.40 / 217.19 ms β 147.01 / 232.46 Β±53.60 / 300.51 ms β 1.38x slower β
β QQuery 42 β 115.00 / 167.76 Β±50.83 / 249.36 ms β 144.23 / 232.09 Β±54.55 / 300.88 ms β 1.38x slower β
βββββββββββββ΄βββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 88400.69ms β
β Total Time (alamb_pushdown_by_default) β 85297.19ms β
β Average Time (HEAD) β 2055.83ms β
β Average Time (alamb_pushdown_by_default) β 1983.66ms β
β Queries Faster β 1 β
β Queries Slower β 9 β
β Queries with No Change β 33 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 445.1s |
| Peak memory | 8.9 GiB |
| Avg memory | 1.8 GiB |
| CPU user | 982.8s |
| CPU sys | 107.8s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 430.1s |
| Peak memory | 9.5 GiB |
| Avg memory | 1.8 GiB |
| CPU user | 847.7s |
| CPU sys | 97.8s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5400741243
> the filter column must be fully decoded to evaluate the threshold regardless; and for a high-cardinality key like SearchPhrase, RG pruning drops nothing β so the selection/mask construction is paid for zero payback.
Thanks @zhuqi-lucas - I think I had claude try this - it is what is listed as
> and a refinement (106653a) that gates only static conjuncts when a dynamic (TopK/aggregate) filter is present
And it didn't seem to work (though perhaps it wasn't implemented correctly)
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317073802
`β QQuery 23 β 3021.65 ms β 148.70 ms β +20.32x faster β`
π€―
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317304712
run benchmark clickbench_partitioned
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5315251680
run benchmark clickbench_partitioned
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5332451311
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5332422047)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5332422047-1660-swhrj 6.12.85+ #1 SMP Wed Jun 17 20:31:55 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (106653a393b8815e09c09db239afa966f0f627b3) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..106653a393b8815e09c09db239afa966f0f627b3)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5332450194
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5332422781)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5332422781-1661-qp6jh 6.12.85+ #1 SMP Wed Jun 17 20:31:55 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (106653a393b8815e09c09db239afa966f0f627b3) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..106653a393b8815e09c09db239afa966f0f627b3)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5315280518
π€ Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5315251974)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c5315251974-1622-rfbp8 6.12.85+ #1 SMP Wed Jun 17 20:31:55 UTC 2026 aarch64 GNU/Linux`
<details><summary>CPU Details (lscpu)</summary>
```
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>
Comparing alamb/pushdown_by_default (79dd91dc7c0119253e85e64cf0521617cd513680) to 00eba79 (merge-base) [diff](https://github.com/apache/datafusion/compare/00eba790b17f7f0a62e5b81627981197f9bfadb0..79dd91dc7c0119253e85e64cf0521617cd513680)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
Results will be posted here when complete
