Skip to content

feat: reuse partial aggregate hashes across repartition#23446

Draft
Rachelint wants to merge 1 commit into
mainfrom
reuse-hash-through-hold-aggr
Draft

feat: reuse partial aggregate hashes across repartition#23446
Rachelint wants to merge 1 commit into
mainfrom
reuse-hash-through-hold-aggr

Conversation

@Rachelint

@Rachelint Rachelint commented Jul 10, 2026

Copy link
Copy Markdown
Contributor

Which issue does this PR close?

  • Closes #.

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

@github-actions github-actions Bot added the physical-plan Changes to the physical-plan crate label Jul 10, 2026
@Rachelint

Rachelint commented Jul 10, 2026

Copy link
Copy Markdown
Contributor Author

run benchmarks clickbench_partitioned

1 similar comment
@Rachelint

Rachelint commented Jul 10, 2026

Copy link
Copy Markdown
Contributor Author

run benchmarks clickbench_partitioned

@adriangbot

adriangbot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4935112942-957-kv7hs 6.12.85+ #1 SMP Mon May 11 08:17:35 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu) 
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

Comparing reuse-hash-through-hold-aggr (82bdb2b) to f34a676 (merge-base) diff using: clickbench_partitioned
Results will be posted here when complete

File an issue against this benchmark runner

@adriangbot

adriangbot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4935121013-958-wd2gp 6.12.85+ #1 SMP Mon May 11 08:17:35 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu) 
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

Comparing reuse-hash-through-hold-aggr (82bdb2b) to f34a676 (merge-base) diff using: clickbench_partitioned
Results will be posted here when complete

File an issue against this benchmark runner

@adriangbot

adriangbot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Instance: c4a-highmem-16 (12 vCPU / 65 GiB)

