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 2026/04/02 19:28:21 UTC
[PR] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Dandandan opened a new pull request, #21328:
URL: https://github.com/apache/datafusion/pull/21328
## Which issue does this PR close?
N/A - performance improvement
## Rationale for this change
`SortPreservingMergeExec::execute()` eagerly executes all input partitions and spawns buffered tasks immediately, even before any batch is polled from the output stream. This wastes resources when streams are created but never consumed (e.g., query cancelled before first poll), and creates an unnecessary burst of concurrent tasks.
## What changes are included in this PR?
Introduces a `LazySortPreservingMergeStream` that defers executing input partitions, spawning buffered tasks, and building the streaming merge until the first `poll_next()` call. The single-partition and zero-partition paths are unchanged.
## How are these changes tested?
All existing tests pass (17/17), covering:
- Multi-partition merge with various overlap patterns
- Drop cancellation
- Fetch limits
- Congestion handling
- Stable sort / round-robin tie breaking
- Metrics
๐ค Generated with [Claude Code](https://claude.com/claude-code)
--
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180036944
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180026900)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4180026900-671-cf8zj 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 lazy-sort-preserving-merge (6eff5503672506aa9e96627e40b577937bb32151) to 4084a18 (merge-base) [diff](https://github.com/apache/datafusion/compare/4084a18bc709d65f9901fa0d51932867ce9f6342..6eff5503672506aa9e96627e40b577937bb32151) 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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180222373
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180142002)
**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 lazy-sort-preserving-merge
--------------------
Benchmark tpcds_sf1.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโ
โ Query โ HEAD โ lazy-sort-preserving-merge โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 43.56 / 44.32 ยฑ0.69 / 45.55 ms โ 43.12 / 43.75 ยฑ0.72 / 44.97 ms โ no change โ
โ QQuery 2 โ 146.49 / 146.81 ยฑ0.29 / 147.18 ms โ 146.45 / 147.29 ยฑ0.63 / 148.06 ms โ no change โ
โ QQuery 3 โ 114.03 / 114.59 ยฑ0.49 / 115.29 ms โ 114.56 / 115.13 ยฑ0.59 / 116.27 ms โ no change โ
โ QQuery 4 โ 1342.85 / 1374.40 ยฑ18.31 / 1396.47 ms โ 1326.82 / 1347.07 ยฑ10.48 / 1356.54 ms โ no change โ
โ QQuery 5 โ 172.34 / 173.96 ยฑ0.83 / 174.59 ms โ 173.48 / 174.67 ยฑ0.77 / 175.63 ms โ no change โ
โ QQuery 6 โ 1024.44 / 1054.63 ยฑ26.54 / 1093.30 ms โ 1000.99 / 1026.86 ยฑ22.58 / 1063.44 ms โ no change โ
โ QQuery 7 โ 353.51 / 357.41 ยฑ3.34 / 362.65 ms โ 353.15 / 356.32 ยฑ3.29 / 361.81 ms โ no change โ
โ QQuery 8 โ 115.91 / 116.95 ยฑ0.76 / 118.09 ms โ 117.61 / 118.16 ยฑ0.50 / 119.04 ms โ no change โ
โ QQuery 9 โ 100.20 / 107.90 ยฑ6.85 / 117.35 ms โ 100.89 / 104.82 ยฑ3.02 / 107.90 ms โ no change โ
โ QQuery 10 โ 107.22 / 108.21 ยฑ0.59 / 109.06 ms โ 109.46 / 110.72 ยฑ0.80 / 111.83 ms โ no change โ
โ QQuery 11 โ 968.82 / 976.91 ยฑ5.39 / 985.62 ms โ 926.49 / 942.95 ยฑ9.23 / 952.01 ms โ no change โ
โ QQuery 12 โ 45.01 / 46.00 ยฑ0.63 / 46.80 ms โ 44.99 / 46.41 ยฑ1.87 / 49.99 ms โ no change โ
โ QQuery 13 โ 400.05 / 403.44 ยฑ2.51 / 406.80 ms โ 405.84 / 407.19 ยฑ1.23 / 409.14 ms โ no change โ
โ QQuery 14 โ 1011.08 / 1027.17 ยฑ8.09 / 1032.27 ms โ 1008.18 / 1021.33 ยฑ9.57 / 1037.40 ms โ no change โ
โ QQuery 15 โ 15.83 / 17.24 ยฑ1.41 / 19.17 ms โ 15.76 / 16.41 ยฑ0.35 / 16.81 ms โ no change โ
โ QQuery 16 โ 40.92 / 42.66 ยฑ1.07 / 43.85 ms โ 41.01 / 41.85 ยฑ0.67 / 43.00 ms โ no change โ
โ QQuery 17 โ 242.63 / 244.98 ยฑ1.24 / 246.05 ms โ 239.52 / 241.49 ยฑ1.32 / 243.53 ms โ no change โ
โ QQuery 18 โ 128.38 / 129.83 ยฑ0.83 / 130.82 ms โ 130.44 / 131.73 ยฑ0.72 / 132.45 ms โ no change โ
โ QQuery 19 โ 154.82 / 156.24 ยฑ1.07 / 157.17 ms โ 155.69 / 157.16 ยฑ1.26 / 158.59 ms โ no change โ
โ QQuery 20 โ 14.04 / 14.37 ยฑ0.34 / 15.00 ms โ 13.34 / 14.06 ยฑ0.57 / 15.01 ms โ no change โ
โ QQuery 21 โ 19.83 / 20.44 ยฑ0.90 / 22.23 ms โ 19.47 / 20.13 ยฑ0.45 / 20.86 ms โ no change โ
โ QQuery 22 โ 488.22 / 491.30 ยฑ2.46 / 493.55 ms โ 482.07 / 493.81 ยฑ13.60 / 519.20 ms โ no change โ
โ QQuery 23 โ 899.10 / 905.10 ยฑ5.03 / 912.46 ms โ 887.54 / 899.61 ยฑ12.99 / 923.29 ms โ no change โ
โ QQuery 24 โ 414.71 / 418.76 ยฑ3.81 / 423.55 ms โ 417.28 / 435.76 ยฑ32.59 / 500.89 ms โ no change โ
โ QQuery 25 โ 349.43 / 355.20 ยฑ3.39 / 359.05 ms โ 353.41 / 355.58 ยฑ1.98 / 358.10 ms โ no change โ
โ QQuery 26 โ 82.62 / 83.45 ยฑ0.80 / 84.76 ms โ 83.09 / 84.04 ยฑ0.71 / 85.24 ms โ no change โ
โ QQuery 27 โ 347.96 / 351.77 ยฑ2.33 / 354.85 ms โ 350.42 / 353.63 ยฑ2.23 / 356.23 ms โ no change โ
โ QQuery 28 โ 149.68 / 150.21 ยฑ0.53 / 151.02 ms โ 149.42 / 150.53 ยฑ0.70 / 151.48 ms โ no change โ
โ QQuery 29 โ 298.97 / 301.42 ยฑ2.83 / 306.56 ms โ 299.17 / 303.33 ยฑ4.01 / 308.89 ms โ no change โ
โ QQuery 30 โ 44.38 / 45.31 ยฑ1.26 / 47.64 ms โ 44.94 / 45.86 ยฑ0.82 / 47.29 ms โ no change โ
โ QQuery 31 โ 170.60 / 172.27 ยฑ1.67 / 174.38 ms โ 168.95 / 173.98 ยฑ2.89 / 177.76 ms โ no change โ
โ QQuery 32 โ 56.84 / 58.18 ยฑ0.77 / 59.14 ms โ 57.22 / 58.10 ยฑ0.68 / 58.83 ms โ no change โ
โ QQuery 33 โ 140.85 / 141.94 ยฑ1.23 / 144.24 ms โ 142.23 / 143.43 ยฑ0.93 / 145.05 ms โ no change โ
โ QQuery 34 โ 107.11 / 108.50 ยฑ1.13 / 110.24 ms โ 107.34 / 108.35 ยฑ1.29 / 110.70 ms โ no change โ
โ QQuery 35 โ 105.98 / 109.20 ยฑ1.96 / 112.19 ms โ 107.01 / 109.61 ยฑ2.69 / 114.66 ms โ no change โ
โ QQuery 36 โ 211.11 / 220.15 ยฑ5.10 / 225.30 ms โ 217.79 / 220.32 ยฑ2.91 / 224.84 ms โ no change โ
โ QQuery 37 โ 179.54 / 181.08 ยฑ1.63 / 183.92 ms โ 179.67 / 182.78 ยฑ3.21 / 188.57 ms โ no change โ
โ QQuery 38 โ 85.89 / 88.95 ยฑ2.64 / 93.27 ms โ 84.86 / 88.51 ยฑ3.93 / 95.15 ms โ no change โ
โ QQuery 39 โ 126.29 / 128.67 ยฑ1.29 / 130.14 ms โ 125.02 / 129.24 ยฑ3.27 / 134.72 ms โ no change โ
โ QQuery 40 โ 111.57 / 116.76 ยฑ4.45 / 124.80 ms โ 112.38 / 116.79 ยฑ5.19 / 126.53 ms โ no change โ
โ QQuery 41 โ 14.31 / 15.52 ยฑ1.00 / 16.95 ms โ 14.27 / 15.25 ยฑ1.02 / 17.04 ms โ no change โ
โ QQuery 42 โ 109.14 / 110.04 ยฑ0.67 / 111.18 ms โ 108.86 / 110.54 ยฑ1.18 / 112.34 ms โ no change โ
โ QQuery 43 โ 84.01 / 84.96 ยฑ0.98 / 86.59 ms โ 84.42 / 85.23 ยฑ0.60 / 86.29 ms โ no change โ
โ QQuery 44 โ 11.43 / 11.76 ยฑ0.23 / 12.05 ms โ 13.51 / 14.25 ยฑ0.83 / 15.47 ms โ 1.21x slower โ
โ QQuery 45 โ 52.92 / 53.95 ยฑ1.09 / 55.68 ms โ 53.07 / 54.29 ยฑ1.09 / 56.21 ms โ no change โ
โ QQuery 46 โ 229.23 / 236.34 ยฑ4.03 / 240.89 ms โ 234.67 / 238.73 ยฑ2.35 / 240.72 ms โ no change โ
โ QQuery 47 โ 709.07 / 720.49 ยฑ8.41 / 732.17 ms โ 712.71 / 721.99 ยฑ5.99 / 730.11 ms โ no change โ
โ QQuery 48 โ 288.25 / 295.36 ยฑ4.07 / 299.21 ms โ 287.84 / 294.65 ยฑ4.45 / 299.87 ms โ no change โ