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5317304146
> Oh wait this is just mostly filter pushdown itself I guess
Yes, exactly. However, the results are very close to not regressing with filter pushdown
I think a few more tweaks and we'll have it
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5439287413
π€ Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/24426#issuecomment-5438849398)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
Comparing alamb/pushdown_by_default (1eb7f67b4c8822c26dfab75fcc303cb2f2f0702d) to a38bb10 (merge-base) [diff](https://github.com/apache/datafusion/compare/a38bb10c33cff4e99a2707059df99f6971c96f33..1eb7f67b4c8822c26dfab75fcc303cb2f2f0702d)
<details><summary>Run configuration</summary>
```yaml
run benchmark clickbench_partitioned
env:
SIMULATE_LATENCY: "true"
```
</details>
<details><summary>CPU Details (lscpu)</summary>
```
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>Details</summary>
<p>
```
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³βββββββββββββ³ββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.18 ms β 1.23 ms β no change β
β QQuery 1 β 1869.54 ms β 1883.01 ms β no change β
β QQuery 2 β 2162.51 ms β 2158.29 ms β no change β
β QQuery 3 β 2163.38 ms β 2166.80 ms β no change β
β QQuery 4 β 2277.90 ms β 2278.54 ms β no change β
β QQuery 5 β 2351.24 ms β 2361.04 ms β no change β
β QQuery 6 β 1.26 ms β 1.26 ms β no change β
β QQuery 7 β 1966.55 ms β 1930.10 ms β no change β
β QQuery 8 β 2369.52 ms β 2405.99 ms β no change β
β QQuery 9 β 2625.43 ms β 2581.00 ms β no change β
β QQuery 10 β 2195.80 ms β 2209.39 ms β no change β
β QQuery 11 β 2237.78 ms β 2219.62 ms β no change β
β QQuery 12 β 2357.85 ms β 2410.90 ms β no change β
β QQuery 13 β 2434.23 ms β 2433.97 ms β no change β
β QQuery 14 β 2357.92 ms β 2394.84 ms β no change β
β QQuery 15 β 2344.58 ms β 2440.35 ms β no change β
β QQuery 16 β 2593.50 ms β 2646.62 ms β no change β
β QQuery 17 β 2618.66 ms β 2658.45 ms β no change β
β QQuery 18 β 3166.59 ms β 3098.94 ms β no change β
β QQuery 19 β 1745.20 ms β 1759.89 ms β no change β
β QQuery 20 β 2626.97 ms β 2606.88 ms β no change β
β QQuery 21 β 2636.25 ms β 2634.29 ms β no change β
β QQuery 22 β 3065.06 ms β 3043.38 ms β no change β
β QQuery 23 β 4824.21 ms β 835.94 ms β +5.77x faster β
β QQuery 24 β 488.74 ms β 725.75 ms β 1.48x slower β
β QQuery 25 β 2233.05 ms β 2292.26 ms β no change β
β QQuery 26 β 488.00 ms β 632.58 ms β 1.30x slower β
β QQuery 27 β 2659.03 ms β 2678.83 ms β no change β
β QQuery 28 β 4987.28 ms β 4995.28 ms β no change β
β QQuery 29 β 2260.85 ms β 2189.80 ms β no change β
β QQuery 30 β 2499.95 ms β 2470.01 ms β no change β
β QQuery 31 β 2383.65 ms β 2477.83 ms β no change β
β QQuery 32 β 2477.18 ms β 2500.50 ms β no change β
β QQuery 33 β 3421.34 ms β 3400.38 ms β no change β
β QQuery 34 β 3408.25 ms β 3413.68 ms β no change β
β QQuery 35 β 2420.03 ms β 2440.91 ms β no change β
β QQuery 36 β 141.28 ms β 141.38 ms β no change β
β QQuery 37 β 134.30 ms β 134.99 ms β no change β
β QQuery 38 β 129.91 ms β 128.36 ms β no change β
β QQuery 39 β 193.55 ms β 172.52 ms β +1.12x faster β
β QQuery 40 β 116.43 ms β 115.06 ms β no change β