CPU Details (lscpu) 
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 HEAD and reuse-hash-through-hold-aggr
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                   HEAD ┃           reuse-hash-through-hold-aggr ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │           1.33 / 4.27 ±5.71 / 15.68 ms │           1.51 / 4.52 ±5.86 / 16.25 ms │  1.06x slower │
│ QQuery 1  │         14.10 / 14.51 ±0.57 / 15.64 ms │         14.68 / 14.83 ±0.11 / 14.99 ms │     no change │
│ QQuery 2  │         37.08 / 37.73 ±0.54 / 38.65 ms │         38.04 / 38.63 ±0.41 / 39.32 ms │     no change │
│ QQuery 3  │         31.13 / 32.28 ±1.29 / 34.39 ms │         31.77 / 33.25 ±0.95 / 34.52 ms │     no change │
│ QQuery 4  │      227.46 / 235.56 ±8.78 / 247.71 ms │      267.18 / 269.99 ±3.17 / 275.85 ms │  1.15x slower │
│ QQuery 5  │      270.96 / 275.54 ±4.86 / 284.39 ms │      275.06 / 284.50 ±7.53 / 291.76 ms │     no change │
│ QQuery 6  │            1.27 / 1.42 ±0.23 / 1.88 ms │            1.28 / 1.43 ±0.24 / 1.91 ms │     no change │
│ QQuery 7  │         14.81 / 15.00 ±0.13 / 15.17 ms │         15.05 / 15.08 ±0.03 / 15.12 ms │     no change │
│ QQuery 8  │      323.15 / 327.08 ±2.82 / 330.37 ms │      336.23 / 339.40 ±1.61 / 340.59 ms │     no change │
│ QQuery 9  │      455.17 / 463.01 ±8.29 / 476.14 ms │      416.87 / 429.72 ±9.62 / 440.46 ms │ +1.08x faster │
│ QQuery 10 │         71.32 / 72.61 ±1.04 / 74.08 ms │         70.07 / 73.61 ±4.10 / 81.36 ms │     no change │
│ QQuery 11 │         84.67 / 85.01 ±0.20 / 85.22 ms │         81.59 / 82.32 ±0.76 / 83.73 ms │     no change │
│ QQuery 12 │      269.38 / 271.85 ±3.08 / 277.40 ms │      269.05 / 277.34 ±6.72 / 286.76 ms │     no change │
│ QQuery 13 │     368.10 / 379.56 ±10.85 / 397.61 ms │     371.39 / 392.02 ±20.27 / 426.38 ms │     no change │
│ QQuery 14 │     285.82 / 295.92 ±11.61 / 318.37 ms │      279.05 / 285.69 ±5.24 / 294.34 ms │     no change │
│ QQuery 15 │     272.73 / 314.52 ±25.47 / 352.47 ms │     318.51 / 347.36 ±24.76 / 379.93 ms │  1.10x slower │
│ QQuery 16 │      611.86 / 624.88 ±8.71 / 636.44 ms │     619.21 / 646.77 ±33.10 / 706.71 ms │     no change │
│ QQuery 17 │      620.36 / 637.45 ±9.12 / 646.90 ms │      621.42 / 626.93 ±5.74 / 637.85 ms │     no change │
│ QQuery 18 │   1295.80 / 1310.26 ±8.42 / 1318.80 ms │  1245.60 / 1264.35 ±16.37 / 1287.06 ms │     no change │
│ QQuery 19 │         28.94 / 29.39 ±0.41 / 29.89 ms │         28.97 / 29.64 ±0.74 / 30.73 ms │     no change │
│ QQuery 20 │     518.82 / 550.00 ±40.15 / 628.88 ms │     517.22 / 535.91 ±10.52 / 547.09 ms │     no change │
│ QQuery 21 │     516.29 / 537.30 ±15.89 / 555.97 ms │      517.53 / 524.66 ±6.57 / 537.01 ms │     no change │
│ QQuery 22 │   1059.35 / 1070.83 ±8.53 / 1081.43 ms │    986.43 / 999.32 ±10.55 / 1015.39 ms │ +1.07x faster │