โ QQuery 49 โ 257.61 / 258.79 ยฑ0.86 / 259.70 ms โ 255.23 / 257.43 ยฑ2.15 / 260.60 ms โ no change โ
โ QQuery 50 โ 230.35 / 239.25 ยฑ6.16 / 247.33 ms โ 235.37 / 237.67 ยฑ1.59 / 240.28 ms โ no change โ
โ QQuery 51 โ 180.72 / 185.66 ยฑ2.97 / 190.06 ms โ 184.70 / 188.34 ยฑ1.95 / 190.34 ms โ no change โ
โ QQuery 52 โ 109.39 / 110.18 ยฑ0.69 / 111.05 ms โ 108.31 / 111.25 ยฑ1.81 / 113.16 ms โ no change โ
โ QQuery 53 โ 104.06 / 104.72 ยฑ0.57 / 105.55 ms โ 103.25 / 104.48 ยฑ1.14 / 106.40 ms โ no change โ
โ QQuery 54 โ 149.52 / 152.16 ยฑ2.15 / 155.30 ms โ 147.47 / 149.39 ยฑ1.35 / 151.33 ms โ no change โ
โ QQuery 55 โ 108.14 / 109.77 ยฑ1.04 / 111.12 ms โ 107.45 / 110.13 ยฑ2.96 / 115.20 ms โ no change โ
โ QQuery 56 โ 143.04 / 143.83 ยฑ0.88 / 145.14 ms โ 140.89 / 144.07 ยฑ1.98 / 147.09 ms โ no change โ
โ QQuery 57 โ 176.14 / 178.52 ยฑ2.08 / 181.84 ms โ 173.20 / 175.32 ยฑ1.14 / 176.60 ms โ no change โ
โ QQuery 58 โ 289.36 / 298.02 ยฑ6.31 / 304.76 ms โ 294.24 / 301.66 ยฑ4.79 / 307.53 ms โ no change โ
โ QQuery 59 โ 197.16 / 199.68 ยฑ1.96 / 201.62 ms โ 199.73 / 200.96 ยฑ1.31 / 203.46 ms โ no change โ
โ QQuery 60 โ 143.54 / 145.67 ยฑ2.07 / 149.18 ms โ 143.38 / 145.77 ยฑ1.48 / 147.25 ms โ no change โ
โ QQuery 61 โ 170.31 / 172.40 ยฑ1.30 / 173.96 ms โ 173.01 / 174.30 ยฑ0.89 / 175.70 ms โ no change โ
โ QQuery 62 โ 894.19 / 921.78 ยฑ22.66 / 954.93 ms โ 894.84 / 933.72 ยฑ26.49 / 969.86 ms โ no change โ
โ QQuery 63 โ 102.92 / 105.72 ยฑ1.85 / 108.57 ms โ 103.85 / 107.53 ยฑ2.90 / 112.75 ms โ no change โ
โ QQuery 64 โ 696.89 / 700.50 ยฑ2.78 / 704.27 ms โ 698.94 / 703.20 ยฑ2.50 / 705.73 ms โ no change โ
โ QQuery 65 โ 251.03 / 257.44 ยฑ4.18 / 261.84 ms โ 250.09 / 254.73 ยฑ4.69 / 263.54 ms โ no change โ
โ QQuery 66 โ 232.20 / 249.57 ยฑ10.22 / 261.93 ms โ 242.47 / 254.11 ยฑ8.22 / 265.61 ms โ no change โ
โ QQuery 67 โ 311.30 / 315.01 ยฑ4.90 / 324.59 ms โ 312.99 / 317.13 ยฑ4.44 / 325.48 ms โ no change โ
โ QQuery 68 โ 280.54 / 283.45 ยฑ1.76 / 286.03 ms โ 279.24 / 280.82 ยฑ0.90 / 281.62 ms โ no change โ
โ QQuery 69 โ 104.04 / 104.91 ยฑ0.76 / 106.00 ms โ 104.91 / 105.96 ยฑ0.92 / 107.35 ms โ no change โ
โ QQuery 70 โ 345.57 / 355.37 ยฑ7.16 / 363.81 ms โ 338.87 / 345.49 ยฑ7.79 / 360.41 ms โ no change โ
โ QQuery 71 โ 134.62 / 135.80 ยฑ1.23 / 137.98 ms โ 134.73 / 137.83 ยฑ2.96 / 142.98 ms โ no change โ
โ QQuery 72 โ 715.22 / 722.92 ยฑ6.36 / 732.08 ms โ 713.49 / 728.48 ยฑ12.77 / 745.27 ms โ no change โ
โ QQuery 73 โ 102.93 / 105.24 ยฑ1.93 / 108.40 ms โ 103.69 / 106.40 ยฑ2.51 / 111.12 ms โ no change โ
โ QQuery 74 โ 584.91 / 588.17 ยฑ2.77 / 592.24 ms โ 593.56 / 624.85 ยฑ22.06 / 661.16 ms โ 1.06x slower โ
โ QQuery 75 โ 274.68 / 277.16 ยฑ2.24 / 280.03 ms โ 279.42 / 281.52 ยฑ1.65 / 284.33 ms โ no change โ
โ QQuery 76 โ 132.16 / 134.83 ยฑ1.80 / 136.89 ms โ 133.48 / 135.74 ยฑ1.65 / 138.09 ms โ no change โ
โ QQuery 77 โ 187.88 / 189.28 ยฑ1.35 / 191.60 ms โ 188.69 / 191.09 ยฑ2.28 / 194.53 ms โ no change โ
โ QQuery 78 โ 349.51 / 357.69 ยฑ5.38 / 365.15 ms โ 355.84 / 362.25 ยฑ5.28 / 370.21 ms โ no change โ
โ QQuery 79 โ 234.54 / 237.01 ยฑ1.48 / 239.09 ms โ 237.84 / 245.15 ยฑ6.69 / 253.85 ms โ no change โ
โ QQuery 80 โ 326.75 / 328.87 ยฑ1.34 / 330.67 ms โ 330.64 / 333.00 ยฑ1.98 / 335.63 ms โ no change โ
โ QQuery 81 โ 26.13 / 27.77 ยฑ0.97 / 29.07 ms โ 26.44 / 27.43 ยฑ0.83 / 28.76 ms โ no change โ
โ QQuery 82 โ 199.96 / 201.30 ยฑ0.86 / 202.42 ms โ 202.77 / 205.21 ยฑ2.11 / 208.50 ms โ no change โ
โ QQuery 83 โ 39.40 / 40.66 ยฑ1.36 / 43.27 ms โ 39.47 / 40.63 ยฑ0.73 / 41.54 ms โ no change โ
โ QQuery 84 โ 48.66 / 49.62 ยฑ0.70 / 50.45 ms โ 48.64 / 50.65 ยฑ1.27 / 52.34 ms โ no change โ
โ QQuery 85 โ 147.77 / 149.84 ยฑ1.64 / 152.80 ms โ 152.40 / 154.33 ยฑ1.90 / 156.72 ms โ no change โ
โ QQuery 86 โ 39.50 / 40.37 ยฑ1.00 / 42.24 ms โ 39.95 / 40.73 ยฑ0.90 / 42.43 ms โ no change โ
โ QQuery 87 โ 87.65 / 90.64 ยฑ2.74 / 95.10 ms โ 86.38 / 90.05 ยฑ3.68 / 95.92 ms โ no change โ
โ QQuery 88 โ 99.00 / 99.96 ยฑ0.53 / 100.54 ms โ 99.59 / 100.31 ยฑ0.81 / 101.88 ms โ no change โ
โ QQuery 89 โ 118.35 / 119.33 ยฑ0.81 / 120.54 ms โ 120.24 / 121.01 ยฑ0.55 / 121.72 ms โ no change โ
โ QQuery 90 โ 23.50 / 23.86 ยฑ0.30 / 24.26 ms โ 24.29 / 24.87 ยฑ0.51 / 25.75 ms โ no change โ
โ QQuery 91 โ 63.30 / 63.82 ยฑ0.36 / 64.26 ms โ 63.36 / 65.44 ยฑ1.37 / 67.17 ms โ no change โ
โ QQuery 92 โ 57.76 / 58.21 ยฑ0.53 / 59.25 ms โ 58.15 / 59.60 ยฑ1.15 / 61.51 ms โ no change โ
โ QQuery 93 โ 189.73 / 192.11 ยฑ2.05 / 195.21 ms โ 194.82 / 195.94 ยฑ1.15 / 198.01 ms โ no change โ
โ QQuery 94 โ 61.72 / 62.26 ยฑ0.33 / 62.68 ms โ 62.35 / 63.48 ยฑ1.03 / 65.30 ms โ no change โ
โ QQuery 95 โ 133.02 / 135.20 ยฑ1.52 / 136.97 ms โ 134.42 / 137.68 ยฑ2.08 / 140.59 ms โ no change โ
โ QQuery 96 โ 71.54 / 73.46 ยฑ0.96 / 74.08 ms โ 73.85 / 74.75 ยฑ0.64 / 75.72 ms โ no change โ
โ QQuery 97 โ 128.89 / 130.65 ยฑ1.89 / 134.30 ms โ 132.86 / 134.63 ยฑ1.30 / 136.15 ms โ no change โ
โ QQuery 98 โ 154.30 / 155.88 ยฑ1.69 / 158.14 ms โ 154.65 / 157.48 ยฑ1.76 / 160.18 ms โ no change โ
โ QQuery 99 โ 10745.75 / 10793.86 ยฑ45.21 / 10878.95 ms โ 10783.20 / 10831.27 ยฑ43.10 / 10889.72 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 33813.26ms โ
โ Total Time (lazy-sort-preserving-merge) โ 33898.99ms โ
โ Average Time (HEAD) โ 341.55ms โ
โ Average Time (lazy-sort-preserving-merge) โ 342.41ms โ
โ Queries Faster โ 0 โ
โ Queries Slower โ 2 โ
โ Queries with No Change โ 97 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpcds โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 169.4s |
| Peak memory | 5.6 GiB |
| Avg memory | 4.5 GiB |
| CPU user | 272.0s |
| CPU sys | 17.6s |
| Disk read | 0 B |
| Disk write | 703.0 MiB |
**tpcds โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 169.8s |
| Peak memory | 5.5 GiB |
| Avg memory | 4.6 GiB |
| CPU user | 272.1s |
| CPU sys | 18.1s |
| Disk read | 0 B |
| Disk write | 148.0 KiB |
</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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on code in PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#discussion_r3035457774
##########
datafusion/physical-plan/src/sorts/sort_preserving_merge.rs:
##########
@@ -433,6 +422,99 @@ impl ExecutionPlan for SortPreservingMergeExec {
}
}
+/// A stream that lazily spawns input partition tasks and builds the streaming
+/// merge on first poll, rather than eagerly in `execute()`.
+struct SortPreservingMergeExecStream {
+ schema: SchemaRef,
+ input: Arc<dyn ExecutionPlan>,
+ context: Arc<TaskContext>,
+ expr: LexOrdering,
+ metrics: BaselineMetrics,
+ batch_size: usize,
+ fetch: Option<usize>,
+ reservation: datafusion_execution::memory_pool::MemoryReservation,
+ enable_round_robin_repartition: bool,
+ state: SPMStreamState,
+}
+
+enum SPMStreamState {
+ /// Tasks have not been spawned yet.
+ Pending,
+ /// The streaming merge has been built and is running.
+ Running(SendableRecordBatchStream),
+ /// Initialization failed.
