You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@datafusion.apache.org by "Dandandan (via GitHub)" <gi...@apache.org> on 2026/04/28 20:17:20 UTC
[PR] perf: strength reduce hash partition modulo [datafusion]
Dandandan opened a new pull request, #21900:
URL: https://github.com/apache/datafusion/pull/21900
## Which issue does this PR close?
- Closes #.
## Rationale for this change
Hash repartition assigns each row with . Because is a runtime value, this can compile to hardware division in the per-row partitioning loop.
This is an alternative to #21830 that preserves the existing modulo partition mapping by strength-reducing the u64 remainder instead of changing the mapping to fastrange.
## What changes are included in this PR?
- Adds a small local u64 strength reducer for exact remainder computation.
- Uses a mask for power-of-two partition counts, including 1.
- Uses a precomputed reciprocal multiply path for other partition counts.
- Validates hash partitioner construction rejects zero partitions.
- Uses the precomputed reducer in 's hash index distribution loop.
## Are these changes tested?
-
running 1 test
test repartition::tests::strength_reduced_u64_remainder_matches_modulo ... ok
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 1391 filtered out; finished in 0.01s
-
running 1 test
test repartition::tests::hash_partitioner_requires_nonzero_partitions ... ok
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 1391 filtered out; finished in 0.00s
-
running 1 test
test repartition::tests::many_to_many_hash_partition ... ok
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 1391 filtered out; finished in 0.01s
-
running 21 tests
test repartition::tests::hash_partitioner_requires_nonzero_partitions ... ok
test repartition::tests::error_for_input_exec ... ok
test repartition::tests::oom ... ok
test repartition::tests::repartition_with_error_in_stream ... ok
test repartition::tests::hash_repartition_avoid_empty_batch ... ok
test repartition::tests::one_to_many_round_robin ... ok
test repartition::tests::many_to_many_round_robin ... ok
test repartition::tests::many_to_many_round_robin_within_tokio_task ... ok
test repartition::tests::many_to_one_round_robin ... ok
test repartition::tests::repartition_without_spilling ... ok
test repartition::tests::hash_repartition_with_dropping_output_stream ... ok
test repartition::tests::unsupported_partitioning ... ok
test repartition::tests::test_repartition_with_coalescing ... ok
test repartition::tests::many_to_many_hash_partition ... ok
test repartition::tests::strength_reduced_u64_remainder_matches_modulo ... ok
test repartition::tests::test_repartition_ordering_with_spilling ... ok
test repartition::tests::repartition_with_partial_spilling ... ok
test repartition::tests::repartition_with_spilling ... ok
test repartition::tests::test_drop_cancel ... ok
test repartition::tests::robin_repartition_with_dropping_output_stream ... ok
test repartition::tests::repartition_with_delayed_stream ... ok
test result: ok. 21 passed; 0 failed; 0 ignored; 0 measured; 1371 filtered out; finished in 0.14s
-
-
- [rust_lint.sh] Running rust_fmt.sh
[rust_fmt.sh] `cargo fmt --all -- --check`
[rust_lint.sh] Running rust_clippy.sh
[rust_clippy.sh] `cargo clippy --all-targets --workspace --features avro,integration-tests,extended_tests -- -D warnings`
[rust_lint.sh] Running rust_toml_fmt.sh
[rust_toml_fmt.sh] `taplo format --check`
[rust_lint.sh] Running license_header.sh
[license_header.sh] `hawkeye check --config licenserc.toml`
[rust_lint.sh] Running typos_check.sh
[typos_check.sh] `typos --config typos.toml`
[rust_lint.sh] Running doc_prettier_check.sh
[doc_prettier_check.sh] prettier@2.7.1 check
Checking formatting...
All matched files use Prettier code style!
[rust_lint.sh] Running rust_docs.sh
## Are there any user-facing changes?
No.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: github-unsubscribe@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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338898059
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338764222)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
<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 perf-strength-reduce-hash-partition
--------------------
Benchmark clickbench_partitioned.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ perf-strength-reduce-hash-partition โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 0 โ 1.15 / 4.55 ยฑ6.70 / 17.96 ms โ 1.24 / 4.86 ยฑ7.04 / 18.94 ms โ 1.07x slower โ
โ QQuery 1 โ 12.68 / 13.11 ยฑ0.22 / 13.27 ms โ 13.05 / 13.62 ยฑ0.34 / 14.08 ms โ no change โ
โ QQuery 2 โ 37.29 / 37.71 ยฑ0.39 / 38.31 ms โ 38.30 / 39.06 ยฑ0.43 / 39.58 ms โ no change โ
โ QQuery 3 โ 31.08 / 32.03 ยฑ1.15 / 34.30 ms โ 33.36 / 33.55 ยฑ0.16 / 33.74 ms โ no change โ
โ QQuery 4 โ 257.53 / 261.21 ยฑ3.00 / 264.64 ms โ 260.08 / 267.18 ยฑ5.40 / 275.48 ms โ no change โ
โ QQuery 5 โ 282.36 / 286.00 ยฑ3.16 / 291.76 ms โ 279.94 / 294.39 ยฑ12.13 / 309.15 ms โ no change โ
โ QQuery 6 โ 6.55 / 7.05 ยฑ0.33 / 7.52 ms โ 6.09 / 6.82 ยฑ0.59 / 7.85 ms โ no change โ
โ QQuery 7 โ 14.01 / 14.12 ยฑ0.09 / 14.23 ms โ 14.21 / 14.26 ยฑ0.07 / 14.39 ms โ no change โ
โ QQuery 8 โ 328.87 / 330.47 ยฑ1.29 / 331.98 ms โ 320.20 / 333.42 ยฑ8.38 / 343.99 ms โ no change โ
โ QQuery 9 โ 444.36 / 448.71 ยฑ3.95 / 455.52 ms โ 448.31 / 458.25 ยฑ8.85 / 474.62 ms โ no change โ
โ QQuery 10 โ 73.91 / 75.83 ยฑ1.91 / 79.20 ms โ 73.31 / 76.45 ยฑ4.22 / 84.72 ms โ no change โ
โ QQuery 11 โ 85.90 / 86.36 ยฑ0.52 / 87.34 ms โ 85.49 / 87.38 ยฑ2.55 / 92.41 ms โ no change โ
โ QQuery 12 โ 273.77 / 284.16 ยฑ6.75 / 293.52 ms โ 272.95 / 277.34 ยฑ3.91 / 282.70 ms โ no change โ
โ QQuery 13 โ 414.87 / 425.52 ยฑ7.96 / 435.77 ms โ 421.90 / 435.66 ยฑ15.66 / 455.41 ms โ no change โ
โ QQuery 14 โ 307.62 / 313.79 ยฑ6.51 / 326.01 ms โ 314.56 / 318.87 ยฑ3.29 / 323.41 ms โ no change โ
โ QQuery 15 โ 324.29 / 329.99 ยฑ4.41 / 335.68 ms โ 284.29 / 296.49 ยฑ11.23 / 317.57 ms โ +1.11x faster โ
โ QQuery 16 โ 628.47 / 645.70 ยฑ21.55 / 687.42 ms โ 609.43 / 623.27 ยฑ10.11 / 641.00 ms โ no change โ
โ QQuery 17 โ 619.90 / 629.71 ยฑ6.49 / 637.56 ms โ 611.73 / 619.24 ยฑ6.76 / 631.23 ms โ no change โ
โ QQuery 18 โ 1251.55 / 1354.64 ยฑ73.10 / 1455.17 ms โ 1218.37 / 1291.83 ยฑ53.71 / 1364.85 ms โ no change โ
โ QQuery 19 โ 30.51 / 31.12 ยฑ0.53 / 31.93 ms โ 30.40 / 44.58 ยฑ25.69 / 95.82 ms โ 1.43x slower โ
โ QQuery 20 โ 523.29 / 535.66 ยฑ10.14 / 551.36 ms โ 529.89 / 539.76 ยฑ6.72 / 549.70 ms โ no change โ
โ QQuery 21 โ 596.38 / 599.46 ยฑ2.91 / 604.75 ms โ 599.98 / 606.79 ยฑ7.07 / 617.30 ms โ no change โ
โ QQuery 22 โ 1058.93 / 1068.59 ยฑ8.14 / 1082.65 ms โ 1071.60 / 1075.88 ยฑ4.52 / 1084.33 ms โ no change โ
โ QQuery 23 โ 3338.81 / 3481.74 ยฑ114.35 / 3651.47 ms โ 3346.62 / 3432.22 ยฑ79.64 / 3545.01 ms โ no change โ
โ QQuery 24 โ 43.98 / 53.08 ยฑ10.12 / 68.40 ms โ 42.04 / 42.74 ยฑ0.58 / 43.65 ms โ +1.24x faster โ
โ QQuery 25 โ 118.02 / 119.56 ยฑ1.05 / 120.94 ms โ 114.08 / 116.42 ยฑ3.03 / 122.27 ms โ no change โ
โ QQuery 26 โ 44.45 / 46.87 ยฑ3.40 / 53.46 ms โ 42.47 / 44.96 ยฑ2.76 / 49.04 ms โ no change โ
โ QQuery 27 โ 673.33 / 686.99 ยฑ8.94 / 697.29 ms โ 671.25 / 680.10 ยฑ7.26 / 689.38 ms โ no change โ
โ QQuery 28 โ 2990.11 / 3039.60 ยฑ46.03 / 3124.08 ms โ 3022.27 / 3104.96 ยฑ47.12 / 3168.68 ms โ no change โ
โ QQuery 29 โ 42.07 / 42.33 ยฑ0.23 / 42.72 ms โ 42.93 / 55.16 ยฑ10.55 / 69.21 ms โ 1.30x slower โ
โ QQuery 30 โ 307.59 / 320.05 ยฑ15.59 / 348.49 ms โ 309.93 / 318.05 ยฑ8.81 / 335.10 ms โ no change โ
โ QQuery 31 โ 304.27 / 308.70 ยฑ5.74 / 319.81 ms โ 302.80 / 311.61 ยฑ6.66 / 322.63 ms โ no change โ
โ QQuery 32 โ 1009.20 / 1052.24 ยฑ27.24 / 1084.42 ms โ 977.95 / 992.86 ยฑ11.66 / 1011.06 ms โ +1.06x faster โ
โ QQuery 33 โ 1457.90 / 1595.22 ยฑ100.03 / 1708.48 ms โ 1464.63 / 1543.88 ยฑ66.18 / 1648.21 ms โ no change โ
โ QQuery 34 โ 1441.22 / 1480.26 ยฑ35.55 / 1533.08 ms โ 1426.47 / 1447.63 ยฑ14.50 / 1469.71 ms โ no change โ
โ QQuery 35 โ 285.58 / 343.31 ยฑ64.65 / 465.44 ms โ 338.22 / 358.36 ยฑ18.49 / 382.68 ms โ no change โ