│ QQuery 23 │  3115.86 / 3175.04 ±35.21 / 3213.68 ms │ 3071.79 / 3199.58 ±137.62 / 3393.58 ms │     no change │
│ QQuery 24 │         42.39 / 49.05 ±9.18 / 66.75 ms │        43.71 / 49.57 ±11.32 / 72.22 ms │     no change │
│ QQuery 25 │      115.74 / 119.33 ±2.87 / 123.03 ms │      119.49 / 126.46 ±6.35 / 136.24 ms │  1.06x slower │
│ QQuery 26 │         41.89 / 43.77 ±1.17 / 45.55 ms │         44.27 / 45.09 ±0.73 / 46.40 ms │     no change │
│ QQuery 27 │      668.75 / 676.91 ±7.49 / 690.79 ms │     675.27 / 689.39 ±13.34 / 710.41 ms │     no change │
│ QQuery 28 │  3050.86 / 3065.25 ±11.01 / 3084.19 ms │  3017.42 / 3060.28 ±38.20 / 3107.57 ms │     no change │
│ QQuery 29 │        43.06 / 49.97 ±13.56 / 77.09 ms │         41.58 / 44.69 ±5.97 / 56.64 ms │ +1.12x faster │
│ QQuery 30 │     304.22 / 319.23 ±15.78 / 347.03 ms │     319.65 / 336.27 ±10.64 / 348.85 ms │  1.05x slower │
│ QQuery 31 │      287.95 / 294.90 ±7.16 / 307.82 ms │     308.73 / 325.62 ±17.41 / 351.45 ms │  1.10x slower │
│ QQuery 32 │    963.54 / 998.41 ±34.38 / 1057.28 ms │  1075.73 / 1100.36 ±18.70 / 1121.98 ms │  1.10x slower │
│ QQuery 33 │  1463.89 / 1488.51 ±23.77 / 1533.33 ms │  1491.50 / 1574.26 ±98.02 / 1762.13 ms │  1.06x slower │
│ QQuery 34 │ 1465.98 / 1676.91 ±124.63 / 1786.89 ms │  1522.75 / 1569.32 ±46.39 / 1643.23 ms │ +1.07x faster │
│ QQuery 35 │     352.90 / 372.52 ±29.97 / 432.15 ms │     297.07 / 323.25 ±27.29 / 360.53 ms │ +1.15x faster │
│ QQuery 36 │         67.38 / 69.82 ±2.01 / 72.73 ms │         61.51 / 68.81 ±8.02 / 84.44 ms │     no change │
│ QQuery 37 │         36.18 / 44.77 ±6.68 / 52.47 ms │         35.64 / 36.66 ±0.90 / 37.69 ms │ +1.22x faster │
│ QQuery 38 │         42.93 / 47.71 ±3.42 / 51.31 ms │         43.21 / 48.46 ±5.89 / 58.82 ms │     no change │
│ QQuery 39 │     138.15 / 151.88 ±10.35 / 168.57 ms │      132.74 / 135.80 ±4.19 / 143.97 ms │ +1.12x faster │
│ QQuery 40 │         14.28 / 14.63 ±0.31 / 15.10 ms │         14.10 / 14.48 ±0.43 / 15.28 ms │     no change │
│ QQuery 41 │         13.92 / 18.42 ±5.14 / 26.89 ms │         13.84 / 15.84 ±3.36 / 22.53 ms │ +1.16x faster │
│ QQuery 42 │         13.35 / 16.51 ±3.44 / 22.02 ms │         13.97 / 15.52 ±2.68 / 20.87 ms │ +1.06x faster │
└───────────┴────────────────────────────────────────┴────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 20279.51ms │
│ Total Time (reuse-hash-through-hold-aggr)   │ 20296.98ms │
│ Average Time (HEAD)                         │   471.62ms │
│ Average Time (reuse-hash-through-hold-aggr) │   472.02ms │
│ Queries Faster                              │          9 │
│ Queries Slower                              │          8 │
│ Queries with No Change                      │         26 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Resource Usage