+ Failed,
+}
+
+impl SortPreservingMergeExecStream {
+ fn start(&mut self) -> Result<&mut SendableRecordBatchStream> {
+ let input_partitions = self.input.output_partitioning().partition_count();
+
+ let receivers = (0..input_partitions)
+ .map(|partition| {
+ let stream = self.input.execute(partition, Arc::clone(&self.context))?;
+ Ok(spawn_buffered(stream, 1))
+ })
+ .collect::<Result<_>>()?;
+
+ debug!("Done setting up sender-receiver for SortPreservingMergeExec::execute");
+
+ // Take reservation out of self via mem::replace to pass ownership
+ let reservation = std::mem::replace(
+ &mut self.reservation,
+ MemoryConsumer::new("empty")
+ .register(&self.context.runtime_env().memory_pool),
+ );
+
+ let result = StreamingMergeBuilder::new()
+ .with_streams(receivers)
+ .with_schema(Arc::clone(&self.schema))
+ .with_expressions(&self.expr)
+ .with_metrics(self.metrics.clone())
+ .with_batch_size(self.batch_size)
+ .with_fetch(self.fetch)
+ .with_reservation(reservation)
+ .with_round_robin_tie_breaker(self.enable_round_robin_repartition)
+ .build()?;
+
+ debug!("Got stream result from SortPreservingMergeStream::new_from_receivers");
+
+ self.state = SPMStreamState::Running(result);
+ match &mut self.state {
+ SPMStreamState::Running(s) => Ok(s),
+ _ => unreachable!(),
+ }
+ }
+}
+
+impl RecordBatchStream for SortPreservingMergeExecStream {
+ fn schema(&self) -> SchemaRef {
+ Arc::clone(&self.schema)
+ }
+}
+
+impl Stream for SortPreservingMergeExecStream {
+ type Item = Result<arrow::array::RecordBatch>;
+
+ fn poll_next(
+ mut self: Pin<&mut Self>,
+ cx: &mut Context<'_>,
+ ) -> Poll<Option<Self::Item>> {
+ let stream = match &mut self.state {
+ SPMStreamState::Running(s) => s,
+ SPMStreamState::Failed => return Poll::Ready(None),
+ SPMStreamState::Pending => match self.start() {
+ Ok(s) => s,
+ Err(e) => {
+ self.state = SPMStreamState::Failed;
+ return Poll::Ready(Some(Err(e)));
+ }
+ },
+ };
+ stream.poll_next_unpin(cx)
+ }
+}
+
#[cfg(test)]
Review Comment:
yeah
--
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180029246
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180026900)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4180026900-670-wmbtx 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 lazy-sort-preserving-merge (6eff5503672506aa9e96627e40b577937bb32151) to 4084a18 (merge-base) [diff](https://github.com/apache/datafusion/compare/4084a18bc709d65f9901fa0d51932867ce9f6342..6eff5503672506aa9e96627e40b577937bb32151) 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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180142002
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180193758
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180142002)
**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 lazy-sort-preserving-merge
--------------------
Benchmark tpch_sf1.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโ
โ Query โ HEAD โ lazy-sort-preserving-merge โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 45.15 / 45.71 ยฑ0.78 / 47.19 ms โ 45.21 / 45.72 ยฑ0.67 / 47.03 ms โ no change โ
โ QQuery 2 โ 20.70 / 21.07 ยฑ0.27 / 21.40 ms โ 20.76 / 20.93 ยฑ0.26 / 21.44 ms โ no change โ
โ QQuery 3 โ 31.46 / 31.69 ยฑ0.20 / 31.95 ms โ 31.45 / 33.08 ยฑ2.30 / 37.61 ms โ no change โ
โ QQuery 4 โ 20.48 / 20.99 ยฑ0.51 / 21.95 ms โ 20.81 / 22.16 ยฑ1.24 / 23.90 ms โ 1.06x slower โ
โ QQuery 5 โ 47.23 / 48.26 ยฑ0.94 / 49.87 ms โ 47.42 / 49.55 ยฑ1.69 / 51.08 ms โ no change โ
โ QQuery 6 โ 17.04 / 17.19 ยฑ0.17 / 17.52 ms โ 17.06 / 18.02 ยฑ1.13 / 20.05 ms โ no change โ
โ QQuery 7 โ 54.27 / 55.03 ยฑ0.84 / 56.60 ms โ 53.77 / 54.61 ยฑ0.76 / 55.59 ms โ no change โ
โ QQuery 8 โ 47.66 / 48.22 ยฑ0.39 / 48.80 ms โ 47.51 / 47.93 ยฑ0.27 / 48.33 ms โ no change โ
โ QQuery 9 โ 55.29 / 55.70 ยฑ0.25 / 56.08 ms โ 53.37 / 54.35 ยฑ0.56 / 55.03 ms โ no change โ
โ QQuery 10 โ 70.33 / 71.83 ยฑ1.31 / 74.24 ms โ 70.30 / 71.01 ยฑ0.52 / 71.90 ms โ no change โ
โ QQuery 11 โ 13.61 / 13.92 ยฑ0.27 / 14.40 ms โ 13.74 / 13.89 ยฑ0.15 / 14.13 ms โ no change โ
โ QQuery 12 โ 27.79 / 28.57 ยฑ0.81 / 29.90 ms โ 27.68 / 27.95 ยฑ0.26 / 28.28 ms โ no change โ
โ QQuery 13 โ 38.05 / 39.58 ยฑ1.32 / 41.79 ms โ 37.72 / 38.82 ยฑ0.77 / 39.68 ms โ no change โ
โ QQuery 14 โ 28.18 / 28.42 ยฑ0.21 / 28.76 ms โ 28.25 / 28.81 ยฑ0.59 / 29.95 ms โ no change โ
โ QQuery 15 โ 33.47 / 33.98 ยฑ0.79 / 35.54 ms โ 33.38 / 33.62 ยฑ0.18 / 33.84 ms โ no change โ
โ QQuery 16 โ 15.82 / 16.20 ยฑ0.35 / 16.74 ms โ 15.59 / 16.18 ยฑ0.39 / 16.73 ms โ no change โ
โ QQuery 17 โ 71.79 / 72.31 ยฑ0.30 / 72.59 ms โ 72.14 / 73.12 ยฑ0.58 / 73.76 ms โ no change โ
โ QQuery 18 โ 76.41 / 77.63 ยฑ1.52 / 80.53 ms โ 76.90 / 78.12 ยฑ0.80 / 79.27 ms โ no change โ
โ QQuery 19 โ 37.01 / 37.72 ยฑ0.89 / 39.42 ms โ 37.10 / 37.44 ยฑ0.34 / 38.06 ms โ no change โ
โ QQuery 20 โ 39.25 / 40.71 ยฑ1.15 / 42.33 ms โ 39.82 / 40.48 ยฑ0.51 / 41.00 ms โ no change โ
โ QQuery 21 โ 62.19 / 63.92 ยฑ1.54 / 66.72 ms โ 62.15 / 63.53 ยฑ1.60 / 66.48 ms โ no change โ
โ QQuery 22 โ 17.72 / 18.09 ยฑ0.44 / 18.95 ms โ 17.20 / 17.83 ยฑ0.67 / 19.06 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 886.73ms โ
โ Total Time (lazy-sort-preserving-merge) โ 887.16ms โ
โ Average Time (HEAD) โ 40.31ms โ
โ Average Time (lazy-sort-preserving-merge) โ 40.33ms โ
โ Queries Faster โ 0 โ
โ Queries Slower โ 1 โ
โ Queries with No Change โ 21 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpch โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 4.7s |
| Peak memory | 4.0 GiB |
| Avg memory | 3.6 GiB |
| CPU user | 33.0s |
| CPU sys | 2.8s |
| Disk read | 0 B |
| Disk write | 136.0 KiB |
**tpch โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 4.7s |
| Peak memory | 4.0 GiB |
| Avg memory | 3.6 GiB |
| CPU user | 33.1s |
| CPU sys | 2.8s |
| Disk read | 0 B |
| Disk write | 60.0 KiB |
</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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#discussion_r3033479310
##########
datafusion/physical-plan/src/sorts/sort_preserving_merge.rs:
##########
@@ -433,6 +422,99 @@ impl ExecutionPlan for SortPreservingMergeExec {
}
}
+/// A stream that lazily spawns input partition tasks and builds the streaming
+/// merge on first poll, rather than eagerly in `execute()`.
+struct SortPreservingMergeExecStream {
+ schema: SchemaRef,
+ input: Arc<dyn ExecutionPlan>,
+ context: Arc<TaskContext>,
+ expr: LexOrdering,
+ metrics: BaselineMetrics,
+ batch_size: usize,
+ fetch: Option<usize>,
+ reservation: datafusion_execution::memory_pool::MemoryReservation,
+ enable_round_robin_repartition: bool,
+ state: SPMStreamState,
+}
+
+enum SPMStreamState {
+ /// Tasks have not been spawned yet.
+ Pending,
+ /// The streaming merge has been built and is running.
+ Running(SendableRecordBatchStream),
+ /// Initialization failed.
+ Failed,
+}
+
+impl SortPreservingMergeExecStream {
+ fn start(&mut self) -> Result<&mut SendableRecordBatchStream> {
+ let input_partitions = self.input.output_partitioning().partition_count();
+
+ let receivers = (0..input_partitions)
+ .map(|partition| {
+ let stream = self.input.execute(partition, Arc::clone(&self.context))?;
+ Ok(spawn_buffered(stream, 1))
+ })
+ .collect::<Result<_>>()?;
+
+ debug!("Done setting up sender-receiver for SortPreservingMergeExec::execute");
+
+ // Take reservation out of self via mem::replace to pass ownership
+ let reservation = std::mem::replace(
+ &mut self.reservation,
+ MemoryConsumer::new("empty")
+ .register(&self.context.runtime_env().memory_pool),
+ );
+
+ let result = StreamingMergeBuilder::new()
+ .with_streams(receivers)
+ .with_schema(Arc::clone(&self.schema))
+ .with_expressions(&self.expr)
+ .with_metrics(self.metrics.clone())
+ .with_batch_size(self.batch_size)
+ .with_fetch(self.fetch)
+ .with_reservation(reservation)
+ .with_round_robin_tie_breaker(self.enable_round_robin_repartition)
+ .build()?;
+
+ debug!("Got stream result from SortPreservingMergeStream::new_from_receivers");
+
+ self.state = SPMStreamState::Running(result);
+ match &mut self.state {
+ SPMStreamState::Running(s) => Ok(s),
+ _ => unreachable!(),
+ }
+ }
+}
+
+impl RecordBatchStream for SortPreservingMergeExecStream {
+ fn schema(&self) -> SchemaRef {
+ Arc::clone(&self.schema)
+ }
+}
+
+impl Stream for SortPreservingMergeExecStream {
+ type Item = Result<arrow::array::RecordBatch>;
+
+ fn poll_next(
+ mut self: Pin<&mut Self>,
+ cx: &mut Context<'_>,
+ ) -> Poll<Option<Self::Item>> {
+ let stream = match &mut self.state {
+ SPMStreamState::Running(s) => s,
+ SPMStreamState::Failed => return Poll::Ready(None),
+ SPMStreamState::Pending => match self.start() {
+ Ok(s) => s,
+ Err(e) => {
+ self.state = SPMStreamState::Failed;
+ return Poll::Ready(Some(Err(e)));
+ }
+ },
+ };
+ stream.poll_next_unpin(cx)
+ }
+}
+
#[cfg(test)]
Review Comment:
I think it would be good to add a test to this PR to avoid regressions. Specifically, one that calls `execute(0, ...)` without polling and proves no child stream/task has started yet
--
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180111731
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180026900)
**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 lazy-sort-preserving-merge
--------------------
Benchmark tpcds_sf1.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ lazy-sort-preserving-merge โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 43.80 / 44.54 ยฑ0.77 / 45.90 ms โ 43.30 / 44.04 ยฑ0.73 / 45.35 ms โ no change โ
โ QQuery 2 โ 145.43 / 147.10 ยฑ0.92 / 148.05 ms โ 145.34 / 145.94 ยฑ0.56 / 146.84 ms โ no change โ
โ QQuery 3 โ 114.40 / 114.94 ยฑ0.56 / 115.80 ms โ 114.85 / 115.84 ยฑ1.17 / 118.09 ms โ no change โ
โ QQuery 4 โ 1406.12 / 1423.75 ยฑ14.25 / 1445.73 ms โ 1359.98 / 1396.12 ยฑ21.59 / 1421.76 ms โ no change โ
โ QQuery 5 โ 172.36 / 174.33 ยฑ1.58 / 176.85 ms โ 173.75 / 175.10 ยฑ1.17 / 177.19 ms โ no change โ
โ QQuery 6 โ 1001.85 / 1045.05 ยฑ24.70 / 1070.16 ms โ 1036.19 / 1049.60 ยฑ13.95 / 1075.63 ms โ no change โ
โ QQuery 7 โ 354.77 / 356.64 ยฑ1.70 / 359.35 ms โ 356.08 / 361.25 ยฑ6.25 / 373.57 ms โ no change โ
