perf(sort): fast path for full batch when it sorts before all other streams in `SortPreservingMergeStream` by rluvaton · Pull Request #23345 · apache/datafusion

rluvaton · 2026-07-06T17:37:02Z

Comparing HEAD and fast-batch-wise-skip
--------------------
Benchmark sort_tpch10.json
--------------------
┏━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃                                   HEAD ┃                   fast-batch-wise-skip ┃        Change ┃
┡━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ Q2    │  1421.45 / 1431.48 ±10.65 / 1451.01 ms │      201.29 / 208.62 ±6.77 / 217.57 ms │ +6.86x faster │
│ Q4    │  2741.65 / 2770.88 ±16.58 / 2789.94 ms │     392.40 / 423.10 ±20.71 / 453.79 ms │ +6.55x faster │

Which issue does this PR close?

N/A

Rationale for this change

this is part of an effort I'm doing to make sort much much faster.

here I'm making it faster when the data is somewhat sorted

in this there are 2 things that we save on in case we get to the fast path:

Not dealing with the the loser tree
Skip row by row comparison
interleave and data copy - as we output the same batch as input

What changes are included in this PR?

fast skip when can emit entire batch

Are these changes tested?

yes and existing tests

Are there any user-facing changes?

Not API changes but behavioral changes regarding the size of the output batches:

sort may not respect batch_size for already sorted input - it will return the same size as the input stream
sort may produce small batches between full sorted batches
e.g. we have 3 streams and batch size is 4:

s1: 1, 10, 12, 13
s2: 3, 4, 5, 6
s3: 2, 9, 11, 11

the output will be:

batch 1: 1, 2
batch 2: 3, 4, 5, 6
batch 3: 9, 10, 11, 11
batch 4: 12, 13

TODO

Need to add benchmark
- for already sorted
- for random
- for reverse ordering

…treams

rluvaton · 2026-07-06T17:37:21Z

run benchmarks

adriangbot · 2026-07-06T17:39:37Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4895826694-867-l8b6b 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 fast-batch-wise-skip (f3cffeb) to 6a0e76e (merge-base) diff using: tpcds
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-07-06T17:40:05Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4895826694-866-fn2jv 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 fast-batch-wise-skip (f3cffeb) to 6a0e76e (merge-base) diff using: clickbench_partitioned
Results will be posted here when complete

File an issue against this benchmark runner

rluvaton · 2026-07-06T17:40:26Z

run benchmark sort_preserving_merge

adriangbot · 2026-07-06T17:40:28Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4895826694-868-v64xg 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 fast-batch-wise-skip (f3cffeb) to 6a0e76e (merge-base) diff using: tpch
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-07-06T17:43:18Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4895853179-869-8t7j4 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 fast-batch-wise-skip (f3cffeb) to 6a0e76e (merge-base) diff using: sort_preserving_merge
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-07-06T17:53:28Z

🤖 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 fast-batch-wise-skip
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃           fast-batch-wise-skip ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 1  │ 38.31 / 38.95 ±0.94 / 40.82 ms │ 38.31 / 39.26 ±1.15 / 41.12 ms │    no change │
│ QQuery 2  │ 19.10 / 19.59 ±0.35 / 20.14 ms │ 19.05 / 19.40 ±0.24 / 19.79 ms │    no change │
│ QQuery 3  │ 31.43 / 33.05 ±2.10 / 36.86 ms │ 30.84 / 34.16 ±2.84 / 39.23 ms │    no change │
│ QQuery 4  │ 17.58 / 17.81 ±0.14 / 17.97 ms │ 17.68 / 18.60 ±0.66 / 19.45 ms │    no change │
│ QQuery 5  │ 38.59 / 39.27 ±0.78 / 40.77 ms │ 38.96 / 40.24 ±1.06 / 41.38 ms │    no change │
│ QQuery 6  │ 16.32 / 16.46 ±0.14 / 16.66 ms │ 16.28 / 17.51 ±1.45 / 19.89 ms │ 1.06x slower │
│ QQuery 7  │ 43.46 / 45.02 ±1.13 / 46.52 ms │ 45.18 / 48.11 ±2.88 / 53.10 ms │ 1.07x slower │
│ QQuery 8  │ 43.64 / 43.95 ±0.38 / 44.61 ms │ 43.55 / 43.68 ±0.11 / 43.86 ms │    no change │
│ QQuery 9  │ 50.00 / 51.20 ±0.90 / 52.74 ms │ 49.92 / 50.72 ±0.75 / 51.80 ms │    no change │
│ QQuery 10 │ 42.44 / 42.63 ±0.16 / 42.93 ms │ 42.67 / 42.89 ±0.15 / 43.06 ms │    no change │
│ QQuery 11 │ 13.55 / 13.72 ±0.20 / 14.06 ms │ 13.55 / 13.83 ±0.17 / 14.03 ms │    no change │
│ QQuery 12 │ 24.12 / 24.72 ±0.65 / 25.90 ms │ 24.27 / 25.19 ±0.94 / 26.92 ms │    no change │
│ QQuery 13 │ 32.22 / 33.75 ±1.34 / 36.04 ms │ 32.20 / 35.27 ±2.54 / 39.37 ms │    no change │
│ QQuery 14 │ 23.77 / 24.03 ±0.16 / 24.22 ms │ 23.62 / 23.97 ±0.18 / 24.13 ms │    no change │
│ QQuery 15 │ 31.39 / 31.53 ±0.16 / 31.85 ms │ 31.16 / 32.00 ±0.69 / 32.85 ms │    no change │
│ QQuery 16 │ 13.97 / 14.16 ±0.11 / 14.29 ms │ 14.05 / 14.16 ±0.09 / 14.25 ms │    no change │
│ QQuery 17 │ 73.62 / 75.02 ±1.55 / 77.25 ms │ 74.47 / 75.30 ±1.16 / 77.60 ms │    no change │
│ QQuery 18 │ 58.95 / 59.84 ±0.88 / 61.44 ms │ 57.16 / 60.18 ±1.54 / 61.37 ms │    no change │
│ QQuery 19 │ 33.58 / 34.70 ±1.24 / 36.64 ms │ 33.06 / 33.67 ±0.92 / 35.47 ms │    no change │
│ QQuery 20 │ 32.19 / 32.48 ±0.20 / 32.78 ms │ 31.93 / 32.49 ±0.45 / 33.05 ms │    no change │
│ QQuery 21 │ 54.79 / 56.62 ±1.93 / 60.04 ms │ 56.11 / 57.62 ±2.11 / 61.82 ms │    no change │
│ QQuery 22 │ 13.98 / 14.47 ±0.55 / 15.16 ms │ 13.92 / 14.20 ±0.17 / 14.41 ms │    no change │
└───────────┴────────────────────────────────┴────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                   ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                   │ 762.96ms │
│ Total Time (fast-batch-wise-skip)   │ 772.45ms │
│ Average Time (HEAD)                 │  34.68ms │
│ Average Time (fast-batch-wise-skip) │  35.11ms │
│ Queries Faster                      │        0 │
│ Queries Slower                      │        2 │
│ Queries with No Change              │       20 │
│ Queries with Failure                │        0 │
└─────────────────────────────────────┴──────────┘