clickbench_partitioned — base (merge-base)

Metric Value
Wall time 105.0s
Peak memory 12.2 GiB
Avg memory 4.5 GiB
CPU user 1040.0s
CPU sys 73.6s
Peak spill 0 B

clickbench_partitioned — branch

Metric Value
Wall time 105.0s
Peak memory 12.4 GiB
Avg memory 4.7 GiB
CPU user 1054.6s
CPU sys 74.7s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

adriangbot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Instance: c4a-highmem-16 (12 vCPU / 65 GiB)

CPU Details (lscpu) 
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 HEAD and reuse-hash-through-hold-aggr
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃          reuse-hash-through-hold-aggr ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │          1.55 / 4.66 ±5.84 / 16.34 ms │          1.35 / 4.37 ±5.85 / 16.07 ms │ +1.07x faster │
│ QQuery 1  │        13.42 / 13.58 ±0.10 / 13.70 ms │        13.69 / 14.04 ±0.35 / 14.61 ms │     no change │
│ QQuery 2  │        37.30 / 37.58 ±0.15 / 37.71 ms │        36.48 / 37.17 ±0.70 / 38.34 ms │     no change │
│ QQuery 3  │        32.57 / 33.16 ±0.71 / 34.52 ms │        32.32 / 33.24 ±0.60 / 34.01 ms │     no change │
│ QQuery 4  │     248.87 / 258.73 ±6.38 / 267.99 ms │    347.41 / 358.72 ±11.37 / 375.84 ms │  1.39x slower │
│ QQuery 5  │     312.69 / 325.15 ±8.23 / 334.82 ms │    295.92 / 312.59 ±13.93 / 333.67 ms │     no change │
│ QQuery 6  │           1.62 / 1.80 ±0.21 / 2.20 ms │           1.29 / 1.45 ±0.25 / 1.93 ms │ +1.24x faster │
│ QQuery 7  │        16.11 / 16.23 ±0.10 / 16.42 ms │        14.83 / 14.92 ±0.08 / 15.05 ms │ +1.09x faster │
│ QQuery 8  │     392.29 / 403.71 ±8.27 / 416.45 ms │     344.42 / 357.55 ±8.72 / 371.47 ms │ +1.13x faster │
│ QQuery 9  │    481.85 / 560.06 ±53.04 / 605.45 ms │    434.94 / 461.59 ±19.03 / 483.95 ms │ +1.21x faster │
│ QQuery 10 │        73.72 / 76.93 ±2.63 / 81.46 ms │        72.55 / 73.18 ±0.71 / 74.46 ms │     no change │
│ QQuery 11 │        84.60 / 86.14 ±1.38 / 88.17 ms │        83.94 / 86.27 ±2.29 / 89.90 ms │     no change │
│ QQuery 12 │     289.77 / 298.67 ±7.22 / 311.21 ms │     287.16 / 293.39 ±6.20 / 303.34 ms │     no change │
│ QQuery 13 │    378.75 / 400.78 ±13.90 / 420.14 ms │    401.36 / 445.08 ±26.50 / 471.09 ms │  1.11x slower │
│ QQuery 14 │     294.73 / 297.35 ±2.24 / 300.73 ms │    312.59 / 328.32 ±16.20 / 359.41 ms │  1.10x slower │
│ QQuery 15 │    290.76 / 307.76 ±13.16 / 326.22 ms │     376.43 / 390.19 ±8.24 / 398.24 ms │  1.27x slower │
│ QQuery 16 │     650.60 / 661.85 ±6.98 / 672.13 ms │    635.55 / 678.39 ±26.02 / 709.39 ms │     no change │
│ QQuery 17 │    673.03 / 702.44 ±16.93 / 724.01 ms │    640.27 / 675.20 ±21.22 / 703.24 ms │     no change │
│ QQuery 18 │ 1361.09 / 1396.23 ±26.39 / 1434.36 ms │ 1405.47 / 1442.89 ±47.72 / 1525.35 ms │     no change │
│ QQuery 19 │        29.44 / 29.79 ±0.20 / 30.01 ms │        29.10 / 31.03 ±2.48 / 35.74 ms │     no change │
│ QQuery 20 │    521.71 / 540.69 ±13.75 / 563.80 ms │     519.46 / 527.46 ±6.86 / 539.60 ms │     no change │