โ QQuery 8 โ 114.77 / 117.62 ยฑ1.50 / 119.11 ms โ 115.44 / 117.19 ยฑ1.17 / 119.02 ms โ no change โ
โ QQuery 9 โ 101.31 / 105.32 ยฑ3.01 / 108.91 ms โ 101.29 / 105.15 ยฑ2.84 / 108.66 ms โ no change โ
โ QQuery 10 โ 107.99 / 108.89 ยฑ0.74 / 110.13 ms โ 108.34 / 109.64 ยฑ0.71 / 110.33 ms โ no change โ
โ QQuery 11 โ 971.48 / 978.68 ยฑ5.84 / 985.21 ms โ 977.52 / 990.67 ยฑ11.72 / 1009.66 ms โ no change โ
โ QQuery 12 โ 45.45 / 46.92 ยฑ1.10 / 48.26 ms โ 47.07 / 48.05 ยฑ0.82 / 49.13 ms โ no change โ
โ QQuery 13 โ 400.81 / 403.26 ยฑ1.68 / 405.32 ms โ 405.73 / 409.09 ยฑ2.09 / 411.79 ms โ no change โ
โ QQuery 14 โ 1005.54 / 1020.34 ยฑ7.94 / 1028.69 ms โ 1022.64 / 1033.90 ยฑ7.82 / 1044.45 ms โ no change โ
โ QQuery 15 โ 15.63 / 17.14 ยฑ1.12 / 19.06 ms โ 15.76 / 17.16 ยฑ0.92 / 18.43 ms โ no change โ
โ QQuery 16 โ 40.94 / 41.78 ยฑ0.63 / 42.77 ms โ 41.13 / 42.29 ยฑ1.02 / 43.78 ms โ no change โ
โ QQuery 17 โ 242.16 / 243.17 ยฑ1.14 / 245.24 ms โ 241.00 / 243.96 ยฑ2.43 / 247.42 ms โ no change โ
โ QQuery 18 โ 129.70 / 130.98 ยฑ0.80 / 132.09 ms โ 131.23 / 132.38 ยฑ1.11 / 133.83 ms โ no change โ
โ QQuery 19 โ 155.95 / 157.91 ยฑ1.30 / 159.29 ms โ 155.84 / 158.28 ยฑ1.39 / 160.12 ms โ no change โ
โ QQuery 20 โ 13.51 / 14.37 ยฑ0.66 / 15.42 ms โ 13.93 / 14.76 ยฑ0.86 / 16.41 ms โ no change โ
โ QQuery 21 โ 19.85 / 20.30 ยฑ0.42 / 20.94 ms โ 19.99 / 20.58 ยฑ0.33 / 20.89 ms โ no change โ
โ QQuery 22 โ 484.54 / 492.08 ยฑ6.13 / 503.30 ms โ 490.14 / 496.14 ยฑ5.48 / 503.35 ms โ no change โ
โ QQuery 23 โ 894.16 / 900.57 ยฑ4.98 / 908.26 ms โ 906.89 / 918.68 ยฑ7.82 / 930.18 ms โ no change โ
โ QQuery 24 โ 416.94 / 419.75 ยฑ2.86 / 425.09 ms โ 421.12 / 423.69 ยฑ1.76 / 426.47 ms โ no change โ
โ QQuery 25 โ 356.74 / 358.48 ยฑ1.22 / 360.44 ms โ 355.88 / 358.22 ยฑ1.58 / 359.99 ms โ no change โ
โ QQuery 26 โ 80.29 / 82.85 ยฑ2.19 / 85.57 ms โ 83.59 / 85.12 ยฑ1.12 / 86.46 ms โ no change โ
โ QQuery 27 โ 350.37 / 351.99 ยฑ1.36 / 353.83 ms โ 351.67 / 354.58 ยฑ2.31 / 358.49 ms โ no change โ
โ QQuery 28 โ 149.05 / 150.47 ยฑ1.47 / 152.81 ms โ 151.02 / 152.36 ยฑ0.92 / 153.64 ms โ no change โ
โ QQuery 29 โ 297.64 / 300.32 ยฑ2.03 / 303.30 ms โ 300.53 / 303.73 ยฑ2.20 / 307.10 ms โ no change โ
โ QQuery 30 โ 43.24 / 44.52 ยฑ1.14 / 46.56 ms โ 43.34 / 44.42 ยฑ0.61 / 45.24 ms โ no change โ
โ QQuery 31 โ 171.29 / 172.57 ยฑ1.22 / 174.63 ms โ 167.75 / 171.19 ยฑ2.20 / 174.18 ms โ no change โ
โ QQuery 32 โ 57.49 / 58.62 ยฑ1.23 / 60.88 ms โ 57.48 / 58.62 ยฑ1.27 / 60.77 ms โ no change โ
โ QQuery 33 โ 142.86 / 144.07 ยฑ1.51 / 146.85 ms โ 141.35 / 142.47 ยฑ0.59 / 143.01 ms โ no change โ
โ QQuery 34 โ 106.49 / 107.79 ยฑ0.88 / 109.17 ms โ 107.49 / 108.68 ยฑ1.35 / 111.14 ms โ no change โ
โ QQuery 35 โ 109.05 / 111.17 ยฑ1.38 / 113.32 ms โ 108.07 / 110.33 ยฑ1.66 / 112.54 ms โ no change โ
โ QQuery 36 โ 212.58 / 220.59 ยฑ4.24 / 225.19 ms โ 212.77 / 219.68 ยฑ5.18 / 224.77 ms โ no change โ
โ QQuery 37 โ 177.52 / 180.34 ยฑ2.98 / 185.48 ms โ 177.45 / 180.71 ยฑ2.82 / 185.34 ms โ no change โ
โ QQuery 38 โ 84.48 / 87.74 ยฑ3.26 / 93.32 ms โ 85.90 / 87.95 ยฑ2.47 / 92.24 ms โ no change โ
โ QQuery 39 โ 125.11 / 128.17 ยฑ2.88 / 133.01 ms โ 123.86 / 126.63 ยฑ1.85 / 129.32 ms โ no change โ
โ QQuery 40 โ 113.52 / 119.91 ยฑ6.77 / 132.98 ms โ 111.77 / 117.93 ยฑ5.97 / 129.15 ms โ no change โ
โ QQuery 41 โ 15.13 / 15.70 ยฑ0.66 / 16.91 ms โ 14.21 / 15.10 ยฑ0.87 / 16.42 ms โ no change โ
โ QQuery 42 โ 107.51 / 108.93 ยฑ1.31 / 111.10 ms โ 108.07 / 108.98 ยฑ1.04 / 110.94 ms โ no change โ
โ QQuery 43 โ 83.97 / 85.61 ยฑ1.11 / 87.47 ms โ 84.22 / 84.91 ยฑ0.56 / 85.77 ms โ no change โ
โ QQuery 44 โ 11.36 / 12.45 ยฑ1.53 / 15.48 ms โ 13.59 / 14.12 ยฑ0.46 / 14.91 ms โ 1.13x slower โ
โ QQuery 45 โ 53.43 / 55.27 ยฑ1.45 / 57.40 ms โ 52.28 / 53.81 ยฑ1.36 / 55.93 ms โ no change โ
โ QQuery 46 โ 236.18 / 238.95 ยฑ1.80 / 240.98 ms โ 231.31 / 238.14 ยฑ5.91 / 248.87 ms โ no change โ
โ QQuery 47 โ 743.91 / 756.94 ยฑ10.17 / 773.80 ms โ 772.41 / 777.60 ยฑ3.50 / 782.86 ms โ no change โ
โ QQuery 48 โ 291.43 / 293.79 ยฑ1.96 / 296.49 ms โ 290.46 / 302.24 ยฑ6.45 / 309.59 ms โ no change โ
โ QQuery 49 โ 253.43 / 255.27 ยฑ1.24 / 257.34 ms โ 254.34 / 257.09 ยฑ2.13 / 260.71 ms โ no change โ
โ QQuery 50 โ 230.72 / 236.61 ยฑ5.46 / 245.73 ms โ 229.02 / 236.46 ยฑ4.99 / 244.75 ms โ no change โ
โ QQuery 51 โ 182.68 / 187.00 ยฑ2.57 / 190.48 ms โ 184.06 / 186.47 ยฑ1.45 / 188.03 ms โ no change โ
โ QQuery 52 โ 109.28 / 110.15 ยฑ0.59 / 110.96 ms โ 108.01 / 109.00 ยฑ1.32 / 111.61 ms โ no change โ
โ QQuery 53 โ 103.70 / 104.50 ยฑ0.43 / 104.87 ms โ 103.28 / 104.53 ยฑ0.83 / 105.83 ms โ no change โ
โ QQuery 54 โ 148.34 / 149.56 ยฑ0.90 / 151.13 ms โ 148.26 / 150.42 ยฑ1.88 / 153.92 ms โ no change โ
โ QQuery 55 โ 107.86 / 109.14 ยฑ1.15 / 110.69 ms โ 108.27 / 109.80 ยฑ0.91 / 110.98 ms โ no change โ
โ QQuery 56 โ 142.07 / 142.59 ยฑ0.66 / 143.89 ms โ 143.05 / 144.48 ยฑ1.30 / 146.28 ms โ no change โ
โ QQuery 57 โ 174.00 / 176.55 ยฑ1.76 / 179.11 ms โ 175.65 / 177.09 ยฑ1.36 / 179.50 ms โ no change โ
โ QQuery 58 โ 299.52 / 303.69 ยฑ2.47 / 306.97 ms โ 295.49 / 318.38 ยฑ15.51 / 339.42 ms โ no change โ
โ QQuery 59 โ 198.44 / 201.47 ยฑ2.48 / 204.97 ms โ 198.28 / 201.62 ยฑ1.83 / 203.81 ms โ no change โ
โ QQuery 60 โ 142.57 / 145.48 ยฑ1.61 / 147.50 ms โ 143.90 / 146.02 ยฑ1.94 / 149.03 ms โ no change โ
โ QQuery 61 โ 172.80 / 174.38 ยฑ1.17 / 176.18 ms โ 171.46 / 174.37 ยฑ1.93 / 176.81 ms โ no change โ
โ QQuery 62 โ 913.45 / 923.46 ยฑ6.20 / 931.00 ms โ 926.05 / 947.25 ยฑ28.35 / 1002.39 ms โ no change โ
โ QQuery 63 โ 105.20 / 106.92 ยฑ1.74 / 109.62 ms โ 104.81 / 106.65 ยฑ1.40 / 108.55 ms โ no change โ
โ QQuery 64 โ 706.62 / 711.87 ยฑ3.78 / 716.66 ms โ 704.93 / 711.09 ยฑ4.92 / 718.08 ms โ no change โ
โ QQuery 65 โ 261.02 / 263.21 ยฑ1.52 / 265.28 ms โ 254.54 / 260.72 ยฑ4.60 / 268.64 ms โ no change โ
โ QQuery 66 โ 248.14 / 259.56 ยฑ5.77 / 263.38 ms โ 251.54 / 265.50 ยฑ9.44 / 276.98 ms โ no change โ
โ QQuery 67 โ 314.65 / 321.55 ยฑ6.75 / 333.84 ms โ 310.09 / 317.86 ยฑ5.79 / 325.36 ms โ no change โ
โ QQuery 68 โ 284.82 / 289.33 ยฑ3.70 / 294.81 ms โ 282.92 / 287.53 ยฑ3.83 / 292.25 ms โ no change โ
โ QQuery 69 โ 105.33 / 106.46 ยฑ0.83 / 107.55 ms โ 103.94 / 106.36 ยฑ1.59 / 108.60 ms โ no change โ
โ QQuery 70 โ 342.96 / 356.60 ยฑ13.13 / 379.01 ms โ 338.60 / 346.62 ยฑ10.86 / 368.04 ms โ no change โ
โ QQuery 71 โ 138.85 / 139.75 ยฑ0.73 / 141.07 ms โ 136.65 / 138.32 ยฑ1.63 / 140.33 ms โ no change โ
โ QQuery 72 โ 738.58 / 750.82 ยฑ9.99 / 766.63 ms โ 703.10 / 724.11 ยฑ12.14 / 736.33 ms โ no change โ
โ QQuery 73 โ 105.02 / 107.78 ยฑ2.56 / 111.86 ms โ 102.07 / 104.26 ยฑ1.25 / 105.93 ms โ no change โ
โ QQuery 74 โ 616.80 / 634.20 ยฑ9.61 / 645.04 ms โ 608.66 / 615.92 ยฑ7.01 / 627.80 ms โ no change โ
โ QQuery 75 โ 279.77 / 281.00 ยฑ1.70 / 284.29 ms โ 278.94 / 280.58 ยฑ1.48 / 282.62 ms โ no change โ
โ QQuery 76 โ 132.89 / 135.11 ยฑ1.16 / 136.13 ms โ 133.63 / 134.92 ยฑ1.55 / 137.94 ms โ no change โ
โ QQuery 77 โ 189.38 / 190.85 ยฑ1.13 / 192.56 ms โ 187.46 / 190.51 ยฑ2.10 / 193.58 ms โ no change โ
โ QQuery 78 โ 359.37 / 363.36 ยฑ2.98 / 367.15 ms โ 351.47 / 356.17 ยฑ3.56 / 361.16 ms โ no change โ
โ QQuery 79 โ 248.33 / 250.30 ยฑ1.33 / 252.28 ms โ 234.21 / 236.80 ยฑ2.19 / 240.36 ms โ +1.06x faster โ
โ QQuery 80 โ 331.41 / 334.29 ยฑ2.46 / 337.83 ms โ 327.77 / 328.89 ยฑ1.04 / 330.67 ms โ no change โ
โ QQuery 81 โ 27.04 / 27.77 ยฑ0.50 / 28.46 ms โ 27.64 / 28.33 ยฑ0.46 / 28.85 ms โ no change โ
โ QQuery 82 โ 201.88 / 206.36 ยฑ3.33 / 210.30 ms โ 200.68 / 204.76 ยฑ3.59 / 210.84 ms โ no change โ
โ QQuery 83 โ 40.28 / 41.27 ยฑ0.74 / 42.45 ms โ 39.96 / 41.66 ยฑ0.94 / 42.81 ms โ no change โ
โ QQuery 84 โ 48.80 / 49.23 ยฑ0.34 / 49.67 ms โ 50.08 / 51.24 ยฑ1.01 / 52.45 ms โ no change โ
โ QQuery 85 โ 151.02 / 153.94 ยฑ2.26 / 157.59 ms โ 151.91 / 152.90 ยฑ1.10 / 155.00 ms โ no change โ
โ QQuery 86 โ 40.63 / 41.67 ยฑ0.75 / 42.90 ms โ 39.35 / 40.02 ยฑ0.61 / 41.05 ms โ no change โ
โ QQuery 87 โ 85.17 / 89.64 ยฑ3.60 / 95.65 ms โ 87.39 / 89.11 ยฑ2.24 / 93.51 ms โ no change โ
โ QQuery 88 โ 99.87 / 100.75 ยฑ0.58 / 101.42 ms โ 98.83 / 100.06 ยฑ0.65 / 100.76 ms โ no change โ
โ QQuery 89 โ 119.57 / 120.72 ยฑ1.13 / 122.46 ms โ 118.54 / 120.09 ยฑ0.91 / 121.30 ms โ no change โ
โ QQuery 90 โ 25.15 / 25.45 ยฑ0.32 / 26.04 ms โ 23.84 / 24.55 ยฑ0.94 / 26.40 ms โ no change โ
โ QQuery 91 โ 62.35 / 64.96 ยฑ1.40 / 66.45 ms โ 65.44 / 67.38 ยฑ1.24 / 69.34 ms โ no change โ
โ QQuery 92 โ 58.37 / 59.13 ยฑ0.61 / 60.23 ms โ 58.93 / 59.24 ยฑ0.24 / 59.49 ms โ no change โ
โ QQuery 93 โ 196.49 / 197.48 ยฑ0.51 / 197.97 ms โ 193.84 / 195.01 ยฑ1.47 / 197.70 ms โ no change โ
โ QQuery 94 โ 62.58 / 63.22 ยฑ0.90 / 64.99 ms โ 62.15 / 63.36 ยฑ1.06 / 65.15 ms โ no change โ
โ QQuery 95 โ 135.65 / 137.41 ยฑ1.56 / 140.17 ms โ 133.45 / 136.50 ยฑ1.53 / 137.34 ms โ no change โ
โ QQuery 96 โ 73.35 / 74.76 ยฑ0.94 / 76.24 ms โ 73.98 / 75.04 ยฑ0.73 / 76.15 ms โ no change โ
โ QQuery 97 โ 129.92 / 133.53 ยฑ2.16 / 136.55 ms โ 132.55 / 133.54 ยฑ0.83 / 134.99 ms โ no change โ
โ QQuery 98 โ 153.89 / 157.07 ยฑ2.57 / 160.05 ms โ 153.25 / 158.44 ยฑ2.98 / 161.39 ms โ no change โ
โ QQuery 99 โ 10840.72 / 10888.12 ยฑ33.72 / 10927.00 ms โ 10810.18 / 10843.75 ยฑ23.28 / 10872.78 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 34167.95ms โ
โ Total Time (lazy-sort-preserving-merge) โ 34147.82ms โ
โ Average Time (HEAD) โ 345.13ms โ
โ Average Time (lazy-sort-preserving-merge) โ 344.93ms โ
โ Queries Faster โ 1 โ
โ Queries Slower โ 1 โ
โ Queries with No Change โ 97 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpcds โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 171.2s |
| Peak memory | 5.5 GiB |
| Avg memory | 4.5 GiB |
| CPU user | 275.1s |
| CPU sys | 18.1s |
| Disk read | 0 B |
| Disk write | 639.0 MiB |
**tpcds โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 171.1s |
| Peak memory | 5.8 GiB |
| Avg memory | 4.7 GiB |
| CPU user | 274.3s |
| CPU sys | 18.4s |
| Disk read | 0 B |
| Disk write | 796.0 KiB |
</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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4183900202
Do you expect any performance change due to this PR?