β QQuery 41 β 114.45 ms β 114.03 ms β no change β
β QQuery 42 β 117.15 ms β 113.77 ms β no change β
βββββββββββββ΄βββββββββββββ΄ββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 85667.52ms β
β Total Time (alamb_pushdown_by_default) β 82298.66ms β
β Average Time (HEAD) β 1992.27ms β
β Average Time (alamb_pushdown_by_default) β 1913.92ms β
β Queries Faster β 2 β
β Queries Slower β 2 β
β Queries with No Change β 39 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
Distribution per query (min / mean Β±stddev / max):
Comparing HEAD and alamb_pushdown_by_default
--------------------
Benchmark clickbench_partitioned.json
--------------------
βββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββββββββββββββββββββββββββ³ββββββββββββββββ
β Query β HEAD β alamb_pushdown_by_default β Change β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β QQuery 0 β 1.18 / 3.95 Β±5.47 / 14.89 ms β 1.23 / 4.02 Β±5.51 / 15.04 ms β no change β
β QQuery 1 β 1869.54 / 1908.61 Β±29.77 / 1956.44 ms β 1883.01 / 1916.22 Β±24.90 / 1938.90 ms β no change β
β QQuery 2 β 2162.51 / 2224.95 Β±43.51 / 2265.77 ms β 2158.29 / 2217.65 Β±34.82 / 2264.20 ms β no change β
β QQuery 3 β 2163.38 / 2205.32 Β±37.18 / 2266.09 ms β 2166.80 / 2238.39 Β±49.98 / 2297.71 ms β no change β
β QQuery 4 β 2277.90 / 2336.19 Β±47.76 / 2419.32 ms β 2278.54 / 2361.68 Β±57.30 / 2427.97 ms β no change β
β QQuery 5 β 2351.24 / 2431.92 Β±58.91 / 2515.41 ms β 2361.04 / 2421.91 Β±59.41 / 2534.60 ms β no change β
β QQuery 6 β 1.26 / 1.41 Β±0.22 / 1.85 ms β 1.26 / 1.42 Β±0.23 / 1.87 ms β no change β
β QQuery 7 β 1966.55 / 2028.13 Β±56.83 / 2127.67 ms β 1930.10 / 1969.18 Β±45.12 / 2029.99 ms β no change β
β QQuery 8 β 2369.52 / 2421.98 Β±37.26 / 2476.91 ms β 2405.99 / 2444.56 Β±42.37 / 2520.08 ms β no change β
β QQuery 9 β 2625.43 / 2701.34 Β±61.94 / 2799.96 ms β 2581.00 / 2635.89 Β±40.80 / 2707.30 ms β no change β
β QQuery 10 β 2195.80 / 2245.17 Β±32.61 / 2284.62 ms β 2209.39 / 2280.57 Β±56.57 / 2361.97 ms β no change β
β QQuery 11 β 2237.78 / 2315.22 Β±65.76 / 2386.97 ms β 2219.62 / 2257.88 Β±22.05 / 2282.44 ms β no change β
β QQuery 12 β 2357.85 / 2424.76 Β±57.79 / 2497.91 ms β 2410.90 / 2463.31 Β±55.35 / 2555.90 ms β no change β
β QQuery 13 β 2434.23 / 2509.80 Β±56.73 / 2574.50 ms β 2433.97 / 2532.15 Β±75.49 / 2657.25 ms β no change β
β QQuery 14 β 2357.92 / 2451.53 Β±80.39 / 2544.66 ms β 2394.84 / 2431.93 Β±30.89 / 2471.24 ms β no change β
β QQuery 15 β 2344.58 / 2413.94 Β±40.57 / 2468.65 ms β 2440.35 / 2494.97 Β±55.08 / 2594.96 ms β no change β
β QQuery 16 β 2593.50 / 2651.42 Β±55.29 / 2755.42 ms β 2646.62 / 2699.52 Β±47.13 / 2788.11 ms β no change β
β QQuery 17 β 2618.66 / 2710.16 Β±70.93 / 2807.09 ms β 2658.45 / 2719.69 Β±61.07 / 2813.57 ms β no change β
β QQuery 18 β 3166.59 / 3201.86 Β±58.19 / 3317.46 ms β 3098.94 / 3137.99 Β±23.90 / 3167.04 ms β no change β
β QQuery 19 β 1745.20 / 1803.34 Β±61.24 / 1919.29 ms β 1759.89 / 1873.80 Β±62.53 / 1949.80 ms β no change β
β QQuery 20 β 2626.97 / 2677.43 Β±33.34 / 2723.74 ms β 2606.88 / 2668.82 Β±52.99 / 2757.03 ms β no change β
β QQuery 21 β 2636.25 / 2668.41 Β±41.64 / 2749.04 ms β 2634.29 / 2700.69 Β±50.97 / 2788.29 ms β no change β