│ QQuery 21 │     529.43 / 539.43 ±8.47 / 552.01 ms │    527.18 / 547.58 ±13.63 / 567.81 ms │     no change │
│ QQuery 22 │ 1006.00 / 1019.95 ±11.11 / 1039.13 ms │ 1070.29 / 1085.78 ±11.80 / 1105.88 ms │  1.06x slower │
│ QQuery 23 │ 3218.47 / 3250.83 ±35.70 / 3317.11 ms │ 3225.26 / 3351.60 ±92.66 / 3484.31 ms │     no change │
│ QQuery 24 │        42.18 / 43.31 ±0.93 / 45.01 ms │        44.04 / 56.31 ±9.40 / 65.87 ms │  1.30x slower │
│ QQuery 25 │     111.78 / 118.32 ±8.79 / 135.52 ms │     115.92 / 126.14 ±7.81 / 140.10 ms │  1.07x slower │
│ QQuery 26 │        42.15 / 43.89 ±1.68 / 46.71 ms │        44.98 / 47.49 ±3.41 / 54.16 ms │  1.08x slower │
│ QQuery 27 │     672.48 / 687.14 ±9.94 / 700.83 ms │     692.08 / 698.87 ±8.20 / 712.78 ms │     no change │
│ QQuery 28 │ 3052.32 / 3154.91 ±57.33 / 3223.32 ms │ 3056.05 / 3098.91 ±36.68 / 3154.98 ms │     no change │
│ QQuery 29 │        41.71 / 42.22 ±0.38 / 42.71 ms │       42.00 / 56.56 ±17.44 / 89.80 ms │  1.34x slower │
│ QQuery 30 │    312.82 / 323.31 ±11.01 / 343.13 ms │     315.04 / 324.84 ±6.65 / 332.02 ms │     no change │
│ QQuery 31 │    299.78 / 322.40 ±17.09 / 348.50 ms │    311.66 / 334.77 ±21.97 / 367.98 ms │     no change │
│ QQuery 32 │  996.69 / 1060.11 ±42.77 / 1116.24 ms │ 1145.30 / 1176.00 ±35.10 / 1241.84 ms │  1.11x slower │
│ QQuery 33 │ 1547.45 / 1639.72 ±57.76 / 1720.67 ms │ 1521.40 / 1533.21 ±10.36 / 1551.72 ms │ +1.07x faster │
│ QQuery 34 │ 1528.00 / 1568.38 ±48.62 / 1662.00 ms │ 1643.23 / 1687.42 ±50.51 / 1781.83 ms │  1.08x slower │
│ QQuery 35 │    315.53 / 369.22 ±49.71 / 447.55 ms │    326.67 / 362.81 ±28.70 / 409.95 ms │     no change │
│ QQuery 36 │        72.22 / 78.25 ±4.58 / 85.99 ms │        67.64 / 72.25 ±6.07 / 84.11 ms │ +1.08x faster │
│ QQuery 37 │        37.38 / 40.50 ±3.24 / 45.22 ms │        37.09 / 37.34 ±0.21 / 37.69 ms │ +1.08x faster │
│ QQuery 38 │        41.91 / 44.52 ±2.48 / 48.29 ms │        40.49 / 43.99 ±3.10 / 48.51 ms │     no change │
│ QQuery 39 │    151.40 / 164.06 ±11.10 / 183.24 ms │     158.60 / 163.47 ±3.61 / 167.93 ms │     no change │
│ QQuery 40 │        15.04 / 15.83 ±0.57 / 16.55 ms │        15.53 / 16.20 ±0.72 / 17.29 ms │     no change │
│ QQuery 41 │        14.54 / 15.13 ±0.33 / 15.46 ms │        15.29 / 15.71 ±0.49 / 16.63 ms │     no change │
│ QQuery 42 │        14.51 / 20.53 ±6.90 / 29.67 ms │        14.77 / 14.98 ±0.21 / 15.35 ms │ +1.37x faster │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 21015.25ms │
│ Total Time (reuse-hash-through-hold-aggr)   │ 21419.26ms │
│ Average Time (HEAD)                         │   488.73ms │
│ Average Time (reuse-hash-through-hold-aggr) │   498.12ms │
│ Queries Faster                              │          9 │
│ Queries Slower                              │         11 │
│ Queries with No Change                      │         23 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Resource Usage