--
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "github-actions[bot] (via GitHub)" <gi...@apache.org>.
github-actions[bot] closed pull request #21328: Defer task spawning in SortPreservingMergeExec to first poll
URL: https://github.com/apache/datafusion/pull/21328
--
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4186938218
> Do you expect any performance change due to this PR?
Not really - `SortPreservingMergeExec` is almost always on the root of the plan, so it will be executed on the first poll (directly after creating it, instead of using `execute`).
In cases it would do more than once in a plan (e.g. union of 2 `SortPreservingMergeExec`) it could actually slow things down as it will no longer run both in parallel.
CoalescePartitions also does the eager spawning, and in queries with non-optimal concurrency, this leads to more parallelism
But in my opinion we want to avoid this kind of implicit pipeline parallelism and use morsel-based parallelism do improve / reduce memory usage + (data/instruction) cache friendliness.
--
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4181822009
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4181752523)
**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 lazy-sort-preserving-merge
--------------------
Benchmark tpcds_sf1.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโ
โ Query โ HEAD โ lazy-sort-preserving-merge โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 43.33 / 43.90 ยฑ0.59 / 45.01 ms โ 42.60 / 43.61 ยฑ1.10 / 45.74 ms โ no change โ
โ QQuery 2 โ 145.86 / 146.59 ยฑ0.84 / 148.21 ms โ 145.36 / 146.25 ยฑ0.80 / 147.33 ms โ no change โ
โ QQuery 3 โ 113.78 / 114.73 ยฑ0.64 / 115.77 ms โ 114.92 / 115.42 ยฑ0.43 / 116.03 ms โ no change โ
โ QQuery 4 โ 1291.17 / 1326.38 ยฑ27.16 / 1360.87 ms โ 1285.33 / 1319.95 ยฑ26.44 / 1343.59 ms โ no change โ
โ QQuery 5 โ 174.10 / 175.98 ยฑ1.57 / 178.86 ms โ 173.15 / 175.08 ยฑ1.13 / 176.52 ms โ no change โ
โ QQuery 6 โ 978.33 / 1027.61 ยฑ54.28 / 1116.11 ms โ 974.52 / 1002.49 ยฑ18.63 / 1028.45 ms โ no change โ
โ QQuery 7 โ 353.36 / 355.21 ยฑ1.60 / 357.71 ms โ 356.32 / 357.42 ยฑ0.69 / 358.49 ms โ no change โ
โ QQuery 8 โ 115.81 / 117.03 ยฑ0.99 / 118.84 ms โ 114.68 / 116.30 ยฑ1.23 / 118.41 ms โ no change โ
โ QQuery 9 โ 100.58 / 109.13 ยฑ9.90 / 128.35 ms โ 100.75 / 103.73 ยฑ2.31 / 107.46 ms โ no change โ
โ QQuery 10 โ 106.29 / 107.37 ยฑ0.87 / 108.43 ms โ 108.74 / 110.08 ยฑ1.12 / 111.98 ms โ no change โ
โ QQuery 11 โ 907.75 / 917.50 ยฑ6.51 / 924.75 ms โ 898.87 / 915.97 ยฑ12.78 / 933.38 ms โ no change โ
โ QQuery 12 โ 44.85 / 46.34 ยฑ1.31 / 48.54 ms โ 45.50 / 47.11 ยฑ1.21 / 49.21 ms โ no change โ
โ QQuery 13 โ 400.33 / 403.50 ยฑ3.51 / 409.07 ms โ 403.04 / 405.76 ยฑ2.38 / 410.11 ms โ no change โ
โ QQuery 14 โ 1012.83 / 1024.31 ยฑ6.45 / 1032.38 ms โ 1012.03 / 1021.15 ยฑ8.65 / 1035.36 ms โ no change โ
โ QQuery 15 โ 15.15 / 16.28 ยฑ0.89 / 17.58 ms โ 15.39 / 16.53 ยฑ1.04 / 18.20 ms โ no change โ
โ QQuery 16 โ 40.93 / 42.34 ยฑ1.75 / 45.54 ms โ 40.77 / 42.41 ยฑ1.16 / 43.69 ms โ no change โ
โ QQuery 17 โ 240.15 / 243.10 ยฑ1.78 / 245.27 ms โ 239.04 / 241.52 ยฑ2.42 / 245.58 ms โ no change โ
โ QQuery 18 โ 129.34 / 129.89 ยฑ0.49 / 130.77 ms โ 129.88 / 131.23 ยฑ0.90 / 132.70 ms โ no change โ
โ QQuery 19 โ 155.82 / 157.25 ยฑ1.26 / 158.82 ms โ 154.24 / 155.97 ยฑ1.08 / 157.26 ms โ no change โ
โ QQuery 20 โ 13.85 / 14.22 ยฑ0.21 / 14.42 ms โ 13.76 / 14.32 ยฑ0.41 / 14.93 ms โ no change โ
โ QQuery 21 โ 19.28 / 19.74 ยฑ0.40 / 20.17 ms โ 19.07 / 19.46 ยฑ0.33 / 19.92 ms โ no change โ
โ QQuery 22 โ 482.19 / 487.38 ยฑ4.31 / 495.13 ms โ 484.74 / 497.44 ยฑ10.51 / 514.60 ms โ no change โ
โ QQuery 23 โ 881.87 / 894.29 ยฑ10.61 / 910.49 ms โ 874.69 / 895.83 ยฑ12.40 / 909.18 ms โ no change โ
โ QQuery 24 โ 416.24 / 419.83 ยฑ3.48 / 426.42 ms โ 414.89 / 417.76 ยฑ2.70 / 422.45 ms โ no change โ
โ QQuery 25 โ 354.86 / 357.19 ยฑ1.71 / 359.51 ms โ 353.06 / 356.45 ยฑ1.87 / 358.11 ms โ no change โ
โ QQuery 26 โ 82.38 / 84.00 ยฑ1.31 / 85.78 ms โ 83.17 / 85.40 ยฑ2.04 / 88.65 ms โ no change โ
โ QQuery 27 โ 350.26 / 352.45 ยฑ1.88 / 355.62 ms โ 349.09 / 351.61 ยฑ3.11 / 357.67 ms โ no change โ
โ QQuery 28 โ 148.39 / 152.75 ยฑ6.48 / 165.61 ms โ 149.44 / 151.48 ยฑ1.28 / 153.27 ms โ no change โ
โ QQuery 29 โ 298.59 / 299.94 ยฑ0.79 / 300.77 ms โ 297.75 / 299.62 ยฑ1.66 / 302.72 ms โ no change โ
โ QQuery 30 โ 43.68 / 44.81 ยฑ1.53 / 47.74 ms โ 43.22 / 45.64 ยฑ1.49 / 47.68 ms โ no change โ
โ QQuery 31 โ 169.95 / 171.16 ยฑ0.65 / 171.82 ms โ 170.87 / 171.58 ยฑ0.68 / 172.75 ms โ no change โ
โ QQuery 32 โ 56.99 / 58.22 ยฑ1.14 / 60.38 ms โ 58.90 / 60.34 ยฑ1.04 / 61.53 ms โ no change โ
โ QQuery 33 โ 139.93 / 141.33 ยฑ1.24 / 143.43 ms โ 140.41 / 143.01 ยฑ2.16 / 146.93 ms โ no change โ
โ QQuery 34 โ 106.73 / 108.06 ยฑ0.86 / 108.88 ms โ 107.15 / 107.75 ยฑ0.53 / 108.60 ms โ no change โ
โ QQuery 35 โ 107.97 / 109.20 ยฑ0.89 / 110.21 ms โ 107.98 / 109.95 ยฑ1.28 / 111.45 ms โ no change โ
โ QQuery 36 โ 213.24 / 220.07 ยฑ4.91 / 227.00 ms โ 211.95 / 219.95 ยฑ4.89 / 224.58 ms โ no change โ
โ QQuery 37 โ 177.32 / 180.08 ยฑ2.31 / 182.74 ms โ 177.08 / 180.33 ยฑ2.52 / 183.22 ms โ no change โ
โ QQuery 38 โ 81.58 / 85.57 ยฑ2.59 / 89.57 ms โ 86.77 / 88.58 ยฑ2.08 / 92.55 ms โ no change โ
โ QQuery 39 โ 121.53 / 126.67 ยฑ3.28 / 129.58 ms โ 125.75 / 127.19 ยฑ1.43 / 129.60 ms โ no change โ
โ QQuery 40 โ 109.40 / 115.79 ยฑ5.57 / 125.90 ms โ 113.00 / 118.92 ยฑ5.60 / 128.23 ms โ no change โ
โ QQuery 41 โ 14.30 / 14.93 ยฑ0.43 / 15.52 ms โ 14.19 / 15.19 ยฑ0.70 / 16.35 ms โ no change โ
โ QQuery 42 โ 108.00 / 109.87 ยฑ1.30 / 111.43 ms โ 109.29 / 110.10 ยฑ0.47 / 110.68 ms โ no change โ
โ QQuery 43 โ 83.37 / 84.40 ยฑ0.69 / 85.42 ms โ 83.07 / 84.26 ยฑ0.63 / 84.80 ms โ no change โ
โ QQuery 44 โ 11.14 / 12.29 ยฑ1.14 / 14.19 ms โ 12.63 / 13.46 ยฑ0.65 / 14.58 ms โ 1.10x slower โ
โ QQuery 45 โ 52.03 / 53.37 ยฑ0.96 / 54.93 ms โ 52.85 / 54.29 ยฑ0.93 / 55.47 ms โ no change โ
โ QQuery 46 โ 230.81 / 233.84 ยฑ2.54 / 236.85 ms โ 230.42 / 233.34 ยฑ2.20 / 236.82 ms โ no change โ
โ QQuery 47 โ 690.77 / 699.25 ยฑ5.69 / 705.12 ms โ 691.60 / 698.12 ยฑ4.58 / 703.50 ms โ no change โ
โ QQuery 48 โ 287.59 / 292.03 ยฑ2.72 / 294.74 ms โ 285.43 / 292.39 ยฑ4.12 / 297.07 ms โ no change โ
โ QQuery 49 โ 254.20 / 257.86 ยฑ2.43 / 261.70 ms โ 252.86 / 256.64 ยฑ3.15 / 260.64 ms โ no change โ
โ QQuery 50 โ 227.55 / 233.98 ยฑ4.42 / 239.48 ms โ 231.67 / 235.20 ยฑ2.28 / 238.03 ms โ no change โ
โ QQuery 51 โ 183.35 / 186.10 ยฑ2.09 / 189.77 ms โ 182.33 / 187.64 ยฑ3.59 / 193.29 ms โ no change โ
โ QQuery 52 โ 107.69 / 108.46 ยฑ0.61 / 109.49 ms โ 108.11 / 109.76 ยฑ1.42 / 111.77 ms โ no change โ
โ QQuery 53 โ 102.81 / 104.27 ยฑ1.28 / 106.25 ms โ 103.21 / 104.51 ยฑ1.54 / 106.92 ms โ no change โ
โ QQuery 54 โ 147.54 / 150.34 ยฑ2.23 / 154.36 ms โ 145.59 / 149.42 ยฑ2.85 / 153.25 ms โ no change โ