โ QQuery 36 โ 65.96 / 73.07 ยฑ8.95 / 90.17 ms โ 67.21 / 76.04 ยฑ6.02 / 83.88 ms โ no change โ
โ QQuery 37 โ 37.78 / 38.44 ยฑ0.53 / 39.41 ms โ 37.52 / 39.28 ยฑ2.51 / 44.24 ms โ no change โ
โ QQuery 38 โ 42.34 / 46.78 ยฑ2.76 / 50.62 ms โ 42.68 / 44.76 ยฑ1.83 / 47.25 ms โ no change โ
โ QQuery 39 โ 134.21 / 150.69 ยฑ12.26 / 171.70 ms โ 143.40 / 151.87 ยฑ9.66 / 170.73 ms โ no change โ
โ QQuery 40 โ 16.04 / 17.68 ยฑ2.81 / 23.29 ms โ 15.70 / 17.78 ยฑ3.45 / 24.67 ms โ no change โ
โ QQuery 41 โ 15.02 / 17.01 ยฑ2.51 / 21.48 ms โ 15.08 / 15.19 ยฑ0.09 / 15.32 ms โ +1.12x faster โ
โ QQuery 42 โ 14.54 / 14.72 ยฑ0.15 / 14.97 ms โ 14.53 / 14.63 ยฑ0.09 / 14.79 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 20743.84ms โ
โ Total Time (perf-strength-reduce-hash-partition) โ 20567.44ms โ
โ Average Time (HEAD) โ 482.41ms โ
โ Average Time (perf-strength-reduce-hash-partition) โ 478.31ms โ
โ Queries Faster โ 4 โ
โ Queries Slower โ 3 โ
โ Queries with No Change โ 36 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 105.0s |
| Peak memory | 30.1 GiB |
| Avg memory | 23.1 GiB |
| CPU user | 1091.0s |
| CPU sys | 69.0s |
| Peak spill | 0 B |
**clickbench_partitioned โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 105.0s |
| Peak memory | 30.7 GiB |
| Avg memory | 23.2 GiB |
| CPU user | 1085.7s |
| CPU sys | 68.0s |
| 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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338903467
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338749432)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
<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 perf-strength-reduce-hash-partition
--------------------
Benchmark clickbench_partitioned.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ perf-strength-reduce-hash-partition โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 0 โ 1.19 / 4.68 ยฑ6.84 / 18.36 ms โ 1.20 / 4.79 ยฑ7.05 / 18.90 ms โ no change โ
โ QQuery 1 โ 12.37 / 12.82 ยฑ0.24 / 13.09 ms โ 12.95 / 13.62 ยฑ0.65 / 14.84 ms โ 1.06x slower โ
โ QQuery 2 โ 36.62 / 37.05 ยฑ0.60 / 38.22 ms โ 37.26 / 37.53 ยฑ0.27 / 37.97 ms โ no change โ
โ QQuery 3 โ 31.16 / 31.81 ยฑ0.89 / 33.58 ms โ 31.52 / 32.06 ยฑ0.56 / 33.12 ms โ no change โ
โ QQuery 4 โ 245.62 / 251.52 ยฑ4.67 / 256.81 ms โ 239.86 / 244.05 ยฑ4.71 / 250.73 ms โ no change โ
โ QQuery 5 โ 289.57 / 290.88 ยฑ0.97 / 291.87 ms โ 287.31 / 290.43 ยฑ2.78 / 295.52 ms โ no change โ
โ QQuery 6 โ 6.92 / 7.23 ยฑ0.25 / 7.62 ms โ 6.85 / 7.31 ยฑ0.42 / 7.95 ms โ no change โ
โ QQuery 7 โ 13.72 / 13.97 ยฑ0.13 / 14.09 ms โ 14.27 / 15.87 ยฑ2.92 / 21.70 ms โ 1.14x slower โ
โ QQuery 8 โ 334.76 / 338.99 ยฑ2.36 / 342.02 ms โ 330.16 / 335.07 ยฑ4.54 / 343.24 ms โ no change โ
โ QQuery 9 โ 457.86 / 463.65 ยฑ5.71 / 474.45 ms โ 463.05 / 480.58 ยฑ15.00 / 499.99 ms โ no change โ
โ QQuery 10 โ 75.01 / 75.68 ยฑ0.45 / 76.28 ms โ 75.25 / 78.01 ยฑ4.07 / 86.07 ms โ no change โ
โ QQuery 11 โ 86.06 / 86.73 ยฑ0.75 / 87.65 ms โ 85.86 / 86.51 ยฑ0.34 / 86.82 ms โ no change โ
โ QQuery 12 โ 280.43 / 285.45 ยฑ3.39 / 290.05 ms โ 280.05 / 285.24 ยฑ5.07 / 291.33 ms โ no change โ
โ QQuery 13 โ 405.73 / 413.86 ยฑ8.88 / 425.97 ms โ 392.11 / 405.99 ยฑ13.21 / 426.37 ms โ no change โ
โ QQuery 14 โ 293.64 / 299.80 ยฑ5.45 / 307.53 ms โ 296.10 / 299.90 ยฑ2.87 / 304.02 ms โ no change โ
โ QQuery 15 โ 290.90 / 294.42 ยฑ2.85 / 299.10 ms โ 289.84 / 298.74 ยฑ5.73 / 307.32 ms โ no change โ
โ QQuery 16 โ 633.07 / 646.24 ยฑ10.46 / 662.35 ms โ 645.30 / 650.41 ยฑ2.63 / 652.55 ms โ no change โ
โ QQuery 17 โ 637.48 / 644.94 ยฑ5.66 / 653.68 ms โ 648.59 / 656.35 ยฑ6.37 / 667.87 ms โ no change โ
โ QQuery 18 โ 1286.26 / 1309.11 ยฑ12.49 / 1324.05 ms โ 1287.55 / 1314.03 ยฑ31.57 / 1373.95 ms โ no change โ
โ QQuery 19 โ 29.25 / 33.32 ยฑ4.13 / 39.32 ms โ 29.54 / 29.81 ยฑ0.17 / 30.00 ms โ +1.12x faster โ
โ QQuery 20 โ 518.40 / 524.61 ยฑ7.06 / 538.44 ms โ 523.93 / 528.38 ยฑ5.12 / 538.07 ms โ no change โ
โ QQuery 21 โ 598.68 / 603.08 ยฑ4.74 / 611.20 ms โ 608.01 / 613.71 ยฑ5.87 / 622.99 ms โ no change โ
โ QQuery 22 โ 1074.47 / 1083.36 ยฑ9.06 / 1098.06 ms โ 1080.02 / 1083.50 ยฑ2.89 / 1087.50 ms โ no change โ
โ QQuery 23 โ 3378.25 / 3402.99 ยฑ20.27 / 3434.00 ms โ 3394.97 / 3418.28 ยฑ14.71 / 3432.30 ms โ no change โ
โ QQuery 24 โ 41.78 / 42.24 ยฑ0.36 / 42.74 ms โ 41.71 / 43.55 ยฑ2.18 / 47.73 ms โ no change โ
โ QQuery 25 โ 114.38 / 115.79 ยฑ0.92 / 116.65 ms โ 115.11 / 116.48 ยฑ0.73 / 117.29 ms โ no change โ
โ QQuery 26 โ 42.44 / 45.24 ยฑ3.37 / 50.84 ms โ 42.25 / 42.77 ยฑ0.33 / 43.09 ms โ +1.06x faster โ
โ QQuery 27 โ 672.34 / 679.72 ยฑ5.92 / 688.86 ms โ 672.82 / 679.98 ยฑ6.55 / 690.31 ms โ no change โ
โ QQuery 28 โ 3014.01 / 3043.51 ยฑ17.19 / 3060.02 ms โ 3020.56 / 3050.28 ยฑ24.45 / 3082.57 ms โ no change โ
โ QQuery 29 โ 42.52 / 42.77 ยฑ0.28 / 43.23 ms โ 42.62 / 43.25 ยฑ0.68 / 44.47 ms โ no change โ
โ QQuery 30 โ 317.79 / 331.19 ยฑ15.95 / 360.70 ms โ 306.44 / 313.83 ยฑ6.02 / 323.89 ms โ +1.06x faster โ
โ QQuery 31 โ 315.38 / 318.19 ยฑ3.31 / 324.65 ms โ 299.71 / 308.06 ยฑ5.19 / 315.68 ms โ no change โ
โ QQuery 32 โ 1028.31 / 1035.57 ยฑ4.79 / 1041.48 ms โ 980.26 / 985.80 ยฑ7.05 / 998.60 ms โ no change โ
โ QQuery 33 โ 1469.73 / 1491.13 ยฑ14.60 / 1511.06 ms โ 1460.82 / 1477.52 ยฑ9.33 / 1487.21 ms โ no change โ
โ QQuery 34 โ 1486.32 / 1506.16 ยฑ13.86 / 1523.95 ms โ 1478.33 / 1509.96 ยฑ18.53 / 1533.95 ms โ no change โ
โ QQuery 35 โ 301.24 / 326.05 ยฑ34.29 / 393.94 ms โ 289.65 / 301.92 ยฑ19.30 / 340.30 ms โ +1.08x faster โ
โ QQuery 36 โ 63.81 / 66.80 ยฑ2.61 / 70.32 ms โ 63.18 / 73.15 ยฑ13.92 / 100.78 ms โ 1.10x slower โ
โ QQuery 37 โ 36.61 / 39.99 ยฑ3.82 / 46.71 ms โ 37.27 / 38.49 ยฑ1.42 / 40.83 ms โ no change โ
โ QQuery 38 โ 40.67 / 45.36 ยฑ3.42 / 50.44 ms โ 40.47 / 41.83 ยฑ1.06 / 43.46 ms โ +1.08x faster โ
โ QQuery 39 โ 128.76 / 136.11 ยฑ4.87 / 143.87 ms โ 133.39 / 137.27 ยฑ3.20 / 142.81 ms โ no change โ
โ QQuery 40 โ 14.77 / 16.48 ยฑ2.43 / 21.25 ms โ 14.93 / 17.03 ยฑ3.72 / 24.46 ms โ no change โ
โ QQuery 41 โ 14.35 / 14.59 ยฑ0.39 / 15.37 ms โ 14.31 / 14.53 ยฑ0.15 / 14.76 ms โ no change โ
โ QQuery 42 โ 13.89 / 14.95 ยฑ1.70 / 18.33 ms โ 14.07 / 14.20 ยฑ0.09 / 14.33 ms โ +1.05x faster โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 20468.03ms โ
โ Total Time (perf-strength-reduce-hash-partition) โ 20420.07ms โ
โ Average Time (HEAD) โ 476.00ms โ
โ Average Time (perf-strength-reduce-hash-partition) โ 474.89ms โ
โ Queries Faster โ 6 โ
โ Queries Slower โ 3 โ
โ Queries with No Change โ 34 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 105.0s |
| Peak memory | 29.7 GiB |
| Avg memory | 23.2 GiB |
| CPU user | 1082.8s |
| CPU sys | 64.1s |
| Peak spill | 0 B |
**clickbench_partitioned โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 105.0s |
| Peak memory | 29.8 GiB |
| Avg memory | 22.7 GiB |
| CPU user | 1079.1s |
| CPU sys | 65.4s |
| 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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338770045
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338749432)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4338749432-1876-qzs7f 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 perf-strength-reduce-hash-partition (b0091c6806aeefa4475e65427e1458a98d00f800) to 65f337d (merge-base) [diff](https://github.com/apache/datafusion/compare/65f337d6637aa069012e0e9286f859fb3a743253..b0091c6806aeefa4475e65427e1458a98d00f800) using: tpcds
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338785176
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338764222)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4338764222-1880-7vjkj 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 perf-strength-reduce-hash-partition (b0091c6806aeefa4475e65427e1458a98d00f800) to 65f337d (merge-base) [diff](https://github.com/apache/datafusion/compare/65f337d6637aa069012e0e9286f859fb3a743253..b0091c6806aeefa4475e65427e1458a98d00f800) using: tpcds
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] perf: strength reduce hash partition modulo (up to 1.16x faster) [datafusion]
Posted by "Copilot (via GitHub)" <gi...@apache.org>.