β QQuery 22 β 3065.06 / 3103.06 Β±31.84 / 3160.03 ms β 3043.38 / 3127.22 Β±51.44 / 3202.62 ms β no change β
β QQuery 23 β 4824.21 / 4976.96 Β±90.85 / 5092.80 ms β 835.94 / 908.77 Β±51.17 / 984.93 ms β +5.48x faster β
β QQuery 24 β 488.74 / 501.30 Β±13.19 / 526.33 ms β 725.75 / 782.39 Β±61.12 / 882.70 ms β 1.56x slower β
β QQuery 25 β 2233.05 / 2296.48 Β±64.02 / 2418.66 ms β 2292.26 / 2368.85 Β±60.50 / 2457.84 ms β no change β
β QQuery 26 β 488.00 / 552.96 Β±53.46 / 640.64 ms β 632.58 / 735.84 Β±100.95 / 885.74 ms β 1.33x slower β
β QQuery 27 β 2659.03 / 2754.80 Β±58.64 / 2823.95 ms β 2678.83 / 2764.79 Β±61.55 / 2870.56 ms β no change β
β QQuery 28 β 4987.28 / 5078.21 Β±89.80 / 5238.11 ms β 4995.28 / 5113.44 Β±89.83 / 5226.04 ms β no change β
β QQuery 29 β 2260.85 / 2285.54 Β±22.41 / 2327.71 ms β 2189.80 / 2275.91 Β±64.11 / 2343.42 ms β no change β
β QQuery 30 β 2499.95 / 2568.52 Β±68.68 / 2684.64 ms β 2470.01 / 2544.62 Β±64.65 / 2621.42 ms β no change β
β QQuery 31 β 2383.65 / 2433.99 Β±44.16 / 2496.85 ms β 2477.83 / 2509.72 Β±32.33 / 2560.34 ms β no change β
β QQuery 32 β 2477.18 / 2569.77 Β±66.10 / 2648.12 ms β 2500.50 / 2586.27 Β±60.33 / 2668.36 ms β no change β
β QQuery 33 β 3421.34 / 3444.36 Β±28.79 / 3497.41 ms β 3400.38 / 3465.95 Β±59.49 / 3566.06 ms β no change β
β QQuery 34 β 3408.25 / 3459.52 Β±51.16 / 3551.45 ms β 3413.68 / 3497.20 Β±61.09 / 3578.90 ms β no change β
β QQuery 35 β 2420.03 / 2466.21 Β±43.88 / 2534.53 ms β 2440.91 / 2497.37 Β±30.65 / 2530.93 ms β no change β
β QQuery 36 β 141.28 / 219.94 Β±42.51 / 256.51 ms β 141.38 / 238.25 Β±73.11 / 367.14 ms β 1.08x slower β
β QQuery 37 β 134.30 / 190.86 Β±36.44 / 236.73 ms β 134.99 / 192.96 Β±39.36 / 240.63 ms β no change β
β QQuery 38 β 129.91 / 191.31 Β±36.90 / 231.88 ms β 128.36 / 190.40 Β±37.23 / 231.01 ms β no change β
β QQuery 39 β 193.55 / 280.50 Β±50.59 / 346.99 ms β 172.52 / 268.12 Β±51.81 / 314.78 ms β no change β
β QQuery 40 β 116.43 / 174.71 Β±37.90 / 221.06 ms β 115.06 / 169.23 Β±37.24 / 221.25 ms β no change β
β QQuery 41 β 114.45 / 173.64 Β±36.73 / 214.91 ms β 114.03 / 194.30 Β±61.20 / 295.48 ms β 1.12x slower β
β QQuery 42 β 117.15 / 188.17 Β±62.24 / 295.43 ms β 113.77 / 167.22 Β±35.69 / 214.84 ms β +1.13x faster β
βββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββββββββββββββββββββββββββ΄ββββββββββββββββ
ββββββββββββββββββββββββββββββββββββββββββββ³βββββββββββββ
β Benchmark Summary β β
β‘ββββββββββββββββββββββββββββββββββββββββββββββββββββββββ©
β Total Time (HEAD) β 88247.63ms β
β Total Time (alamb_pushdown_by_default) β 85071.00ms β
β Average Time (HEAD) β 2052.27ms β
β Average Time (alamb_pushdown_by_default) β 1978.40ms β
β Queries Faster β 2 β
β Queries Slower β 4 β
β Queries with No Change β 37 β
β Queries with Failure β 0 β
ββββββββββββββββββββββββββββββββββββββββββββ΄βββββββββββββ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned β base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 445.1s |
| Peak memory | 9.0 GiB |
| Avg memory | 1.8 GiB |
| CPU user | 974.8s |
| CPU sys | 109.7s |
| Peak spill | 0 B |
**clickbench_partitioned β branch**
| Metric | Value |
|--------|-------|
| Wall time | 430.1s |
| Peak memory | 9.3 GiB |
| Avg memory | 1.8 GiB |
| CPU user | 830.7s |
| CPU sys | 95.0s |