โ QQuery 55 โ 107.95 / 111.01 ยฑ2.21 / 113.70 ms โ 107.71 / 108.16 ยฑ0.31 / 108.63 ms โ no change โ
โ QQuery 56 โ 142.57 / 144.67 ยฑ1.35 / 146.26 ms โ 141.59 / 143.55 ยฑ1.71 / 145.84 ms โ no change โ
โ QQuery 57 โ 174.74 / 177.49 ยฑ1.92 / 179.96 ms โ 173.33 / 174.95 ยฑ1.58 / 177.35 ms โ no change โ
โ QQuery 58 โ 300.57 / 311.41 ยฑ7.57 / 321.60 ms โ 298.56 / 302.55 ยฑ3.97 / 308.96 ms โ no change โ
โ QQuery 59 โ 200.11 / 202.23 ยฑ1.33 / 203.75 ms โ 196.55 / 198.58 ยฑ2.43 / 203.16 ms โ no change โ
โ QQuery 60 โ 144.03 / 147.49 ยฑ1.73 / 148.53 ms โ 145.73 / 148.05 ยฑ1.28 / 149.60 ms โ no change โ
โ QQuery 61 โ 173.23 / 174.99 ยฑ1.12 / 176.12 ms โ 170.52 / 172.47 ยฑ1.21 / 174.30 ms โ no change โ
โ QQuery 62 โ 882.40 / 940.95 ยฑ38.42 / 991.67 ms โ 862.62 / 913.48 ยฑ40.19 / 960.99 ms โ no change โ
โ QQuery 63 โ 103.64 / 105.04 ยฑ1.28 / 107.13 ms โ 103.52 / 106.79 ยฑ2.74 / 110.93 ms โ no change โ
โ QQuery 64 โ 695.26 / 700.96 ยฑ4.85 / 708.80 ms โ 702.18 / 707.20 ยฑ4.53 / 712.87 ms โ no change โ
โ QQuery 65 โ 249.76 / 253.22 ยฑ2.52 / 257.47 ms โ 251.05 / 256.02 ยฑ3.12 / 260.46 ms โ no change โ
โ QQuery 66 โ 245.34 / 263.55 ยฑ13.98 / 281.04 ms โ 253.60 / 257.57 ยฑ4.16 / 265.08 ms โ no change โ
โ QQuery 67 โ 306.88 / 315.17 ยฑ6.36 / 325.21 ms โ 305.82 / 317.69 ยฑ9.95 / 333.95 ms โ no change โ
โ QQuery 68 โ 280.35 / 283.72 ยฑ2.36 / 287.14 ms โ 282.41 / 285.69 ยฑ2.20 / 288.34 ms โ no change โ
โ QQuery 69 โ 104.08 / 104.52 ยฑ0.35 / 104.98 ms โ 104.07 / 104.95 ยฑ0.69 / 105.85 ms โ no change โ
โ QQuery 70 โ 332.94 / 343.97 ยฑ8.90 / 358.95 ms โ 342.38 / 350.19 ยฑ5.05 / 355.69 ms โ no change โ
โ QQuery 71 โ 133.53 / 135.86 ยฑ1.91 / 138.29 ms โ 136.04 / 137.94 ยฑ1.66 / 140.80 ms โ no change โ
โ QQuery 72 โ 719.23 / 726.48 ยฑ6.45 / 735.78 ms โ 695.92 / 724.46 ยฑ19.52 / 753.85 ms โ no change โ
โ QQuery 73 โ 104.33 / 104.71 ยฑ0.39 / 105.43 ms โ 103.25 / 104.47 ยฑ1.10 / 106.41 ms โ no change โ
โ QQuery 74 โ 547.94 / 561.60 ยฑ9.54 / 577.65 ms โ 547.60 / 556.21 ยฑ7.51 / 565.68 ms โ no change โ
โ QQuery 75 โ 277.23 / 278.48 ยฑ1.32 / 280.85 ms โ 274.99 / 278.26 ยฑ2.03 / 281.39 ms โ no change โ
โ QQuery 76 โ 131.87 / 134.41 ยฑ2.02 / 137.28 ms โ 132.33 / 133.98 ยฑ1.13 / 135.57 ms โ no change โ
โ QQuery 77 โ 188.52 / 190.41 ยฑ1.90 / 193.52 ms โ 188.78 / 190.32 ยฑ0.85 / 191.26 ms โ no change โ
โ QQuery 78 โ 351.04 / 356.39 ยฑ2.92 / 359.04 ms โ 350.14 / 356.15 ยฑ5.56 / 364.46 ms โ no change โ
โ QQuery 79 โ 234.10 / 236.10 ยฑ2.30 / 240.22 ms โ 232.58 / 236.22 ยฑ2.72 / 239.48 ms โ no change โ
โ QQuery 80 โ 327.96 / 331.10 ยฑ1.88 / 333.34 ms โ 328.56 / 331.32 ยฑ1.88 / 333.61 ms โ no change โ
โ QQuery 81 โ 26.06 / 27.20 ยฑ1.10 / 29.20 ms โ 27.20 / 29.02 ยฑ1.72 / 31.89 ms โ 1.07x slower โ
โ QQuery 82 โ 200.51 / 203.04 ยฑ1.79 / 205.85 ms โ 198.98 / 201.55 ยฑ2.45 / 204.70 ms โ no change โ
โ QQuery 83 โ 39.51 / 40.04 ยฑ0.50 / 40.89 ms โ 40.23 / 41.36 ยฑ0.70 / 42.21 ms โ no change โ
โ QQuery 84 โ 47.95 / 49.26 ยฑ1.10 / 50.81 ms โ 48.81 / 50.19 ยฑ0.91 / 51.68 ms โ no change โ
โ QQuery 85 โ 146.73 / 152.21 ยฑ2.91 / 154.68 ms โ 150.00 / 151.62 ยฑ1.45 / 154.35 ms โ no change โ
โ QQuery 86 โ 39.07 / 40.11 ยฑ0.84 / 41.50 ms โ 39.48 / 40.42 ยฑ0.74 / 41.61 ms โ no change โ
โ QQuery 87 โ 85.16 / 88.74 ยฑ3.50 / 95.17 ms โ 85.41 / 89.92 ยฑ4.35 / 97.86 ms โ no change โ
โ QQuery 88 โ 99.34 / 99.84 ยฑ0.48 / 100.46 ms โ 98.97 / 99.87 ยฑ0.63 / 100.49 ms โ no change โ
โ QQuery 89 โ 118.40 / 120.21 ยฑ1.00 / 121.21 ms โ 118.47 / 121.11 ยฑ1.65 / 123.66 ms โ no change โ
โ QQuery 90 โ 23.02 / 24.44 ยฑ1.03 / 26.20 ms โ 23.48 / 23.99 ยฑ0.39 / 24.55 ms โ no change โ
โ QQuery 91 โ 62.73 / 65.39 ยฑ1.71 / 67.79 ms โ 63.88 / 65.70 ยฑ1.43 / 67.85 ms โ no change โ
โ QQuery 92 โ 58.18 / 58.54 ยฑ0.22 / 58.87 ms โ 57.06 / 58.16 ยฑ0.86 / 59.39 ms โ no change โ
โ QQuery 93 โ 191.21 / 193.79 ยฑ1.60 / 195.93 ms โ 193.12 / 194.16 ยฑ1.11 / 195.88 ms โ no change โ
โ QQuery 94 โ 60.77 / 62.08 ยฑ0.87 / 63.35 ms โ 61.04 / 62.44 ยฑ0.81 / 63.21 ms โ no change โ
โ QQuery 95 โ 136.40 / 137.05 ยฑ0.41 / 137.42 ms โ 133.59 / 136.29 ยฑ2.07 / 139.81 ms โ no change โ
โ QQuery 96 โ 71.92 / 74.42 ยฑ1.31 / 75.59 ms โ 72.89 / 73.98 ยฑ0.57 / 74.49 ms โ no change โ
โ QQuery 97 โ 129.90 / 132.28 ยฑ1.88 / 135.62 ms โ 130.06 / 132.19 ยฑ1.10 / 133.13 ms โ no change โ
โ QQuery 98 โ 155.24 / 157.75 ยฑ2.57 / 161.85 ms โ 154.37 / 155.70 ยฑ1.10 / 157.25 ms โ no change โ
โ QQuery 99 โ 10792.16 / 10816.19 ยฑ23.48 / 10855.52 ms โ 10767.72 / 10780.65 ยฑ8.63 / 10792.51 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 33666.69ms โ
โ Total Time (lazy-sort-preserving-merge) โ 33580.44ms โ
โ Average Time (HEAD) โ 340.07ms โ
โ Average Time (lazy-sort-preserving-merge) โ 339.20ms โ
โ Queries Faster โ 0 โ
โ Queries Slower โ 2 โ
โ Queries with No Change โ 97 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpcds โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 168.7s |
| Peak memory | 5.5 GiB |
| Avg memory | 4.7 GiB |
| CPU user | 270.2s |
| CPU sys | 18.6s |
| Disk read | 0 B |
| Disk write | 703.0 MiB |
**tpcds โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 168.2s |
| Peak memory | 5.6 GiB |
| Avg memory | 4.7 GiB |
| CPU user | 268.7s |
| CPU sys | 19.0s |
| Disk read | 0 B |
| Disk write | 148.0 KiB |
</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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4182041172
cc @neilconway
--
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180154858
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180142002)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4180142002-674-fflk2 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 lazy-sort-preserving-merge (6eff5503672506aa9e96627e40b577937bb32151) to 4084a18 (merge-base) [diff](https://github.com/apache/datafusion/compare/4084a18bc709d65f9901fa0d51932867ce9f6342..6eff5503672506aa9e96627e40b577937bb32151) 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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180154803
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180142002)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4180142002-673-xf9zw 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 lazy-sort-preserving-merge (6eff5503672506aa9e96627e40b577937bb32151) to 4084a18 (merge-base) [diff](https://github.com/apache/datafusion/compare/4084a18bc709d65f9901fa0d51932867ce9f6342..6eff5503672506aa9e96627e40b577937bb32151) 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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "github-actions[bot] (via GitHub)" <gi...@apache.org>.
github-actions[bot] commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4618379997
Thank you for your contribution. Unfortunately, this pull request is stale because it has been open 60 days with no activity. Please remove the stale label or comment or this will be closed in 7 days.