Copilot commented on code in PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#discussion_r3161606233
##########
datafusion/physical-plan/src/repartition/mod.rs:
##########
@@ -446,22 +511,27 @@ impl BatchPartitioner {
/// - `num_partitions`: Total number of output partitions.
/// - `timer`: Metric used to record time spent during repartitioning.
///
- /// # Notes
- /// This constructor cannot fail and performs no validation.
+ /// # Errors
+ /// Returns an error if `num_partitions` is zero.
pub fn new_hash_partitioner(
exprs: Vec<Arc<dyn PhysicalExpr>>,
num_partitions: usize,
timer: metrics::Time,
- ) -> Self {
- Self {
+ ) -> Result<Self> {
+ if num_partitions == 0 {
+ return internal_err!("Hash repartition requires at least one partition");
+ }
+
+ Ok(Self {
state: BatchPartitionerState::Hash {
exprs,
num_partitions,
+ partition_reducer: StrengthReducedU64::new(num_partitions as u64),
hash_buffer: vec![],
indices: vec![vec![]; num_partitions],
},
timer,
- }
+ })
}
Review Comment:
`new_hash_partitioner` is `pub` in the `datafusion-physical-plan` crate, but this change makes it fallible (`-> Result<Self>`) and introduces a new runtime error for `num_partitions == 0`. That is a breaking public API change for downstream crates. Consider preserving the existing signature (e.g., keep `new_hash_partitioner` infallible and add a new `try_new_hash_partitioner` / `new_hash_partitioner_checked`), or reduce visibility if this is intended to be internal-only.
--
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] perf: strength reduce hash partition modulo (up to 1.16x faster) [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan merged PR #21900:
URL: https://github.com/apache/datafusion/pull/21900
--
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338784488
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338764222)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4338764222-1879-9x8g9 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 perf-strength-reduce-hash-partition (b0091c6806aeefa4475e65427e1458a98d00f800) to 65f337d (merge-base) [diff](https://github.com/apache/datafusion/compare/65f337d6637aa069012e0e9286f859fb3a743253..b0091c6806aeefa4475e65427e1458a98d00f800) using: clickbench_partitioned
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338764766
run benchmark tpch10
--
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338769339
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338749432)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4338749432-1875-wrwgm 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 perf-strength-reduce-hash-partition (b0091c6806aeefa4475e65427e1458a98d00f800) to 65f337d (merge-base) [diff](https://github.com/apache/datafusion/compare/65f337d6637aa069012e0e9286f859fb3a743253..b0091c6806aeefa4475e65427e1458a98d00f800) using: clickbench_partitioned
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338898501
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338749432)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
<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 perf-strength-reduce-hash-partition
--------------------
Benchmark tpcds_sf1.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ perf-strength-reduce-hash-partition โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 6.82 / 7.41 ยฑ0.84 / 9.07 ms โ 6.83 / 7.37 ยฑ0.82 / 9.00 ms โ no change โ
โ QQuery 2 โ 145.73 / 146.44 ยฑ0.56 / 147.35 ms โ 146.47 / 146.76 ยฑ0.18 / 146.97 ms โ no change โ
โ QQuery 3 โ 113.42 / 114.31 ยฑ0.63 / 115.00 ms โ 113.92 / 114.85 ยฑ0.87 / 115.99 ms โ no change โ
โ QQuery 4 โ 1340.02 / 1351.65 ยฑ11.62 / 1370.27 ms โ 1332.84 / 1341.27 ยฑ9.04 / 1358.82 ms โ no change โ
โ QQuery 5 โ 173.26 / 173.66 ยฑ0.21 / 173.88 ms โ 174.21 / 174.79 ยฑ0.61 / 175.71 ms โ no change โ
โ QQuery 6 โ 841.37 / 851.48 ยฑ9.94 / 868.90 ms โ 825.18 / 865.84 ยฑ22.20 / 886.66 ms โ no change โ
โ QQuery 7 โ 337.50 / 338.11 ยฑ0.50 / 338.80 ms โ 320.73 / 324.30 ยฑ2.91 / 328.38 ms โ no change โ
โ QQuery 8 โ 114.14 / 114.74 ยฑ0.42 / 115.29 ms โ 113.71 / 114.60 ยฑ0.82 / 116.09 ms โ no change โ
โ QQuery 9 โ 105.08 / 111.31 ยฑ10.78 / 132.85 ms โ 101.80 / 105.96 ยฑ2.20 / 108.06 ms โ no change โ
โ QQuery 10 โ 104.08 / 104.58 ยฑ0.41 / 105.27 ms โ 103.52 / 106.05 ยฑ3.09 / 111.86 ms โ no change โ
โ QQuery 11 โ 948.13 / 951.29 ยฑ4.98 / 961.18 ms โ 943.96 / 956.74 ยฑ8.36 / 969.55 ms โ no change โ
โ QQuery 12 โ 43.57 / 44.24 ยฑ0.41 / 44.85 ms โ 44.03 / 44.56 ยฑ0.44 / 45.05 ms โ no change โ
โ QQuery 13 โ 392.59 / 394.72 ยฑ1.09 / 395.59 ms โ 395.04 / 398.84 ยฑ2.71 / 402.94 ms โ no change โ
โ QQuery 14 โ 981.18 / 983.80 ยฑ4.41 / 992.59 ms โ 956.80 / 960.00 ยฑ2.61 / 963.55 ms โ no change โ
โ QQuery 15 โ 15.10 / 15.43 ยฑ0.22 / 15.68 ms โ 14.74 / 14.92 ยฑ0.15 / 15.18 ms โ no change โ
โ QQuery 16 โ 7.25 / 7.47 ยฑ0.20 / 7.83 ms โ 7.33 / 7.56 ยฑ0.25 / 8.04 ms โ no change โ
โ QQuery 17 โ 223.83 / 224.77 ยฑ0.78 / 225.98 ms โ 204.50 / 206.55 ยฑ2.27 / 210.93 ms โ +1.09x faster โ
โ QQuery 18 โ 126.94 / 128.56 ยฑ0.92 / 129.57 ms โ 124.61 / 125.51 ยฑ0.95 / 127.24 ms โ no change โ
โ QQuery 19 โ 154.44 / 155.39 ยฑ0.98 / 157.29 ms โ 153.48 / 157.17 ยฑ2.78 / 160.60 ms โ no change โ
โ QQuery 20 โ 13.41 / 13.53 ยฑ0.12 / 13.77 ms โ 13.10 / 13.33 ยฑ0.16 / 13.50 ms โ no change โ
โ QQuery 21 โ 19.21 / 19.48 ยฑ0.22 / 19.72 ms โ 19.36 / 19.55 ยฑ0.18 / 19.88 ms โ no change โ
โ QQuery 22 โ 488.71 / 491.44 ยฑ3.13 / 497.30 ms โ 480.97 / 486.03 ยฑ5.51 / 496.26 ms โ no change โ
โ QQuery 23 โ 834.52 / 839.02 ยฑ3.12 / 843.96 ms โ 794.99 / 797.74 ยฑ1.41 / 798.96 ms โ no change โ
โ QQuery 24 โ 374.02 / 380.66 ยฑ5.43 / 389.20 ms โ 338.21 / 345.69 ยฑ4.40 / 350.78 ms โ +1.10x faster โ
โ QQuery 25 โ 331.50 / 334.79 ยฑ2.54 / 339.01 ms โ 310.41 / 312.22 ยฑ1.26 / 313.99 ms โ +1.07x faster โ
โ QQuery 26 โ 78.41 / 79.19 ยฑ0.66 / 80.29 ms โ 76.23 / 76.61 ยฑ0.31 / 77.08 ms โ no change โ
โ QQuery 27 โ 6.95 / 7.12 ยฑ0.16 / 7.42 ms โ 6.99 / 7.13 ยฑ0.17 / 7.46 ms โ no change โ
โ QQuery 28 โ 148.95 / 149.48 ยฑ0.61 / 150.64 ms โ 149.73 / 149.99 ยฑ0.24 / 150.29 ms โ no change โ
โ QQuery 29 โ 271.79 / 275.14 ยฑ2.03 / 277.57 ms โ 247.75 / 250.82 ยฑ2.37 / 254.57 ms โ +1.10x faster โ
โ QQuery 30 โ 42.13 / 42.50 ยฑ0.46 / 43.39 ms โ 42.03 / 42.46 ยฑ0.46 / 43.29 ms โ no change โ
โ QQuery 31 โ 163.67 / 166.19 ยฑ2.10 / 169.91 ms โ 165.05 / 167.41 ยฑ2.55 / 172.12 ms โ no change โ
โ QQuery 32 โ 13.37 / 14.40 ยฑ1.42 / 17.19 ms โ 13.54 / 13.75 ยฑ0.20 / 14.06 ms โ no change โ
โ QQuery 33 โ 138.01 / 139.76 ยฑ1.36 / 141.92 ms โ 139.64 / 140.66 ยฑ0.52 / 141.08 ms โ no change โ
โ QQuery 34 โ 7.02 / 7.19 ยฑ0.16 / 7.49 ms โ 7.31 / 7.40 ยฑ0.13 / 7.65 ms โ no change โ
โ QQuery 35 โ 101.43 / 102.49 ยฑ0.83 / 103.87 ms โ 101.83 / 103.21 ยฑ1.26 / 105.53 ms โ no change โ
โ QQuery 36 โ 6.55 / 6.75 ยฑ0.13 / 6.88 ms โ 6.79 / 6.94 ยฑ0.09 / 7.07 ms โ no change โ
โ QQuery 37 โ 8.24 / 8.30 ยฑ0.03 / 8.34 ms โ 8.36 / 8.45 ยฑ0.08 / 8.61 ms โ no change โ
โ QQuery 38 โ 87.70 / 89.51 ยฑ1.20 / 91.00 ms โ 85.38 / 87.62 ยฑ2.76 / 92.95 ms โ no change โ
โ QQuery 39 โ 117.89 / 123.51 ยฑ6.05 / 135.15 ms โ 118.30 / 120.08 ยฑ2.46 / 124.89 ms โ no change โ
โ QQuery 40 โ 101.76 / 106.48 ยฑ4.47 / 112.78 ms โ 94.74 / 98.06 ยฑ2.65 / 102.04 ms โ +1.09x faster โ
โ QQuery 41 โ 14.52 / 15.03 ยฑ0.86 / 16.74 ms โ 14.42 / 14.55 ยฑ0.15 / 14.84 ms โ no change โ
โ QQuery 42 โ 107.22 / 107.54 ยฑ0.24 / 107.83 ms โ 107.42 / 109.38 ยฑ2.46 / 114.19 ms โ no change โ
โ QQuery 43 โ 5.63 / 5.84 ยฑ0.25 / 6.33 ms โ 5.82 / 5.94 ยฑ0.13 / 6.19 ms โ no change โ
โ QQuery 44 โ 11.68 / 11.77 ยฑ0.08 / 11.90 ms โ 11.88 / 12.12 ยฑ0.16 / 12.37 ms โ no change โ
โ QQuery 45 โ 49.07 / 49.48 ยฑ0.39 / 50.21 ms โ 45.12 / 45.53 ยฑ0.30 / 45.99 ms โ +1.09x faster โ