clickbench_partitioned — base (merge-base)

Metric Value
Wall time 110.0s
Peak memory 12.5 GiB
Avg memory 4.4 GiB
CPU user 1076.7s
CPU sys 76.7s
Peak spill 0 B

clickbench_partitioned — branch

Metric Value
Wall time 110.0s
Peak memory 12.5 GiB
Avg memory 4.6 GiB
CPU user 1111.7s
CPU sys 82.7s
Peak spill 0 B

File an issue against this benchmark runner

@Rachelint

Rachelint commented Jul 10, 2026

Copy link
Copy Markdown
Contributor Author

run benchmarks clickbench_partitioned

@adriangbot

adriangbot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4936468797-959-d22h8 6.12.85+ #1 SMP Mon May 11 08:17:35 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu) 
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

Comparing reuse-hash-through-hold-aggr (82bdb2b) to f34a676 (merge-base) diff using: clickbench_partitioned
Results will be posted here when complete

File an issue against this benchmark runner

@adriangbot

adriangbot commented Jul 10, 2026

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Instance: c4a-highmem-16 (12 vCPU / 65 GiB)

CPU Details (lscpu) 
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 HEAD and reuse-hash-through-hold-aggr
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃          reuse-hash-through-hold-aggr ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │          1.23 / 3.94 ±5.37 / 14.68 ms │          1.21 / 3.92 ±5.35 / 14.62 ms │     no change │
│ QQuery 1  │        12.51 / 12.82 ±0.23 / 13.19 ms │        13.21 / 13.57 ±0.21 / 13.80 ms │  1.06x slower │
│ QQuery 2  │        36.83 / 37.34 ±0.57 / 38.43 ms │        36.24 / 36.60 ±0.33 / 37.00 ms │     no change │
│ QQuery 3  │        30.93 / 31.85 ±0.95 / 33.63 ms │        30.79 / 31.11 ±0.24 / 31.44 ms │     no change │
│ QQuery 4  │     221.36 / 222.84 ±2.29 / 227.37 ms │    265.32 / 278.63 ±22.41 / 323.40 ms │  1.25x slower │
│ QQuery 5  │     272.20 / 273.47 ±1.41 / 276.10 ms │     313.17 / 316.27 ±2.05 / 319.28 ms │  1.16x slower │
│ QQuery 6  │           1.27 / 1.42 ±0.23 / 1.88 ms │           1.28 / 1.53 ±0.40 / 2.33 ms │  1.08x slower │
│ QQuery 7  │        13.70 / 13.81 ±0.08 / 13.94 ms │        14.47 / 14.61 ±0.08 / 14.68 ms │  1.06x slower │
│ QQuery 8  │     317.64 / 320.36 ±2.54 / 324.98 ms │    338.99 / 375.78 ±23.15 / 398.14 ms │  1.17x slower │
│ QQuery 9  │     449.96 / 458.05 ±5.05 / 465.37 ms │     417.44 / 428.70 ±8.78 / 441.34 ms │ +1.07x faster │
│ QQuery 10 │        70.87 / 71.96 ±0.82 / 73.13 ms │        68.65 / 70.59 ±2.68 / 75.89 ms │     no change │
│ QQuery 11 │        83.44 / 84.31 ±0.68 / 85.18 ms │        79.93 / 83.68 ±6.57 / 96.79 ms │     no change │
│ QQuery 12 │     263.21 / 270.06 ±4.61 / 276.13 ms │     265.60 / 274.03 ±6.69 / 283.47 ms │     no change │
│ QQuery 13 │    353.76 / 381.69 ±21.06 / 412.91 ms │     368.70 / 375.13 ±6.01 / 385.39 ms │     no change │
│ QQuery 14 │     279.06 / 283.91 ±7.09 / 297.96 ms │     278.16 / 282.46 ±3.36 / 286.64 ms │     no change │
│ QQuery 15 │     268.19 / 273.11 ±2.55 / 275.18 ms │     309.33 / 320.41 ±8.10 / 331.09 ms │  1.17x slower │
│ QQuery 16 │     615.17 / 621.44 ±4.17 / 628.23 ms │    615.01 / 625.65 ±11.65 / 648.25 ms │     no change │
│ QQuery 17 │     611.55 / 618.98 ±5.24 / 623.59 ms │     615.67 / 626.18 ±7.04 / 635.46 ms │     no change │
│ QQuery 18 │ 1251.75 / 1277.98 ±18.33 / 1300.55 ms │ 1231.37 / 1253.87 ±13.75 / 1268.61 ms │     no change │
│ QQuery 19 │        27.65 / 29.90 ±3.94 / 37.77 ms │        27.87 / 28.30 ±0.43 / 29.07 ms │ +1.06x faster │
│ QQuery 20 │     514.96 / 521.69 ±6.09 / 531.85 ms │    516.81 / 531.73 ±18.40 / 566.39 ms │     no change │