--
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180112117
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180026900)
**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 lazy-sort-preserving-merge
--------------------
Benchmark clickbench_partitioned.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ lazy-sort-preserving-merge โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 0 โ 1.32 / 4.53 ยฑ6.34 / 17.22 ms โ 1.29 / 4.46 ยฑ6.24 / 16.94 ms โ no change โ
โ QQuery 1 โ 14.26 / 14.32 ยฑ0.06 / 14.43 ms โ 14.31 / 14.63 ยฑ0.35 / 15.22 ms โ no change โ
โ QQuery 2 โ 43.50 / 43.99 ยฑ0.29 / 44.30 ms โ 44.65 / 44.89 ยฑ0.21 / 45.24 ms โ no change โ
โ QQuery 3 โ 42.60 / 44.10 ยฑ1.05 / 45.64 ms โ 44.74 / 46.10 ยฑ1.17 / 48.20 ms โ no change โ
โ QQuery 4 โ 284.07 / 294.39 ยฑ5.26 / 298.70 ms โ 284.03 / 288.81 ยฑ5.55 / 299.45 ms โ no change โ
โ QQuery 5 โ 343.27 / 349.08 ยฑ4.20 / 356.08 ms โ 343.96 / 346.46 ยฑ2.99 / 352.17 ms โ no change โ
โ QQuery 6 โ 5.43 / 6.62 ยฑ0.97 / 7.65 ms โ 5.56 / 8.40 ยฑ3.44 / 15.04 ms โ 1.27x slower โ
โ QQuery 7 โ 16.40 / 16.80 ยฑ0.32 / 17.36 ms โ 16.61 / 17.13 ยฑ0.59 / 18.26 ms โ no change โ
โ QQuery 8 โ 411.18 / 417.90 ยฑ5.23 / 424.82 ms โ 415.21 / 443.66 ยฑ24.16 / 487.53 ms โ 1.06x slower โ
โ QQuery 9 โ 621.24 / 633.96 ยฑ10.59 / 650.55 ms โ 636.44 / 644.05 ยฑ6.47 / 655.07 ms โ no change โ
โ QQuery 10 โ 89.86 / 92.43 ยฑ2.58 / 96.80 ms โ 91.28 / 93.97 ยฑ2.99 / 99.69 ms โ no change โ
โ QQuery 11 โ 102.96 / 103.79 ยฑ0.62 / 104.50 ms โ 104.08 / 106.06 ยฑ2.49 / 110.93 ms โ no change โ
โ QQuery 12 โ 337.21 / 346.86 ยฑ6.25 / 353.55 ms โ 342.60 / 346.61 ยฑ2.44 / 350.22 ms โ no change โ
โ QQuery 13 โ 446.13 / 456.41 ยฑ11.22 / 477.92 ms โ 457.53 / 469.03 ยฑ11.93 / 485.25 ms โ no change โ
โ QQuery 14 โ 348.18 / 352.23 ยฑ2.50 / 354.77 ms โ 343.58 / 352.71 ยฑ5.81 / 360.78 ms โ no change โ
โ QQuery 15 โ 361.82 / 372.69 ยฑ10.23 / 390.58 ms โ 356.76 / 375.85 ยฑ20.43 / 411.72 ms โ no change โ
โ QQuery 16 โ 721.28 / 745.37 ยฑ24.91 / 790.84 ms โ 718.26 / 750.57 ยฑ30.65 / 797.57 ms โ no change โ
โ QQuery 17 โ 718.07 / 725.05 ยฑ4.61 / 732.25 ms โ 709.67 / 720.86 ยฑ6.87 / 729.55 ms โ no change โ
โ QQuery 18 โ 1446.03 / 1469.53 ยฑ19.14 / 1495.12 ms โ 1385.38 / 1472.36 ยฑ49.24 / 1529.28 ms โ no change โ
โ QQuery 19 โ 34.89 / 42.01 ยฑ7.77 / 55.18 ms โ 34.89 / 39.20 ยฑ6.16 / 51.40 ms โ +1.07x faster โ
โ QQuery 20 โ 710.86 / 728.65 ยฑ16.02 / 753.45 ms โ 714.91 / 724.05 ยฑ10.57 / 744.00 ms โ no change โ
โ QQuery 21 โ 756.38 / 761.28 ยฑ3.28 / 765.11 ms โ 752.40 / 759.53 ยฑ5.97 / 766.99 ms โ no change โ
โ QQuery 22 โ 1116.82 / 1125.07 ยฑ6.94 / 1137.42 ms โ 1120.38 / 1127.16 ยฑ5.22 / 1132.06 ms โ no change โ
โ QQuery 23 โ 3017.85 / 3046.09 ยฑ29.48 / 3098.88 ms โ 3048.50 / 3085.51 ยฑ26.50 / 3127.09 ms โ no change โ
โ QQuery 24 โ 93.49 / 99.75 ยฑ3.45 / 102.69 ms โ 98.60 / 100.83 ยฑ1.85 / 103.55 ms โ no change โ
โ QQuery 25 โ 138.66 / 141.29 ยฑ1.88 / 143.55 ms โ 137.64 / 139.90 ยฑ1.83 / 142.26 ms โ no change โ
โ QQuery 26 โ 97.38 / 100.15 ยฑ1.74 / 102.67 ms โ 97.34 / 100.87 ยฑ2.06 / 103.55 ms โ no change โ
โ QQuery 27 โ 844.72 / 849.46 ยฑ3.92 / 856.45 ms โ 843.28 / 848.90 ยฑ6.37 / 861.06 ms โ no change โ
โ QQuery 28 โ 7671.11 / 7732.92 ยฑ34.70 / 7774.83 ms โ 7685.77 / 7735.32 ยฑ27.53 / 7770.18 ms โ no change โ
โ QQuery 29 โ 50.46 / 53.13 ยฑ2.47 / 57.11 ms โ 50.30 / 58.03 ยฑ8.62 / 74.88 ms โ 1.09x slower โ
โ QQuery 30 โ 354.21 / 363.29 ยฑ6.86 / 375.08 ms โ 359.46 / 365.45 ยฑ4.53 / 371.06 ms โ no change โ
โ QQuery 31 โ 361.59 / 376.27 ยฑ9.32 / 388.30 ms โ 376.30 / 389.06 ยฑ10.84 / 407.57 ms โ no change โ
โ QQuery 32 โ 1028.28 / 1058.14 ยฑ28.22 / 1106.02 ms โ 1284.87 / 1292.00 ยฑ4.47 / 1298.32 ms โ 1.22x slower โ
โ QQuery 33 โ 1443.84 / 1455.00 ยฑ8.74 / 1463.78 ms โ 1442.67 / 1502.95 ยฑ40.15 / 1565.26 ms โ no change โ
โ QQuery 34 โ 1436.41 / 1447.06 ยฑ9.23 / 1458.79 ms โ 1445.50 / 1463.29 ยฑ12.64 / 1482.56 ms โ no change โ
โ QQuery 35 โ 383.94 / 386.50 ยฑ1.30 / 387.56 ms โ 380.85 / 388.68 ยฑ4.94 / 394.35 ms โ no change โ
โ QQuery 36 โ 116.47 / 120.38 ยฑ3.13 / 124.25 ms โ 121.81 / 122.82 ยฑ1.16 / 125.05 ms โ no change โ
โ QQuery 37 โ 46.13 / 49.62 ยฑ2.02 / 51.84 ms โ 49.31 / 50.12 ยฑ0.68 / 51.35 ms โ no change โ
โ QQuery 38 โ 76.61 / 77.14 ยฑ0.85 / 78.84 ms โ 75.21 / 76.53 ยฑ0.99 / 78.10 ms โ no change โ
โ QQuery 39 โ 203.54 / 213.22 ยฑ6.50 / 223.27 ms โ 213.54 / 221.07 ยฑ4.50 / 225.31 ms โ no change โ
โ QQuery 40 โ 23.82 / 25.79 ยฑ1.66 / 28.77 ms โ 22.73 / 24.95 ยฑ1.91 / 27.42 ms โ no change โ
โ QQuery 41 โ 19.13 / 20.56 ยฑ0.98 / 22.07 ms โ 18.94 / 20.07 ยฑ0.79 / 21.33 ms โ no change โ
โ QQuery 42 โ 18.72 / 19.53 ยฑ0.56 / 20.35 ms โ 18.96 / 19.55 ยฑ0.35 / 19.87 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 26782.34ms โ
โ Total Time (lazy-sort-preserving-merge) โ 27202.41ms โ
โ Average Time (HEAD) โ 622.85ms โ
โ Average Time (lazy-sort-preserving-merge) โ 632.61ms โ
โ Queries Faster โ 1 โ
โ Queries Slower โ 4 โ
โ Queries with No Change โ 38 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 135.0s |
| Peak memory | 41.0 GiB |
| Avg memory | 31.5 GiB |
| CPU user | 1276.4s |
| CPU sys | 86.8s |
| Disk read | 0 B |
| Disk write | 4.5 GiB |
**clickbench_partitioned โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 137.1s |
| Peak memory | 41.0 GiB |
| Avg memory | 31.8 GiB |
| CPU user | 1277.3s |
| CPU sys | 101.0s |
| Disk read | 0 B |
| Disk write | 532.0 KiB |
</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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180087552
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180026900)
**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 lazy-sort-preserving-merge
--------------------
Benchmark tpch_sf1.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโ
โ Query โ HEAD โ lazy-sort-preserving-merge โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 1 โ 45.64 / 46.21 ยฑ0.73 / 47.65 ms โ 45.51 / 46.17 ยฑ0.85 / 47.83 ms โ no change โ
โ QQuery 2 โ 21.26 / 21.68 ยฑ0.43 / 22.50 ms โ 21.32 / 21.54 ยฑ0.15 / 21.74 ms โ no change โ
โ QQuery 3 โ 31.86 / 32.17 ยฑ0.17 / 32.35 ms โ 31.97 / 32.81 ยฑ1.05 / 34.85 ms โ no change โ
โ QQuery 4 โ 20.66 / 21.58 ยฑ0.69 / 22.57 ms โ 20.57 / 21.69 ยฑ1.10 / 23.21 ms โ no change โ
โ QQuery 5 โ 48.61 / 50.56 ยฑ1.27 / 51.75 ms โ 48.18 / 49.26 ยฑ1.26 / 51.68 ms โ no change โ
โ QQuery 6 โ 17.40 / 17.61 ยฑ0.16 / 17.88 ms โ 17.23 / 17.69 ยฑ0.37 / 18.37 ms โ no change โ
โ QQuery 7 โ 54.42 / 55.97 ยฑ1.02 / 57.47 ms โ 54.33 / 56.74 ยฑ1.57 / 59.17 ms โ no change โ
โ QQuery 8 โ 48.83 / 49.03 ยฑ0.18 / 49.36 ms โ 48.64 / 48.87 ยฑ0.17 / 49.14 ms โ no change โ
โ QQuery 9 โ 53.92 / 55.48 ยฑ0.95 / 56.84 ms โ 54.97 / 55.26 ยฑ0.37 / 55.87 ms โ no change โ
โ QQuery 10 โ 72.46 / 73.76 ยฑ1.21 / 76.03 ms โ 70.77 / 73.62 ยฑ1.77 / 76.31 ms โ no change โ
โ QQuery 11 โ 14.25 / 14.42 ยฑ0.17 / 14.67 ms โ 14.30 / 14.84 ยฑ0.57 / 15.84 ms โ no change โ
โ QQuery 12 โ 28.18 / 28.33 ยฑ0.09 / 28.43 ms โ 28.02 / 28.41 ยฑ0.31 / 28.92 ms โ no change โ
โ QQuery 13 โ 39.36 / 40.09 ยฑ0.57 / 40.74 ms โ 38.61 / 39.47 ยฑ0.75 / 40.65 ms โ no change โ
โ QQuery 14 โ 28.61 / 29.13 ยฑ0.40 / 29.84 ms โ 28.85 / 29.10 ยฑ0.19 / 29.36 ms โ no change โ
โ QQuery 15 โ 33.94 / 34.36 ยฑ0.40 / 35.07 ms โ 33.96 / 34.82 ยฑ0.57 / 35.49 ms โ no change โ
โ QQuery 16 โ 16.46 / 16.79 ยฑ0.40 / 17.54 ms โ 15.93 / 16.54 ยฑ0.35 / 16.98 ms โ no change โ
โ QQuery 17 โ 74.11 / 74.76 ยฑ0.68 / 75.98 ms โ 73.41 / 74.18 ยฑ0.53 / 75.08 ms โ no change โ
โ QQuery 18 โ 77.89 / 79.93 ยฑ1.61 / 81.94 ms โ 78.89 / 80.51 ยฑ0.86 / 81.33 ms โ no change โ
โ QQuery 19 โ 37.54 / 37.87 ยฑ0.23 / 38.14 ms โ 37.80 / 38.17 ยฑ0.27 / 38.51 ms โ no change โ
โ QQuery 20 โ 40.91 / 41.55 ยฑ0.57 / 42.50 ms โ 40.55 / 41.32 ยฑ0.49 / 41.80 ms โ no change โ
โ QQuery 21 โ 64.27 / 66.27 ยฑ1.05 / 67.14 ms โ 64.62 / 67.78 ยฑ2.84 / 72.99 ms โ no change โ
โ QQuery 22 โ 17.99 / 18.35 ยฑ0.28 / 18.69 ms โ 18.03 / 18.56 ยฑ0.64 / 19.78 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 905.89ms โ
โ Total Time (lazy-sort-preserving-merge) โ 907.38ms โ
โ Average Time (HEAD) โ 41.18ms โ
โ Average Time (lazy-sort-preserving-merge) โ 41.24ms โ
โ Queries Faster โ 0 โ
โ Queries Slower โ 0 โ
โ Queries with No Change โ 22 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**tpch โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 4.8s |
| Peak memory | 4.1 GiB |
| Avg memory | 3.6 GiB |
| CPU user | 33.7s |
| CPU sys | 3.0s |
| Disk read | 0 B |
| Disk write | 136.0 KiB |