โ QQuery 46 โ 8.45 / 8.62 ยฑ0.17 / 8.94 ms โ 8.40 / 8.93 ยฑ0.72 / 10.34 ms โ no change โ
โ QQuery 47 โ 721.11 / 728.81 ยฑ6.71 / 739.05 ms โ 712.66 / 727.85 ยฑ8.87 / 737.07 ms โ no change โ
โ QQuery 48 โ 273.56 / 279.47 ยฑ3.40 / 282.46 ms โ 275.74 / 279.85 ยฑ3.00 / 282.91 ms โ no change โ
โ QQuery 49 โ 249.38 / 250.79 ยฑ1.38 / 253.29 ms โ 243.57 / 245.73 ยฑ3.01 / 251.62 ms โ no change โ
โ QQuery 50 โ 211.11 / 215.58 ยฑ5.61 / 226.62 ms โ 188.68 / 191.47 ยฑ3.15 / 197.40 ms โ +1.13x faster โ
โ QQuery 51 โ 177.55 / 181.32 ยฑ4.09 / 188.35 ms โ 171.87 / 174.52 ยฑ3.25 / 180.31 ms โ no change โ
โ QQuery 52 โ 107.27 / 107.78 ยฑ0.39 / 108.43 ms โ 107.69 / 107.95 ยฑ0.15 / 108.09 ms โ no change โ
โ QQuery 53 โ 102.58 / 104.49 ยฑ1.11 / 106.02 ms โ 102.28 / 103.20 ยฑ0.87 / 104.83 ms โ no change โ
โ QQuery 54 โ 147.05 / 147.81 ยฑ1.00 / 149.68 ms โ 144.82 / 146.59 ยฑ1.36 / 148.40 ms โ no change โ
โ QQuery 55 โ 106.73 / 107.14 ยฑ0.39 / 107.83 ms โ 105.87 / 107.30 ยฑ0.82 / 108.23 ms โ no change โ
โ QQuery 56 โ 141.38 / 142.06 ยฑ0.53 / 142.85 ms โ 139.46 / 140.46 ยฑ0.81 / 141.61 ms โ no change โ
โ QQuery 57 โ 165.39 / 168.77 ยฑ2.14 / 171.94 ms โ 162.27 / 164.31 ยฑ1.51 / 166.76 ms โ no change โ
โ QQuery 58 โ 310.40 / 311.72 ยฑ1.00 / 313.40 ms โ 282.71 / 284.64 ยฑ1.19 / 286.12 ms โ +1.10x faster โ
โ QQuery 59 โ 196.80 / 199.64 ยฑ1.89 / 202.36 ms โ 195.70 / 197.68 ยฑ2.01 / 201.06 ms โ no change โ
โ QQuery 60 โ 142.99 / 144.05 ยฑ0.69 / 145.05 ms โ 141.39 / 141.92 ยฑ0.43 / 142.48 ms โ no change โ
โ QQuery 61 โ 13.51 / 13.72 ยฑ0.15 / 13.89 ms โ 13.40 / 14.59 ยฑ2.12 / 18.82 ms โ 1.06x slower โ
โ QQuery 62 โ 894.52 / 901.10 ยฑ6.57 / 912.82 ms โ 867.53 / 876.05 ยฑ6.90 / 884.20 ms โ no change โ
โ QQuery 63 โ 101.81 / 103.36 ยฑ1.23 / 105.54 ms โ 102.56 / 103.27 ยฑ0.67 / 104.17 ms โ no change โ
โ QQuery 64 โ 672.30 / 677.18 ยฑ4.57 / 683.40 ms โ 633.93 / 643.07 ยฑ6.86 / 651.69 ms โ +1.05x faster โ
โ QQuery 65 โ 252.05 / 257.11 ยฑ3.88 / 261.72 ms โ 247.95 / 250.70 ยฑ2.21 / 254.42 ms โ no change โ
โ QQuery 66 โ 217.25 / 224.32 ยฑ10.33 / 244.59 ms โ 213.85 / 221.04 ยฑ6.56 / 231.91 ms โ no change โ
โ QQuery 67 โ 300.52 / 309.45 ยฑ10.36 / 323.98 ms โ 295.22 / 303.00 ยฑ11.41 / 325.42 ms โ no change โ
โ QQuery 68 โ 8.75 / 8.91 ยฑ0.18 / 9.25 ms โ 8.61 / 8.77 ยฑ0.17 / 9.11 ms โ no change โ
โ QQuery 69 โ 98.76 / 100.40 ยฑ1.50 / 102.95 ms โ 99.20 / 101.70 ยฑ4.19 / 110.06 ms โ no change โ
โ QQuery 70 โ 316.30 / 330.46 ยฑ12.84 / 354.73 ms โ 312.75 / 323.98 ยฑ9.85 / 338.46 ms โ no change โ
โ QQuery 71 โ 133.22 / 134.92 ยฑ0.93 / 135.95 ms โ 134.13 / 135.45 ยฑ1.90 / 139.19 ms โ no change โ
โ QQuery 72 โ 592.47 / 604.29 ยฑ8.50 / 616.24 ms โ 546.97 / 553.13 ยฑ6.66 / 565.53 ms โ +1.09x faster โ
โ QQuery 73 โ 6.85 / 6.96 ยฑ0.11 / 7.16 ms โ 6.64 / 6.83 ยฑ0.18 / 7.15 ms โ no change โ
โ QQuery 74 โ 601.89 / 612.30 ยฑ7.09 / 623.95 ms โ 590.60 / 594.95 ยฑ2.81 / 599.17 ms โ no change โ
โ QQuery 75 โ 268.04 / 270.51 ยฑ1.81 / 273.04 ms โ 268.07 / 269.62 ยฑ1.35 / 271.89 ms โ no change โ
โ QQuery 76 โ 130.05 / 132.34 ยฑ1.78 / 135.12 ms โ 130.54 / 131.90 ยฑ1.46 / 134.63 ms โ no change โ
โ QQuery 77 โ 187.37 / 188.89 ยฑ1.69 / 192.08 ms โ 186.96 / 188.49 ยฑ1.14 / 189.92 ms โ no change โ
โ QQuery 78 โ 332.61 / 335.31 ยฑ2.79 / 339.19 ms โ 310.77 / 313.49 ยฑ2.37 / 317.50 ms โ +1.07x faster โ
โ QQuery 79 โ 232.16 / 237.36 ยฑ4.05 / 244.01 ms โ 230.79 / 233.28 ยฑ1.96 / 236.40 ms โ no change โ
โ QQuery 80 โ 319.69 / 321.23 ยฑ1.40 / 323.88 ms โ 301.52 / 302.78 ยฑ1.00 / 304.19 ms โ +1.06x faster โ
โ QQuery 81 โ 26.20 / 26.78 ยฑ0.43 / 27.42 ms โ 26.07 / 26.30 ยฑ0.15 / 26.46 ms โ no change โ
โ QQuery 82 โ 39.23 / 39.68 ยฑ0.39 / 40.23 ms โ 39.67 / 40.06 ยฑ0.43 / 40.85 ms โ no change โ
โ QQuery 83 โ 37.72 / 38.02 ยฑ0.22 / 38.31 ms โ 36.00 / 36.37 ยฑ0.46 / 37.24 ms โ no change โ
โ QQuery 84 โ 46.96 / 47.19 ยฑ0.20 / 47.56 ms โ 46.75 / 46.86 ยฑ0.10 / 47.00 ms โ no change โ
โ QQuery 85 โ 141.94 / 143.24 ยฑ1.11 / 145.23 ms โ 142.33 / 143.94 ยฑ1.83 / 147.04 ms โ no change โ
โ QQuery 86 โ 37.65 / 38.06 ยฑ0.27 / 38.44 ms โ 37.56 / 38.11 ยฑ0.44 / 38.83 ms โ no change โ
โ QQuery 87 โ 3.46 / 3.60 ยฑ0.18 / 3.93 ms โ 3.49 / 3.60 ยฑ0.19 / 3.98 ms โ no change โ
โ QQuery 88 โ 100.42 / 103.31 ยฑ3.11 / 107.75 ms โ 99.69 / 100.18 ยฑ0.31 / 100.60 ms โ no change โ
โ QQuery 89 โ 116.87 / 117.76 ยฑ1.09 / 119.87 ms โ 117.15 / 118.86 ยฑ1.94 / 122.49 ms โ no change โ
โ QQuery 90 โ 22.76 / 23.01 ยฑ0.24 / 23.43 ms โ 22.72 / 23.02 ยฑ0.20 / 23.27 ms โ no change โ
โ QQuery 91 โ 58.55 / 59.88 ยฑ0.88 / 60.75 ms โ 58.69 / 59.29 ยฑ0.57 / 60.17 ms โ no change โ
โ QQuery 92 โ 56.85 / 57.11 ยฑ0.24 / 57.53 ms โ 56.63 / 57.41 ยฑ0.78 / 58.81 ms โ no change โ
โ QQuery 93 โ 180.76 / 182.78 ยฑ1.43 / 184.56 ms โ 162.42 / 163.37 ยฑ0.94 / 164.84 ms โ +1.12x faster โ
โ QQuery 94 โ 60.78 / 61.39 ยฑ0.35 / 61.76 ms โ 61.18 / 61.53 ยฑ0.53 / 62.59 ms โ no change โ
โ QQuery 95 โ 126.80 / 127.74 ยฑ0.78 / 128.60 ms โ 111.81 / 112.73 ยฑ0.68 / 113.42 ms โ +1.13x faster โ
โ QQuery 96 โ 68.10 / 69.58 ยฑ1.26 / 71.14 ms โ 67.25 / 69.44 ยฑ2.07 / 73.33 ms โ no change โ
โ QQuery 97 โ 117.11 / 118.55 ยฑ1.57 / 121.55 ms โ 112.06 / 113.17 ยฑ0.89 / 114.31 ms โ no change โ
โ QQuery 98 โ 149.65 / 151.58 ยฑ1.23 / 153.40 ms โ 149.67 / 151.59 ยฑ1.15 / 152.76 ms โ no change โ
โ QQuery 99 โ 10817.90 / 10877.58 ยฑ82.16 / 11036.81 ms โ 10801.80 / 10876.01 ยฑ72.49 / 10980.77 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 30960.91ms โ
โ Total Time (perf-strength-reduce-hash-partition) โ 30456.67ms โ
โ Average Time (HEAD) โ 312.74ms โ
โ Average Time (perf-strength-reduce-hash-partition) โ 307.64ms โ
โ Queries Faster โ 14 โ
โ Queries Slower โ 1 โ
โ Queries with No Change โ 84 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpcds โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 160.0s |
| Peak memory | 6.3 GiB |
| Avg memory | 5.5 GiB |
| CPU user | 259.1s |
| CPU sys | 8.5s |
| Peak spill | 0 B |
**tpcds โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 155.0s |
| Peak memory | 6.6 GiB |
| Avg memory | 5.5 GiB |
| CPU user | 254.5s |
| CPU sys | 8.1s |
| 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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338939588
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338923527)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4338923527-1883-8bpt5 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 perf-strength-reduce-hash-partition (b0091c6806aeefa4475e65427e1458a98d00f800) to 65f337d (merge-base) [diff](https://github.com/apache/datafusion/compare/65f337d6637aa069012e0e9286f859fb3a743253..b0091c6806aeefa4475e65427e1458a98d00f800) using: tpch10
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338862027
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338749432)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
<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 perf-strength-reduce-hash-partition
--------------------
Benchmark tpch_sf1.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ perf-strength-reduce-hash-partition โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 40.64 / 41.83 ยฑ1.35 / 43.54 ms โ 40.56 / 41.44 ยฑ0.96 / 43.27 ms โ no change โ
โ QQuery 2 โ 20.89 / 21.14 ยฑ0.14 / 21.32 ms โ 21.05 / 21.35 ยฑ0.21 / 21.69 ms โ no change โ
โ QQuery 3 โ 37.89 / 39.86 ยฑ1.28 / 41.59 ms โ 36.72 / 38.25 ยฑ1.17 / 39.48 ms โ no change โ
โ QQuery 4 โ 18.37 / 18.86 ยฑ0.56 / 19.89 ms โ 18.53 / 19.00 ยฑ0.53 / 20.01 ms โ no change โ
โ QQuery 5 โ 49.28 / 51.76 ยฑ1.46 / 53.79 ms โ 44.11 / 46.32 ยฑ1.29 / 47.71 ms โ +1.12x faster โ
โ QQuery 6 โ 17.13 / 17.33 ยฑ0.11 / 17.44 ms โ 17.43 / 18.02 ยฑ0.63 / 19.25 ms โ no change โ
โ QQuery 7 โ 54.18 / 56.84 ยฑ2.45 / 60.69 ms โ 51.53 / 53.42 ยฑ1.46 / 55.52 ms โ +1.06x faster โ
โ QQuery 8 โ 48.08 / 48.44 ยฑ0.23 / 48.72 ms โ 47.34 / 48.06 ยฑ0.70 / 49.19 ms โ no change โ