│ QQuery 21 │    515.04 / 521.76 ±10.73 / 543.14 ms │     513.01 / 516.83 ±3.03 / 522.24 ms │     no change │
│ QQuery 22 │    983.18 / 991.08 ±6.76 / 1001.48 ms │     976.75 / 989.20 ±6.76 / 995.64 ms │     no change │
│ QQuery 23 │ 3103.28 / 3126.30 ±14.78 / 3143.25 ms │ 3022.12 / 3043.77 ±13.91 / 3063.95 ms │     no change │
│ QQuery 24 │        41.38 / 42.78 ±0.95 / 44.33 ms │        40.75 / 46.13 ±6.57 / 56.37 ms │  1.08x slower │
│ QQuery 25 │     113.11 / 114.12 ±1.02 / 116.00 ms │     111.62 / 115.91 ±4.38 / 122.70 ms │     no change │
│ QQuery 26 │        41.60 / 42.63 ±1.09 / 44.72 ms │        41.66 / 42.50 ±1.15 / 44.77 ms │     no change │
│ QQuery 27 │     668.72 / 678.85 ±9.34 / 694.75 ms │     669.36 / 680.50 ±8.06 / 692.04 ms │     no change │
│ QQuery 28 │ 2998.12 / 3016.99 ±11.30 / 3033.14 ms │  2990.85 / 2997.59 ±5.84 / 3007.88 ms │     no change │
│ QQuery 29 │       40.73 / 50.40 ±12.56 / 72.24 ms │       40.66 / 49.47 ±12.44 / 73.03 ms │     no change │
│ QQuery 30 │     302.06 / 307.28 ±3.60 / 312.87 ms │     307.44 / 311.50 ±4.07 / 318.78 ms │     no change │
│ QQuery 31 │    288.36 / 302.28 ±14.04 / 327.38 ms │    302.32 / 311.83 ±10.26 / 331.15 ms │     no change │
│ QQuery 32 │   956.87 / 982.30 ±23.49 / 1025.19 ms │ 1042.16 / 1060.99 ±12.04 / 1074.80 ms │  1.08x slower │
│ QQuery 33 │ 1454.97 / 1466.92 ±14.01 / 1493.83 ms │ 1481.37 / 1519.26 ±30.50 / 1569.91 ms │     no change │
│ QQuery 34 │ 1451.04 / 1469.50 ±14.01 / 1484.15 ms │ 1504.49 / 1529.12 ±23.25 / 1565.56 ms │     no change │
│ QQuery 35 │    268.52 / 309.93 ±44.75 / 376.33 ms │    295.44 / 332.08 ±42.77 / 405.41 ms │  1.07x slower │
│ QQuery 36 │        71.59 / 74.31 ±1.66 / 76.29 ms │        69.02 / 72.02 ±4.02 / 79.89 ms │     no change │
│ QQuery 37 │        35.54 / 40.09 ±7.40 / 54.79 ms │        35.06 / 35.76 ±0.56 / 36.40 ms │ +1.12x faster │
│ QQuery 38 │        39.98 / 41.37 ±1.34 / 43.75 ms │        43.78 / 46.70 ±2.21 / 49.51 ms │  1.13x slower │
│ QQuery 39 │     138.16 / 152.28 ±8.45 / 161.04 ms │     127.12 / 132.69 ±5.88 / 143.36 ms │ +1.15x faster │
│ QQuery 40 │        14.34 / 14.67 ±0.31 / 15.16 ms │        13.79 / 14.16 ±0.27 / 14.54 ms │     no change │
│ QQuery 41 │        13.54 / 13.94 ±0.30 / 14.31 ms │        13.43 / 14.31 ±1.29 / 16.89 ms │     no change │
│ QQuery 42 │        13.29 / 13.52 ±0.15 / 13.75 ms │        13.58 / 15.71 ±3.52 / 22.69 ms │  1.16x slower │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 19584.21ms │
│ Total Time (reuse-hash-through-hold-aggr)   │ 19850.80ms │
│ Average Time (HEAD)                         │   455.45ms │
│ Average Time (reuse-hash-through-hold-aggr) │   461.65ms │
│ Queries Faster                              │          4 │
│ Queries Slower                              │         12 │
│ Queries with No Change                      │         27 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Resource Usage

clickbench_partitioned — base (merge-base)

Metric Value
Wall time 100.0s
Peak memory 12.9 GiB
Avg memory 4.7 GiB
CPU user 1012.0s
CPU sys 66.3s
Peak spill 0 B

clickbench_partitioned — branch

Metric Value
Wall time 100.0s
Peak memory 12.6 GiB
Avg memory 4.5 GiB
CPU user 1037.3s
CPU sys 70.4s
Peak spill 0 B

File an issue against this benchmark runner

@Rachelint Rachelint force-pushed the reuse-hash-through-hold-aggr branch from 82bdb2b to 0200925 Compare July 11, 2026 10:21
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

physical-plan Changes to the physical-plan crate

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@Rachelint @adriangbot