**tpch โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 4.8s |
| Peak memory | 4.1 GiB |
| Avg memory | 3.6 GiB |
| CPU user | 33.6s |
| CPU sys | 3.0s |
| Disk read | 0 B |
| Disk write | 72.0 KiB |
</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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180223355
๐ค Benchmark completed (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180142002)
**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 lazy-sort-preserving-merge
--------------------
Benchmark clickbench_partitioned.json
--------------------
โโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโโโโ
โ Query โ HEAD โ lazy-sort-preserving-merge โ Change โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ QQuery 0 โ 1.28 / 4.56 ยฑ6.40 / 17.35 ms โ 1.33 / 4.62 ยฑ6.45 / 17.51 ms โ no change โ
โ QQuery 1 โ 14.32 / 14.77 ยฑ0.26 / 15.03 ms โ 14.43 / 14.73 ยฑ0.18 / 14.94 ms โ no change โ
โ QQuery 2 โ 44.78 / 45.12 ยฑ0.31 / 45.67 ms โ 44.80 / 44.99 ยฑ0.20 / 45.32 ms โ no change โ
โ QQuery 3 โ 44.69 / 45.32 ยฑ0.66 / 46.53 ms โ 43.80 / 45.20 ยฑ0.84 / 46.39 ms โ no change โ
โ QQuery 4 โ 292.74 / 300.41 ยฑ6.35 / 309.33 ms โ 293.46 / 303.78 ยฑ9.47 / 319.34 ms โ no change โ
โ QQuery 5 โ 346.77 / 353.15 ยฑ4.55 / 359.72 ms โ 346.87 / 352.60 ยฑ7.05 / 365.62 ms โ no change โ
โ QQuery 6 โ 5.92 / 6.64 ยฑ0.58 / 7.54 ms โ 5.09 / 6.84 ยฑ1.12 / 8.50 ms โ no change โ
โ QQuery 7 โ 16.82 / 17.24 ยฑ0.40 / 17.77 ms โ 17.08 / 17.90 ยฑ0.72 / 19.04 ms โ no change โ
โ QQuery 8 โ 416.34 / 423.89 ยฑ4.29 / 428.17 ms โ 421.55 / 429.08 ยฑ6.37 / 438.74 ms โ no change โ
โ QQuery 9 โ 645.55 / 658.00 ยฑ7.10 / 665.34 ms โ 644.18 / 662.08 ยฑ12.51 / 677.49 ms โ no change โ
โ QQuery 10 โ 91.10 / 94.53 ยฑ3.54 / 101.35 ms โ 94.32 / 96.67 ยฑ2.77 / 101.97 ms โ no change โ
โ QQuery 11 โ 106.17 / 107.42 ยฑ1.12 / 109.47 ms โ 106.35 / 107.83 ยฑ1.51 / 110.72 ms โ no change โ
โ QQuery 12 โ 343.79 / 351.30 ยฑ5.19 / 359.45 ms โ 349.30 / 354.82 ยฑ3.91 / 359.46 ms โ no change โ
โ QQuery 13 โ 463.95 / 469.29 ยฑ3.49 / 473.39 ms โ 458.95 / 471.51 ยฑ10.55 / 490.32 ms โ no change โ
โ QQuery 14 โ 348.51 / 355.80 ยฑ4.87 / 361.33 ms โ 350.91 / 356.20 ยฑ4.46 / 362.28 ms โ no change โ
โ QQuery 15 โ 374.95 / 397.08 ยฑ16.02 / 419.14 ms โ 361.04 / 366.96 ยฑ5.84 / 376.04 ms โ +1.08x faster โ
โ QQuery 16 โ 733.53 / 747.87 ยฑ13.48 / 773.39 ms โ 718.97 / 726.30 ยฑ4.73 / 733.61 ms โ no change โ
โ QQuery 17 โ 718.30 / 722.31 ยฑ4.40 / 730.28 ms โ 714.86 / 727.88 ยฑ8.63 / 740.01 ms โ no change โ
โ QQuery 18 โ 1449.05 / 1486.25 ยฑ35.15 / 1544.04 ms โ 1406.57 / 1410.54 ยฑ3.24 / 1416.33 ms โ +1.05x faster โ
โ QQuery 19 โ 36.78 / 38.16 ยฑ1.46 / 40.15 ms โ 36.72 / 40.40 ยฑ3.56 / 45.15 ms โ 1.06x slower โ
โ QQuery 20 โ 717.03 / 732.32 ยฑ16.82 / 754.06 ms โ 710.26 / 720.21 ยฑ11.70 / 739.57 ms โ no change โ
โ QQuery 21 โ 759.28 / 766.56 ยฑ5.98 / 777.25 ms โ 755.43 / 758.58 ยฑ2.35 / 760.95 ms โ no change โ
โ QQuery 22 โ 1129.70 / 1135.22 ยฑ5.65 / 1145.59 ms โ 1125.50 / 1130.81 ยฑ3.57 / 1136.21 ms โ no change โ
โ QQuery 23 โ 3063.67 / 3092.32 ยฑ30.87 / 3147.47 ms โ 3094.27 / 3106.40 ยฑ6.19 / 3111.59 ms โ no change โ
โ QQuery 24 โ 101.24 / 103.39 ยฑ1.31 / 105.09 ms โ 97.60 / 101.16 ยฑ2.25 / 103.55 ms โ no change โ
โ QQuery 25 โ 137.99 / 139.95 ยฑ2.03 / 142.98 ms โ 138.26 / 140.92 ยฑ1.46 / 142.08 ms โ no change โ
โ QQuery 26 โ 97.32 / 102.20 ยฑ2.89 / 106.16 ms โ 100.80 / 104.18 ยฑ1.86 / 105.77 ms โ no change โ
โ QQuery 27 โ 848.67 / 855.31 ยฑ4.25 / 861.35 ms โ 841.92 / 851.34 ยฑ7.11 / 859.64 ms โ no change โ
โ QQuery 28 โ 7726.17 / 7789.43 ยฑ38.55 / 7840.38 ms โ 7697.56 / 7772.98 ยฑ39.87 / 7816.02 ms โ no change โ
โ QQuery 29 โ 50.30 / 55.49 ยฑ3.32 / 60.76 ms โ 52.18 / 56.37 ยฑ5.13 / 65.83 ms โ no change โ
โ QQuery 30 โ 364.43 / 367.96 ยฑ4.09 / 375.75 ms โ 365.84 / 376.41 ยฑ7.90 / 387.67 ms โ no change โ
โ QQuery 31 โ 365.59 / 380.19 ยฑ7.79 / 388.48 ms โ 352.26 / 379.27 ยฑ17.30 / 404.46 ms โ no change โ
โ QQuery 32 โ 1194.60 / 1265.06 ยฑ48.62 / 1318.77 ms โ 1038.85 / 1054.63 ยฑ17.78 / 1087.37 ms โ +1.20x faster โ
โ QQuery 33 โ 1471.45 / 1515.35 ยฑ42.95 / 1594.90 ms โ 1463.32 / 1480.52 ยฑ20.31 / 1519.49 ms โ no change โ
โ QQuery 34 โ 1479.15 / 1494.24 ยฑ8.83 / 1506.62 ms โ 1466.78 / 1484.75 ยฑ14.27 / 1503.75 ms โ no change โ
โ QQuery 35 โ 396.66 / 400.11 ยฑ1.98 / 402.13 ms โ 393.33 / 401.85 ยฑ8.51 / 416.51 ms โ no change โ
โ QQuery 36 โ 115.41 / 122.18 ยฑ3.78 / 126.87 ms โ 111.74 / 123.56 ยฑ6.49 / 131.46 ms โ no change โ
โ QQuery 37 โ 49.80 / 51.90 ยฑ1.51 / 53.95 ms โ 48.99 / 51.08 ยฑ1.38 / 52.91 ms โ no change โ
โ QQuery 38 โ 75.55 / 76.40 ยฑ0.82 / 77.82 ms โ 74.68 / 75.72 ยฑ0.84 / 77.25 ms โ no change โ
โ QQuery 39 โ 215.43 / 225.19 ยฑ5.40 / 231.67 ms โ 214.25 / 222.56 ยฑ6.42 / 228.45 ms โ no change โ
โ QQuery 40 โ 25.02 / 26.50 ยฑ1.15 / 27.70 ms โ 22.99 / 25.63 ยฑ1.71 / 27.89 ms โ no change โ
โ QQuery 41 โ 20.17 / 21.55 ยฑ1.41 / 24.24 ms โ 20.69 / 22.68 ยฑ1.41 / 24.54 ms โ 1.05x slower โ
โ QQuery 42 โ 19.98 / 20.41 ยฑ0.38 / 20.96 ms โ 20.24 / 20.67 ยฑ0.30 / 21.10 ms โ no change โ
โโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโโโโ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโณโโโโโโโโโโโโโ
โ Benchmark Summary โ โ
โกโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโฉ
โ Total Time (HEAD) โ 27378.37ms โ
โ Total Time (lazy-sort-preserving-merge) โ 27003.20ms โ
โ Average Time (HEAD) โ 636.71ms โ
โ Average Time (lazy-sort-preserving-merge) โ 627.98ms โ
โ Queries Faster โ 3 โ
โ Queries Slower โ 2 โ
โ Queries with No Change โ 38 โ
โ Queries with Failure โ 0 โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโดโโโโโโโโโโโโโ
```
</p>
</details>
<details><summary>Resource Usage</summary>
**clickbench_partitioned โ base (merge-base)**
| Metric | Value |
|--------|-------|
| Wall time | 138.2s |
| Peak memory | 40.5 GiB |
| Avg memory | 31.4 GiB |
| CPU user | 1289.2s |
| CPU sys | 101.5s |
| Disk read | 0 B |
| Disk write | 5.4 GiB |
**clickbench_partitioned โ branch**
| Metric | Value |
|--------|-------|
| Wall time | 136.2s |
| Peak memory | 42.6 GiB |
| Avg memory | 34.8 GiB |
| CPU user | 1290.4s |
| CPU sys | 84.0s |
| Disk read | 0 B |
| Disk write | 120.0 KiB |
</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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4181771958
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4181752523)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4181752523-688-j9zrp 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 lazy-sort-preserving-merge (6eff5503672506aa9e96627e40b577937bb32151) to 4084a18 (merge-base) [diff](https://github.com/apache/datafusion/compare/4084a18bc709d65f9901fa0d51932867ce9f6342..6eff5503672506aa9e96627e40b577937bb32151) 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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180037291
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180026900)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4180026900-669-n59rl 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 lazy-sort-preserving-merge (6eff5503672506aa9e96627e40b577937bb32151) to 4084a18 (merge-base) [diff](https://github.com/apache/datafusion/compare/4084a18bc709d65f9901fa0d51932867ce9f6342..6eff5503672506aa9e96627e40b577937bb32151) 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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180026900
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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "adriangbot (via GitHub)" <gi...@apache.org>.
adriangbot commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4180154559
๐ค Benchmark running (GKE) | [trigger](https://github.com/apache/datafusion/pull/21328#issuecomment-4180142002)
**Instance:** `c4a-highmem-16` (12 vCPU / 65 GiB) | `Linux bench-c4180142002-672-fb76c 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 lazy-sort-preserving-merge (6eff5503672506aa9e96627e40b577937bb32151) to 4084a18 (merge-base) [diff](https://github.com/apache/datafusion/compare/4084a18bc709d65f9901fa0d51932867ce9f6342..6eff5503672506aa9e96627e40b577937bb32151) 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] Defer task spawning in SortPreservingMergeExec to first poll [datafusion]
Posted by "Dandandan (via GitHub)" <gi...@apache.org>.
Dandandan commented on PR #21328:
URL: https://github.com/apache/datafusion/pull/21328#issuecomment-4181752523
run benchmark tpcds
--
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