โ QQuery 9 โ 53.33 / 54.47 ยฑ0.66 / 55.12 ms โ 52.22 / 52.46 ยฑ0.18 / 52.67 ms โ no change โ
โ QQuery 10 โ 65.82 / 67.42 ยฑ1.32 / 69.69 ms โ 66.21 / 67.29 ยฑ1.26 / 68.88 ms โ no change โ
โ QQuery 11 โ 14.35 / 14.44 ยฑ0.09 / 14.55 ms โ 14.20 / 14.63 ยฑ0.28 / 14.96 ms โ no change โ
โ QQuery 12 โ 27.46 / 27.80 ยฑ0.43 / 28.65 ms โ 26.68 / 27.30 ยฑ0.92 / 29.12 ms โ no change โ
โ QQuery 13 โ 38.07 / 39.02 ยฑ0.86 / 40.62 ms โ 35.58 / 36.44 ยฑ0.57 / 37.31 ms โ +1.07x faster โ
โ QQuery 14 โ 27.87 / 28.14 ยฑ0.29 / 28.70 ms โ 27.07 / 27.37 ยฑ0.31 / 27.97 ms โ no change โ
โ QQuery 15 โ 33.61 / 34.11 ยฑ0.67 / 35.42 ms โ 33.99 / 34.33 ยฑ0.51 / 35.35 ms โ no change โ
โ QQuery 16 โ 15.38 / 15.70 ยฑ0.43 / 16.56 ms โ 15.39 / 15.52 ยฑ0.13 / 15.78 ms โ no change โ
โ QQuery 17 โ 79.25 / 80.58 ยฑ0.95 / 82.02 ms โ 78.36 / 79.06 ยฑ0.67 / 80.18 ms โ no change โ
โ QQuery 18 โ 75.87 / 76.76 ยฑ0.64 / 77.61 ms โ 69.57 / 70.15 ยฑ0.59 / 71.09 ms โ +1.09x faster โ
โ QQuery 19 โ 37.59 / 37.92 ยฑ0.18 / 38.14 ms โ 38.09 / 38.65 ยฑ0.50 / 39.37 ms โ no change โ
โ QQuery 20 โ 39.66 / 41.60 ยฑ2.98 / 47.42 ms โ 39.38 / 39.60 ยฑ0.21 / 40.01 ms โ no change โ
โ QQuery 21 โ 65.16 / 66.87 ยฑ1.05 / 68.00 ms โ 61.41 / 62.27 ยฑ0.74 / 63.23 ms โ +1.07x faster โ
โ QQuery 22 โ 17.23 / 17.74 ยฑ0.44 / 18.49 ms โ 17.40 / 17.47 ยฑ0.08 / 17.62 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 898.64ms โ
โ Total Time (perf-strength-reduce-hash-partition) โ 868.39ms โ
โ Average Time (HEAD) โ 40.85ms โ
โ Average Time (perf-strength-reduce-hash-partition) โ 39.47ms โ
โ Queries Faster โ 5 โ
โ Queries Slower โ 0 โ
โ Queries with No Change โ 17 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpch โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 5.5 GiB |
| Avg memory | 4.8 GiB |
| CPU user | 33.9s |
| CPU sys | 2.4s |
| Peak spill | 0 B |
**tpch โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 5.2 GiB |
| Avg memory | 4.8 GiB |
| CPU user | 32.8s |
| CPU sys | 2.6s |
| 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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338889625
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338764766)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4338764766-1882-b8hbm 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 perf-strength-reduce-hash-partition (b0091c6806aeefa4475e65427e1458a98d00f800) to 65f337d (merge-base) [diff](https://github.com/apache/datafusion/compare/65f337d6637aa069012e0e9286f859fb3a743253..b0091c6806aeefa4475e65427e1458a98d00f800) using: tpch10
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] perf: strength reduce hash partition modulo (up to 1.16x faster) [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on code in PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#discussion_r3161571847
##########
datafusion/physical-plan/src/repartition/mod.rs:
##########
@@ -446,22 +511,27 @@ impl BatchPartitioner {
/// - `num_partitions`: Total number of output partitions.
/// - `timer`: Metric used to record time spent during repartitioning.
///
- /// # Notes
- /// This constructor cannot fail and performs no validation.
+ /// # Errors
+ /// Returns an error if `num_partitions` is zero.
pub fn new_hash_partitioner(
exprs: Vec<Arc<dyn PhysicalExpr>>,
num_partitions: usize,
timer: metrics::Time,
- ) -> Self {
- Self {
+ ) -> Result<Self> {
Review Comment:
Technically a breaking change but I think this is one of those things that is more so public for the workspace than it is for end users. So may not be worth calling out.
##########
datafusion/physical-plan/src/repartition/mod.rs:
##########
@@ -575,7 +645,8 @@ impl BatchPartitioner {
}
BatchPartitionerState::Hash {
exprs,
- num_partitions: partitions,
+ num_partitions: _,
Review Comment:
Dead code? Should we remove?
--
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338749432
run benchmarks
--
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338764222
run benchmarks
--
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338961346
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338764766)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
<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 perf-strength-reduce-hash-partition
--------------------
Benchmark tpch_sf10.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ perf-strength-reduce-hash-partition โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 327.64 / 329.66 ยฑ1.44 / 331.84 ms โ 327.82 / 330.00 ยฑ1.68 / 331.48 ms โ no change โ
โ QQuery 2 โ 131.35 / 138.32 ยฑ3.91 / 143.24 ms โ 125.94 / 126.49 ยฑ0.61 / 127.42 ms โ +1.09x faster โ
โ QQuery 3 โ 286.47 / 300.24 ยฑ8.19 / 308.99 ms โ 273.76 / 276.01 ยฑ1.80 / 277.98 ms โ +1.09x faster โ
โ QQuery 4 โ 158.30 / 160.81 ยฑ2.21 / 163.54 ms โ 137.56 / 138.79 ยฑ0.98 / 139.80 ms โ +1.16x faster โ
โ QQuery 5 โ 428.90 / 437.68 ยฑ4.45 / 440.83 ms โ 390.52 / 396.51 ยฑ4.30 / 403.67 ms โ +1.10x faster โ
โ QQuery 6 โ 131.88 / 132.83 ยฑ1.17 / 134.81 ms โ 133.06 / 134.70 ยฑ1.22 / 135.83 ms โ no change โ
โ QQuery 7 โ 541.09 / 545.88 ยฑ4.21 / 552.67 ms โ 508.51 / 531.75 ยฑ16.82 / 548.21 ms โ no change โ
โ QQuery 8 โ 467.86 / 476.44 ยฑ6.95 / 483.87 ms โ 427.19 / 439.03 ยฑ9.88 / 453.56 ms โ +1.09x faster โ
โ QQuery 9 โ 649.16 / 660.07 ยฑ10.12 / 676.72 ms โ 605.25 / 611.87 ยฑ5.72 / 620.70 ms โ +1.08x faster โ
โ QQuery 10 โ 327.64 / 339.90 ยฑ6.92 / 348.85 ms โ 321.66 / 330.67 ยฑ4.76 / 334.89 ms โ no change โ
โ QQuery 11 โ 104.93 / 107.54 ยฑ1.71 / 110.18 ms โ 92.80 / 101.35 ยฑ12.27 / 125.63 ms โ +1.06x faster โ
โ QQuery 12 โ 198.96 / 202.37 ยฑ2.45 / 206.21 ms โ 195.07 / 197.77 ยฑ4.26 / 206.26 ms โ no change โ
โ QQuery 13 โ 300.44 / 312.37 ยฑ6.87 / 321.90 ms โ 291.85 / 308.47 ยฑ10.23 / 317.55 ms โ no change โ
โ QQuery 14 โ 188.06 / 193.71 ยฑ4.81 / 200.69 ms โ 182.89 / 186.72 ยฑ3.67 / 192.75 ms โ no change โ
โ QQuery 15 โ 334.88 / 339.95 ยฑ5.81 / 350.78 ms โ 330.79 / 336.21 ยฑ4.31 / 342.71 ms โ no change โ
โ QQuery 16 โ 78.38 / 81.25 ยฑ2.51 / 84.55 ms โ 74.35 / 76.61 ยฑ2.80 / 81.94 ms โ +1.06x faster โ
โ QQuery 17 โ 744.08 / 761.70 ยฑ12.84 / 781.69 ms โ 703.40 / 724.66 ยฑ23.35 / 770.05 ms โ no change โ
โ QQuery 18 โ 760.17 / 782.23 ยฑ12.12 / 796.85 ms โ 725.45 / 744.71 ยฑ15.59 / 765.59 ms โ no change โ
โ QQuery 19 โ 267.90 / 280.99 ยฑ14.61 / 306.80 ms โ 275.58 / 298.23 ยฑ27.69 / 351.75 ms โ 1.06x slower โ
โ QQuery 20 โ 311.46 / 323.12 ยฑ10.13 / 341.26 ms โ 312.13 / 319.42 ยฑ4.39 / 324.46 ms โ no change โ
โ QQuery 21 โ 816.40 / 837.33 ยฑ19.78 / 870.18 ms โ 766.23 / 778.58 ยฑ8.98 / 792.31 ms โ +1.08x faster โ
โ QQuery 22 โ 81.46 / 84.94 ยฑ2.58 / 88.20 ms โ 75.31 / 77.73 ยฑ1.39 / 79.55 ms โ +1.09x faster โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 7829.34ms โ
โ Total Time (perf-strength-reduce-hash-partition) โ 7466.29ms โ
โ Average Time (HEAD) โ 355.88ms โ
โ Average Time (perf-strength-reduce-hash-partition) โ 339.38ms โ
โ Queries Faster โ 10 โ
โ Queries Slower โ 1 โ
โ Queries with No Change โ 11 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpch10 โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 40.0s |
| Peak memory | 11.8 GiB |
| Avg memory | 8.1 GiB |
| CPU user | 419.5s |
| CPU sys | 21.6s |
| Peak spill | 0 B |
**tpch10 โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 40.0s |
| Peak memory | 11.7 GiB |
| Avg memory | 8.0 GiB |
| CPU user | 399.7s |
| CPU sys | 21.6s |
| 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] perf: strength reduce hash partition modulo [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338923527
run benchmark tpch10
--
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] perf: strength reduce hash partition modulo (up to 1.16x faster) [datafusion]
Posted by "adriangb (via GitHub)" <gi...@apache.org>.