| Peak spill | 0 B |
</details>
---
[File an issue](https://github.com/adriangb/datafusion-benchmarking/issues) against this benchmark runner
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org
Re: [PR] WIP: DO NOT REVIEW: Benchmark enabling parquet filter pushdown by default with one-shot row group I/O [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #24426:
URL: https://github.com/apache/datafusion/pull/24426#issuecomment-5443606081
## Local Q25 results for the five combinations
Five `--profile=profiling` builds of `datafusion-cli` (one per combination in https://github.com/apache/datafusion/pull/24426#issuecomment-5438472525), each run at its own defaults against `hits_partitioned` with the harness-matched setup (`binary_as_string=true`, `hits` view with the `EventDate` cast). 3 interleaved rounds Γ 10 measured runs per binary after 2 warmups; per-round averages agreed within ~2ms, stdev β€3.5ms.
Q25: `SELECT "SearchPhrase" FROM hits WHERE "SearchPhrase" <> '' ORDER BY "SearchPhrase" LIMIT 10;`
| Combination | min ms | avg ms | vs main |
|---|---|---|---|
| 1. main (merge-base `a38bb10`) | 81.0 | 84.2 | 1.00x |
| 2. + `progressive_io`, `pushdown_filters` on | 105.0 | 108.4 | 1.29x |
| 3. + `reorder_filters` on | 102.0 | 106.6 | 1.27x |
| 4. + gate from #23420 | 110.0 | 113.1 | **1.34x** |
| 5. + [arrow-rs#10852](https://github.com/apache/arrow-rs/pull/10852) (via [arrow-rs#10885](https://github.com/apache/arrow-rs/pull/10885) pin) | 97.0 | 101.9 | **1.21x** |
Step-by-step isolation:
1. **Pushdown itself costs ~24ms (1.29x) on Q25** β consistent with the earlier profiling (selection construction + cache-replay concat + assembly, see [arrow-rs#10774](https://github.com/apache/arrow-rs/issues/10774) / [arrow-rs#10776](https://github.com/apache/arrow-rs/issues/10776)).
2. **`reorder_filters` is neutral here** (-2ms, borderline noise): Q25 has a single static conjunct.
3. **The gate costs ~6.5ms on Q25** (106.6 β 113.1, reproduced in all three rounds). Q25 takes the gate's *dynamic-filter bypass* path, so the heuristic itself decides nothing here β the earlier samply profiles showed `FilterExec` frames on the gate binary that are absent without the gate, suggesting the gate's per-filter parent responses change where the TopK dynamic filter lands (`FilterExec` above the scan vs absorbed). Needs investigation in `try_pushdown_filters` β this looks like a fixable interaction, not an inherent cost of the heuristic.
4. **arrow-rs#10852 recovers 11.2ms** (113.1 β 101.9), about 40% of the pushdown penalty β matching the prediction from the profile (~100ms CPU/query of booleanβselectorsβmask churn spread over ~10 effective cores). Nice work @hhhizzz!
Remaining ~18ms gap vs main: ~6ms from the gate/TopK placement issue above; the bulk of the rest matches the `CachedArrayReader` replay concat ([arrow-rs#10774](https://github.com/apache/arrow-rs/issues/10774), ~170ms CPU/query).
Next: the same five-way sweep across all 43 ClickBench queries.
--
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@datafusion.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: github-unsubscribe@datafusion.apache.org
For additional commands, e-mail: github-help@datafusion.apache.org