adriangb commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4344444587
I think some more docstrings linking to explainations of what this is, other example implementations, why it helps ,etc. would be nice
--
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338899826
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338764222)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
<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 perf-strength-reduce-hash-partition
--------------------
Benchmark tpcds_sf1.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ perf-strength-reduce-hash-partition โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 6.78 / 7.39 ยฑ0.82 / 9.02 ms โ 7.04 / 7.48 ยฑ0.71 / 8.89 ms โ no change โ
โ QQuery 2 โ 145.52 / 145.78 ยฑ0.22 / 146.08 ms โ 146.38 / 146.80 ยฑ0.25 / 147.15 ms โ no change โ
โ QQuery 3 โ 113.42 / 114.48 ยฑ0.67 / 115.23 ms โ 113.21 / 114.23 ยฑ0.95 / 115.98 ms โ no change โ
โ QQuery 4 โ 1327.95 / 1337.03 ยฑ8.77 / 1351.05 ms โ 1329.81 / 1335.48 ยฑ4.28 / 1341.10 ms โ no change โ
โ QQuery 5 โ 172.79 / 173.72 ยฑ0.48 / 174.19 ms โ 173.62 / 175.64 ยฑ1.90 / 178.74 ms โ no change โ
โ QQuery 6 โ 846.21 / 859.28 ยฑ8.69 / 871.10 ms โ 813.57 / 848.98 ยฑ27.46 / 887.43 ms โ no change โ
โ QQuery 7 โ 334.91 / 336.28 ยฑ0.84 / 337.37 ms โ 322.04 / 324.51 ยฑ3.10 / 330.45 ms โ no change โ
โ QQuery 8 โ 113.10 / 113.61 ยฑ0.54 / 114.60 ms โ 113.57 / 113.84 ยฑ0.31 / 114.45 ms โ no change โ
โ QQuery 9 โ 100.25 / 103.54 ยฑ2.62 / 106.36 ms โ 102.10 / 109.03 ยฑ9.13 / 126.81 ms โ 1.05x slower โ
โ QQuery 10 โ 103.04 / 104.22 ยฑ0.64 / 104.88 ms โ 103.07 / 104.66 ยฑ1.44 / 107.35 ms โ no change โ
โ QQuery 11 โ 935.10 / 939.58 ยฑ3.77 / 943.96 ms โ 932.82 / 942.12 ยฑ8.46 / 955.34 ms โ no change โ
โ QQuery 12 โ 44.26 / 44.73 ยฑ0.42 / 45.48 ms โ 43.84 / 44.07 ยฑ0.19 / 44.33 ms โ no change โ
โ QQuery 13 โ 390.03 / 393.99 ยฑ3.01 / 399.02 ms โ 393.39 / 394.59 ยฑ1.18 / 396.69 ms โ no change โ
โ QQuery 14 โ 982.99 / 992.09 ยฑ9.29 / 1006.57 ms โ 943.88 / 955.36 ยฑ6.25 / 961.80 ms โ no change โ
โ QQuery 15 โ 15.43 / 15.72 ยฑ0.23 / 16.02 ms โ 14.88 / 16.19 ยฑ1.64 / 19.38 ms โ no change โ
โ QQuery 16 โ 7.28 / 7.44 ยฑ0.19 / 7.80 ms โ 7.34 / 7.50 ยฑ0.17 / 7.83 ms โ no change โ
โ QQuery 17 โ 225.60 / 227.00 ยฑ1.16 / 228.47 ms โ 202.74 / 204.41 ยฑ1.10 / 206.13 ms โ +1.11x faster โ
โ QQuery 18 โ 128.32 / 129.56 ยฑ0.66 / 130.26 ms โ 123.09 / 124.38 ยฑ0.67 / 124.98 ms โ no change โ
โ QQuery 19 โ 156.55 / 157.29 ยฑ0.44 / 157.84 ms โ 155.22 / 156.63 ยฑ1.38 / 159.02 ms โ no change โ
โ QQuery 20 โ 13.15 / 13.40 ยฑ0.14 / 13.54 ms โ 13.10 / 13.27 ยฑ0.12 / 13.42 ms โ no change โ
โ QQuery 21 โ 19.23 / 19.93 ยฑ0.75 / 21.39 ms โ 19.43 / 19.60 ยฑ0.12 / 19.78 ms โ no change โ
โ QQuery 22 โ 493.23 / 497.87 ยฑ2.87 / 501.32 ms โ 482.34 / 484.82 ยฑ1.67 / 486.33 ms โ no change โ
โ QQuery 23 โ 826.65 / 834.06 ยฑ6.20 / 844.40 ms โ 787.82 / 792.26 ยฑ3.09 / 795.67 ms โ +1.05x faster โ
โ QQuery 24 โ 372.57 / 376.45 ยฑ3.18 / 381.91 ms โ 337.25 / 341.00 ยฑ2.40 / 344.59 ms โ +1.10x faster โ
โ QQuery 25 โ 333.76 / 335.59 ยฑ1.21 / 337.56 ms โ 308.52 / 310.64 ยฑ1.59 / 313.44 ms โ +1.08x faster โ
โ QQuery 26 โ 78.27 / 78.71 ยฑ0.38 / 79.27 ms โ 76.52 / 77.35 ยฑ1.10 / 79.45 ms โ no change โ
โ QQuery 27 โ 6.86 / 6.96 ยฑ0.19 / 7.35 ms โ 6.76 / 6.93 ยฑ0.16 / 7.23 ms โ no change โ
โ QQuery 28 โ 149.21 / 149.93 ยฑ0.79 / 151.38 ms โ 149.68 / 151.15 ยฑ1.57 / 153.13 ms โ no change โ
โ QQuery 29 โ 274.31 / 277.54 ยฑ3.45 / 283.42 ms โ 247.03 / 248.21 ยฑ1.22 / 249.79 ms โ +1.12x faster โ
โ QQuery 30 โ 42.02 / 42.89 ยฑ1.10 / 45.03 ms โ 41.83 / 42.27 ยฑ0.34 / 42.82 ms โ no change โ
โ QQuery 31 โ 165.43 / 167.33 ยฑ2.24 / 171.70 ms โ 164.82 / 166.06 ยฑ1.30 / 168.38 ms โ no change โ
โ QQuery 32 โ 13.82 / 13.96 ยฑ0.21 / 14.38 ms โ 13.68 / 13.98 ยฑ0.20 / 14.20 ms โ no change โ
โ QQuery 33 โ 139.80 / 141.38 ยฑ1.57 / 144.15 ms โ 139.20 / 140.47 ยฑ0.80 / 141.35 ms โ no change โ
โ QQuery 34 โ 7.12 / 7.23 ยฑ0.16 / 7.53 ms โ 6.93 / 7.08 ยฑ0.18 / 7.42 ms โ no change โ
โ QQuery 35 โ 101.33 / 102.34 ยฑ0.97 / 103.85 ms โ 100.57 / 102.96 ยฑ2.24 / 106.01 ms โ no change โ
โ QQuery 36 โ 6.66 / 6.92 ยฑ0.15 / 7.05 ms โ 6.50 / 6.73 ยฑ0.18 / 6.98 ms โ no change โ
โ QQuery 37 โ 8.43 / 8.53 ยฑ0.05 / 8.60 ms โ 8.38 / 8.55 ยฑ0.26 / 9.06 ms โ no change โ
โ QQuery 38 โ 86.69 / 88.73 ยฑ1.08 / 89.88 ms โ 86.30 / 87.95 ยฑ1.54 / 90.02 ms โ no change โ
โ QQuery 39 โ 119.46 / 122.11 ยฑ3.68 / 129.22 ms โ 116.84 / 120.49 ยฑ5.23 / 130.87 ms โ no change โ
โ QQuery 40 โ 104.12 / 108.15 ยฑ3.99 / 115.41 ms โ 97.96 / 100.34 ยฑ1.91 / 102.82 ms โ +1.08x faster โ
โ QQuery 41 โ 14.38 / 15.80 ยฑ2.58 / 20.94 ms โ 14.30 / 14.48 ยฑ0.17 / 14.81 ms โ +1.09x faster โ
โ QQuery 42 โ 107.33 / 108.54 ยฑ0.75 / 109.49 ms โ 107.44 / 108.92 ยฑ2.00 / 112.78 ms โ no change โ
โ QQuery 43 โ 5.65 / 5.78 ยฑ0.11 / 5.97 ms โ 5.53 / 5.74 ยฑ0.18 / 6.07 ms โ no change โ
โ QQuery 44 โ 11.68 / 11.87 ยฑ0.10 / 11.99 ms โ 11.58 / 11.73 ยฑ0.09 / 11.85 ms โ no change โ
โ QQuery 45 โ 48.71 / 49.07 ยฑ0.29 / 49.48 ms โ 45.32 / 46.72 ยฑ2.25 / 51.19 ms โ no change โ
โ QQuery 46 โ 8.39 / 8.56 ยฑ0.17 / 8.88 ms โ 8.31 / 8.52 ยฑ0.18 / 8.81 ms โ no change โ
โ QQuery 47 โ 716.95 / 722.81 ยฑ4.62 / 728.71 ms โ 719.32 / 724.20 ยฑ2.64 / 726.83 ms โ no change โ
โ QQuery 48 โ 273.86 / 279.65 ยฑ5.29 / 287.20 ms โ 276.75 / 278.53 ยฑ1.99 / 281.74 ms โ no change โ
โ QQuery 49 โ 247.57 / 250.62 ยฑ2.40 / 253.81 ms โ 244.58 / 245.53 ยฑ0.50 / 245.96 ms โ no change โ
โ QQuery 50 โ 204.37 / 212.28 ยฑ5.23 / 219.40 ms โ 187.26 / 192.56 ยฑ4.79 / 201.25 ms โ +1.10x faster โ
โ QQuery 51 โ 175.13 / 177.54 ยฑ2.13 / 181.25 ms โ 172.47 / 173.81 ยฑ1.39 / 176.10 ms โ no change โ
โ QQuery 52 โ 106.66 / 107.71 ยฑ0.91 / 108.86 ms โ 107.64 / 108.96 ยฑ1.66 / 112.10 ms โ no change โ
โ QQuery 53 โ 101.03 / 102.34 ยฑ1.87 / 106.01 ms โ 103.07 / 103.62 ยฑ0.40 / 104.28 ms โ no change โ
โ QQuery 54 โ 145.00 / 146.16 ยฑ0.78 / 146.88 ms โ 144.32 / 145.86 ยฑ1.29 / 148.13 ms โ no change โ
โ QQuery 55 โ 106.33 / 107.14 ยฑ0.52 / 107.86 ms โ 106.18 / 107.33 ยฑ0.80 / 108.53 ms โ no change โ
โ QQuery 56 โ 138.88 / 140.51 ยฑ1.29 / 142.32 ms โ 139.86 / 140.31 ยฑ0.41 / 141.06 ms โ no change โ
โ QQuery 57 โ 163.94 / 165.36 ยฑ0.90 / 166.25 ms โ 163.90 / 165.91 ยฑ1.41 / 168.10 ms โ no change โ
โ QQuery 58 โ 309.14 / 310.72 ยฑ1.68 / 312.82 ms โ 283.95 / 285.95 ยฑ1.95 / 289.29 ms โ +1.09x faster โ
โ QQuery 59 โ 193.79 / 195.89 ยฑ1.25 / 197.53 ms โ 193.26 / 194.50 ยฑ0.85 / 195.48 ms โ no change โ
โ QQuery 60 โ 141.43 / 142.37 ยฑ1.22 / 144.61 ms โ 140.77 / 142.28 ยฑ1.11 / 144.01 ms โ no change โ
โ QQuery 61 โ 13.46 / 13.59 ยฑ0.12 / 13.83 ms โ 13.34 / 13.43 ยฑ0.12 / 13.68 ms โ no change โ
โ QQuery 62 โ 885.17 / 892.94 ยฑ7.51 / 903.56 ms โ 876.62 / 884.83 ยฑ10.00 / 903.25 ms โ no change โ
โ QQuery 63 โ 101.57 / 103.90 ยฑ2.91 / 109.59 ms โ 102.90 / 104.56 ยฑ2.42 / 109.35 ms โ no change โ
โ QQuery 64 โ 664.35 / 669.43 ยฑ4.14 / 675.24 ms โ 637.93 / 640.48 ยฑ2.14 / 643.93 ms โ no change โ
โ QQuery 65 โ 247.98 / 250.80 ยฑ2.31 / 254.10 ms โ 248.47 / 251.17 ยฑ2.52 / 255.81 ms โ no change โ
โ QQuery 66 โ 215.83 / 224.60 ยฑ7.11 / 236.58 ms โ 217.18 / 226.88 ยฑ9.06 / 240.34 ms โ no change โ
โ QQuery 67 โ 294.94 / 299.50 ยฑ5.51 / 310.35 ms โ 294.54 / 301.53 ยฑ6.71 / 312.69 ms โ no change โ
โ QQuery 68 โ 8.55 / 11.95 ยฑ4.04 / 18.09 ms โ 8.71 / 10.29 ยฑ2.89 / 16.06 ms โ +1.16x faster โ
โ QQuery 69 โ 98.73 / 99.75 ยฑ0.92 / 101.31 ms โ 98.92 / 100.41 ยฑ1.11 / 102.00 ms โ no change โ
โ QQuery 70 โ 314.33 / 321.93 ยฑ5.59 / 330.31 ms โ 327.39 / 332.12 ยฑ4.10 / 339.77 ms โ no change โ
โ QQuery 71 โ 133.21 / 134.22 ยฑ1.19 / 136.49 ms โ 133.68 / 134.59 ยฑ0.87 / 136.22 ms โ no change โ
โ QQuery 72 โ 588.56 / 596.20 ยฑ7.54 / 609.08 ms โ 538.78 / 550.31 ยฑ5.91 / 555.21 ms โ +1.08x faster โ
โ QQuery 73 โ 6.51 / 6.65 ยฑ0.19 / 7.01 ms โ 6.60 / 6.73 ยฑ0.19 / 7.11 ms โ no change โ
โ QQuery 74 โ 591.10 / 598.82 ยฑ4.84 / 605.87 ms โ 590.35 / 595.52 ยฑ3.17 / 598.83 ms โ no change โ
โ QQuery 75 โ 266.03 / 268.88 ยฑ2.81 / 273.64 ms โ 267.07 / 269.34 ยฑ2.17 / 272.02 ms โ no change โ
โ QQuery 76 โ 129.82 / 130.52 ยฑ0.39 / 130.83 ms โ 129.50 / 131.01 ยฑ1.05 / 132.32 ms โ no change โ
โ QQuery 77 โ 185.28 / 187.90 ยฑ2.44 / 192.32 ms โ 186.43 / 187.45 ยฑ0.67 / 188.39 ms โ no change โ
โ QQuery 78 โ 329.53 / 331.34 ยฑ1.84 / 334.80 ms โ 311.01 / 313.83 ยฑ2.64 / 318.71 ms โ +1.06x faster โ
โ QQuery 79 โ 230.58 / 233.60 ยฑ2.74 / 238.11 ms โ 229.81 / 233.77 ยฑ2.51 / 236.70 ms โ no change โ
โ QQuery 80 โ 317.97 / 319.76 ยฑ1.10 / 320.88 ms โ 300.59 / 302.41 ยฑ1.26 / 303.86 ms โ +1.06x faster โ
โ QQuery 81 โ 26.27 / 26.56 ยฑ0.21 / 26.91 ms โ 26.57 / 26.73 ยฑ0.15 / 26.94 ms โ no change โ
โ QQuery 82 โ 39.30 / 39.59 ยฑ0.30 / 40.17 ms โ 39.49 / 39.77 ยฑ0.30 / 40.32 ms โ no change โ
โ QQuery 83 โ 37.10 / 37.52 ยฑ0.32 / 37.90 ms โ 35.44 / 35.71 ยฑ0.18 / 35.96 ms โ no change โ
โ QQuery 84 โ 46.47 / 46.81 ยฑ0.37 / 47.53 ms โ 46.34 / 47.01 ยฑ0.45 / 47.69 ms โ no change โ
โ QQuery 85 โ 142.17 / 143.57 ยฑ1.21 / 145.68 ms โ 141.44 / 142.56 ยฑ1.02 / 144.02 ms โ no change โ
โ QQuery 86 โ 37.47 / 38.66 ยฑ2.05 / 42.75 ms โ 37.37 / 38.63 ยฑ1.92 / 42.39 ms โ no change โ
โ QQuery 87 โ 3.44 / 3.55 ยฑ0.19 / 3.92 ms โ 3.43 / 3.55 ยฑ0.19 / 3.94 ms โ no change โ
โ QQuery 88 โ 99.19 / 99.69 ยฑ0.51 / 100.54 ms โ 99.54 / 100.47 ยฑ1.45 / 103.36 ms โ no change โ
โ QQuery 89 โ 116.05 / 117.00 ยฑ1.72 / 120.43 ms โ 117.41 / 119.45 ยฑ3.43 / 126.23 ms โ no change โ
โ QQuery 90 โ 22.28 / 23.43 ยฑ1.62 / 26.63 ms โ 22.53 / 22.94 ยฑ0.30 / 23.47 ms โ no change โ
โ QQuery 91 โ 58.97 / 60.09 ยฑ0.70 / 60.93 ms โ 59.57 / 60.60 ยฑ0.99 / 61.99 ms โ no change โ
โ QQuery 92 โ 56.53 / 56.91 ยฑ0.23 / 57.22 ms โ 56.72 / 57.15 ยฑ0.23 / 57.39 ms โ no change โ
โ QQuery 93 โ 179.34 / 181.50 ยฑ1.51 / 183.77 ms โ 162.76 / 165.42 ยฑ2.02 / 168.85 ms โ +1.10x faster โ
โ QQuery 94 โ 60.01 / 60.81 ยฑ0.47 / 61.37 ms โ 60.85 / 61.35 ยฑ0.40 / 61.82 ms โ no change โ
โ QQuery 95 โ 125.77 / 127.03 ยฑ1.00 / 128.43 ms โ 111.62 / 112.63 ยฑ1.18 / 114.87 ms โ +1.13x faster โ
โ QQuery 96 โ 68.94 / 70.22 ยฑ1.06 / 71.91 ms โ 68.27 / 69.55 ยฑ0.89 / 70.74 ms โ no change โ
โ QQuery 97 โ 116.19 / 118.01 ยฑ1.56 / 120.46 ms โ 112.73 / 114.26 ยฑ1.65 / 117.31 ms โ no change โ
โ QQuery 98 โ 150.86 / 152.05 ยฑ1.07 / 154.03 ms โ 149.52 / 152.00 ยฑ2.33 / 155.99 ms โ no change โ
โ QQuery 99 โ 10789.42 / 10840.79 ยฑ35.21 / 10886.03 ms โ 10788.21 / 10860.60 ยฑ108.62 / 11076.55 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 30803.51ms โ
โ Total Time (perf-strength-reduce-hash-partition) โ 30394.50ms โ
โ Average Time (HEAD) โ 311.15ms โ
โ Average Time (perf-strength-reduce-hash-partition) โ 307.02ms โ
โ Queries Faster โ 15 โ
โ Queries Slower โ 1 โ
โ Queries with No Change โ 83 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpcds โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 155.0s |
| Peak memory | 6.5 GiB |
| Avg memory | 5.5 GiB |
| CPU user | 257.7s |
| CPU sys | 8.2s |
| Peak spill | 0 B |
**tpcds โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 155.0s |
| Peak memory | 6.1 GiB |
| Avg memory | 5.5 GiB |
| CPU user | 253.7s |
| CPU sys | 8.2s |
| 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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338862734
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338764222)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
<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 perf-strength-reduce-hash-partition
--------------------
Benchmark tpch_sf1.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ perf-strength-reduce-hash-partition โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 40.38 / 41.37 ยฑ1.48 / 44.26 ms โ 40.24 / 41.14 ยฑ1.45 / 44.02 ms โ no change โ
โ QQuery 2 โ 20.81 / 21.41 ยฑ0.55 / 22.41 ms โ 21.02 / 21.48 ยฑ0.60 / 22.65 ms โ no change โ
โ QQuery 3 โ 38.37 / 39.02 ยฑ0.86 / 40.71 ms โ 36.31 / 37.72 ยฑ1.94 / 41.53 ms โ no change โ
โ QQuery 4 โ 18.32 / 18.55 ยฑ0.29 / 19.12 ms โ 18.54 / 18.76 ยฑ0.26 / 19.24 ms โ no change โ
โ QQuery 5 โ 47.55 / 50.49 ยฑ2.40 / 54.24 ms โ 44.18 / 46.39 ยฑ1.18 / 47.62 ms โ +1.09x faster โ
โ QQuery 6 โ 17.05 / 17.28 ยฑ0.12 / 17.40 ms โ 17.47 / 17.60 ยฑ0.16 / 17.91 ms โ no change โ
โ QQuery 7 โ 54.05 / 56.89 ยฑ2.17 / 59.44 ms โ 49.87 / 51.07 ยฑ1.12 / 53.03 ms โ +1.11x faster โ
โ QQuery 8 โ 47.86 / 48.71 ยฑ0.85 / 50.06 ms โ 46.89 / 47.50 ยฑ0.35 / 47.94 ms โ no change โ
โ QQuery 9 โ 53.14 / 54.17 ยฑ0.92 / 55.37 ms โ 51.77 / 53.27 ยฑ1.20 / 55.25 ms โ no change โ
โ QQuery 10 โ 65.36 / 66.00 ยฑ0.77 / 67.14 ms โ 65.42 / 65.90 ยฑ0.26 / 66.11 ms โ no change โ
โ QQuery 11 โ 13.83 / 14.15 ยฑ0.29 / 14.62 ms โ 14.26 / 14.69 ยฑ0.60 / 15.86 ms โ no change โ
โ QQuery 12 โ 27.27 / 27.55 ยฑ0.22 / 27.93 ms โ 26.75 / 26.95 ยฑ0.17 / 27.14 ms โ no change โ
โ QQuery 13 โ 37.24 / 37.89 ยฑ0.43 / 38.23 ms โ 35.56 / 36.10 ยฑ0.32 / 36.50 ms โ no change โ
โ QQuery 14 โ 27.66 / 28.15 ยฑ0.44 / 28.97 ms โ 27.02 / 27.13 ยฑ0.13 / 27.38 ms โ no change โ
โ QQuery 15 โ 33.39 / 33.84 ยฑ0.43 / 34.62 ms โ 33.75 / 34.52 ยฑ0.81 / 35.85 ms โ no change โ
โ QQuery 16 โ 15.20 / 15.38 ยฑ0.13 / 15.59 ms โ 15.30 / 15.44 ยฑ0.14 / 15.71 ms โ no change โ
โ QQuery 17 โ 78.12 / 78.42 ยฑ0.18 / 78.62 ms โ 77.08 / 77.62 ยฑ0.61 / 78.76 ms โ no change โ
โ QQuery 18 โ 74.61 / 75.35 ยฑ0.55 / 76.03 ms โ 69.09 / 69.77 ยฑ0.57 / 70.82 ms โ +1.08x faster โ
โ QQuery 19 โ 37.53 / 37.76 ยฑ0.26 / 38.15 ms โ 37.56 / 37.99 ยฑ0.29 / 38.28 ms โ no change โ
โ QQuery 20 โ 39.11 / 39.57 ยฑ0.36 / 40.10 ms โ 39.27 / 39.59 ยฑ0.25 / 40.01 ms โ no change โ
โ QQuery 21 โ 63.36 / 64.88 ยฑ1.26 / 66.30 ms โ 61.21 / 62.30 ยฑ0.97 / 64.10 ms โ no change โ
โ QQuery 22 โ 17.12 / 17.42 ยฑ0.24 / 17.78 ms โ 16.96 / 17.24 ยฑ0.15 / 17.42 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 884.24ms โ
โ Total Time (perf-strength-reduce-hash-partition) โ 860.18ms โ
โ Average Time (HEAD) โ 40.19ms โ
โ Average Time (perf-strength-reduce-hash-partition) โ 39.10ms โ
โ Queries Faster โ 3 โ
โ Queries Slower โ 0 โ
โ Queries with No Change โ 19 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpch โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 5.2 GiB |
| Avg memory | 4.8 GiB |
| CPU user | 33.5s |
| CPU sys | 2.4s |
| Peak spill | 0 B |
**tpch โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 5.0s |
| Peak memory | 5.2 GiB |
| Avg memory | 4.8 GiB |
| CPU user | 32.6s |
| CPU sys | 2.4s |
| 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] perf: strength reduce hash partition modulo (up to 1.16x faster) [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4339017111
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338923527)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB)
<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 perf-strength-reduce-hash-partition
--------------------
Benchmark tpch_sf10.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ perf-strength-reduce-hash-partition โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 329.46 / 331.44 ยฑ1.63 / 334.22 ms โ 328.18 / 329.01 ยฑ0.69 / 330.26 ms โ no change โ
โ QQuery 2 โ 144.13 / 147.30 ยฑ2.19 / 149.29 ms โ 125.51 / 129.32 ยฑ3.45 / 134.72 ms โ +1.14x faster โ
โ QQuery 3 โ 317.71 / 323.21 ยฑ5.42 / 330.22 ms โ 278.07 / 286.59 ยฑ5.81 / 292.86 ms โ +1.13x faster โ
โ QQuery 4 โ 163.25 / 167.88 ยฑ3.25 / 172.97 ms โ 142.43 / 144.89 ยฑ1.62 / 147.55 ms โ +1.16x faster โ
โ QQuery 5 โ 428.64 / 453.55 ยฑ19.04 / 471.91 ms โ 414.64 / 429.62 ยฑ8.54 / 438.47 ms โ +1.06x faster โ
โ QQuery 6 โ 131.43 / 133.76 ยฑ2.44 / 138.17 ms โ 139.74 / 143.05 ยฑ2.72 / 147.90 ms โ 1.07x slower โ
โ QQuery 7 โ 542.70 / 549.01 ยฑ6.34 / 560.84 ms โ 520.00 / 542.56 ยฑ25.56 / 588.57 ms โ no change โ
โ QQuery 8 โ 463.76 / 472.94 ยฑ5.62 / 481.42 ms โ 428.34 / 435.88 ยฑ5.16 / 443.86 ms โ +1.09x faster โ
โ QQuery 9 โ 670.60 / 673.19 ยฑ3.73 / 680.54 ms โ 608.61 / 615.30 ยฑ6.48 / 626.15 ms โ +1.09x faster โ
โ QQuery 10 โ 339.72 / 349.34 ยฑ9.17 / 366.40 ms โ 329.81 / 332.03 ยฑ1.70 / 334.11 ms โ no change โ
โ QQuery 11 โ 107.75 / 110.90 ยฑ3.27 / 115.38 ms โ 95.22 / 98.55 ยฑ2.64 / 102.96 ms โ +1.13x faster โ
โ QQuery 12 โ 200.49 / 208.29 ยฑ8.03 / 220.63 ms โ 194.28 / 205.25 ยฑ6.93 / 213.38 ms โ no change โ
โ QQuery 13 โ 307.84 / 323.57 ยฑ12.19 / 341.09 ms โ 328.16 / 337.99 ยฑ5.66 / 345.71 ms โ no change โ
โ QQuery 14 โ 194.52 / 196.14 ยฑ1.92 / 199.78 ms โ 194.49 / 198.40 ยฑ4.08 / 205.80 ms โ no change โ
โ QQuery 15 โ 351.09 / 353.63 ยฑ1.77 / 356.11 ms โ 334.24 / 352.93 ยฑ9.60 / 360.25 ms โ no change โ
โ QQuery 16 โ 83.52 / 88.26 ยฑ6.00 / 99.94 ms โ 74.54 / 77.42 ยฑ3.57 / 84.11 ms โ +1.14x faster โ
โ QQuery 17 โ 762.01 / 829.02 ยฑ52.11 / 873.27 ms โ 693.39 / 703.33 ยฑ11.48 / 725.76 ms โ +1.18x faster โ
โ QQuery 18 โ 771.48 / 797.11 ยฑ25.85 / 845.55 ms โ 716.33 / 731.66 ยฑ9.60 / 746.05 ms โ +1.09x faster โ
โ QQuery 19 โ 286.88 / 298.28 ยฑ10.49 / 317.02 ms โ 270.66 / 286.58 ยฑ19.84 / 324.34 ms โ no change โ
โ QQuery 20 โ 330.80 / 334.80 ยฑ3.30 / 340.27 ms โ 303.50 / 305.68 ยฑ1.73 / 308.76 ms โ +1.10x faster โ
โ QQuery 21 โ 836.96 / 845.87 ยฑ11.01 / 866.15 ms โ 754.59 / 771.43 ยฑ15.27 / 798.16 ms โ +1.10x faster โ
โ QQuery 22 โ 82.92 / 84.82 ยฑ1.05 / 85.86 ms โ 80.98 / 83.96 ยฑ1.79 / 86.15 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 8072.33ms โ
โ Total Time (perf-strength-reduce-hash-partition) โ 7541.41ms โ
โ Average Time (HEAD) โ 366.92ms โ
โ Average Time (perf-strength-reduce-hash-partition) โ 342.79ms โ
โ Queries Faster โ 12 โ
โ Queries Slower โ 1 โ
โ Queries with No Change โ 9 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpch10 โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 45.0s |
| Peak memory | 12.3 GiB |
| Avg memory | 8.3 GiB |
| CPU user | 432.4s |
| CPU sys | 22.1s |
| Peak spill | 0 B |
**tpch10 โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 40.0s |
| Peak memory | 12.4 GiB |
| Avg memory | 8.4 GiB |
| CPU user | 404.4s |
| CPU sys | 21.9s |
| 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] perf: strength reduce hash partition modulo (up to 1.16x faster) [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4339015830
FYI @adriangb
--
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338787038
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338764222)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4338764222-1881-kqfh9 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 perf-strength-reduce-hash-partition (b0091c6806aeefa4475e65427e1458a98d00f800) to 65f337d (merge-base) [diff](https://github.com/apache/datafusion/compare/65f337d6637aa069012e0e9286f859fb3a743253..b0091c6806aeefa4475e65427e1458a98d00f800) using: tpch
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] perf: strength reduce hash partition modulo [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21900:
URL: https://github.com/apache/datafusion/pull/21900#issuecomment-4338773760
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21900#issuecomment-4338749432)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4338749432-1877-kzqzk 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 perf-strength-reduce-hash-partition (b0091c6806aeefa4475e65427e1458a98d00f800) to 65f337d (merge-base) [diff](https://github.com/apache/datafusion/compare/65f337d6637aa069012e0e9286f859fb3a743253..b0091c6806aeefa4475e65427e1458a98d00f800) using: tpch
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