Resource Usage

tpch — base (merge-base)

Metric	Value
Wall time	5.0s
Peak memory	1.1 GiB
Avg memory	496.1 MiB
CPU user	21.9s
CPU sys	1.7s
Peak spill	0 B

tpch — branch

Metric	Value
Wall time	5.0s
Peak memory	1.2 GiB
Avg memory	519.5 MiB
CPU user	22.4s
CPU sys	1.7s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-07-06T17:55:45Z

🤖 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 fast-batch-wise-skip
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                   HEAD ┃                  fast-batch-wise-skip ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │            5.81 / 6.25 ±0.80 / 7.86 ms │           5.62 / 6.13 ±0.88 / 7.89 ms │     no change │
│ QQuery 2  │         80.56 / 80.77 ±0.22 / 81.15 ms │        81.03 / 81.82 ±1.01 / 83.73 ms │     no change │
│ QQuery 3  │         29.92 / 30.08 ±0.15 / 30.32 ms │        31.04 / 31.54 ±0.26 / 31.78 ms │     no change │
│ QQuery 4  │      481.06 / 494.06 ±7.09 / 500.82 ms │    538.73 / 569.12 ±23.25 / 595.98 ms │  1.15x slower │
│ QQuery 5  │         52.08 / 52.48 ±0.36 / 53.11 ms │        53.70 / 53.82 ±0.07 / 53.90 ms │     no change │
│ QQuery 6  │         37.14 / 37.36 ±0.14 / 37.57 ms │        38.13 / 38.48 ±0.28 / 38.79 ms │     no change │
│ QQuery 7  │         94.84 / 95.50 ±0.45 / 96.00 ms │      95.48 / 100.20 ±3.49 / 105.23 ms │     no change │
│ QQuery 8  │         37.06 / 37.55 ±0.38 / 38.08 ms │        36.55 / 37.25 ±0.39 / 37.68 ms │     no change │
│ QQuery 9  │         53.30 / 54.88 ±1.31 / 56.36 ms │        54.00 / 54.99 ±0.83 / 56.26 ms │     no change │
│ QQuery 10 │         63.99 / 64.78 ±0.72 / 66.07 ms │        63.81 / 65.71 ±2.45 / 70.54 ms │     no change │
│ QQuery 11 │      300.79 / 304.59 ±2.95 / 309.30 ms │     307.19 / 311.36 ±3.50 / 316.60 ms │     no change │
│ QQuery 12 │         28.84 / 29.30 ±0.36 / 29.73 ms │        28.91 / 29.06 ±0.12 / 29.26 ms │     no change │
│ QQuery 13 │      118.53 / 121.45 ±4.36 / 130.05 ms │     119.15 / 120.51 ±1.20 / 122.65 ms │     no change │
│ QQuery 14 │      431.47 / 435.81 ±3.99 / 442.30 ms │     415.21 / 419.69 ±2.40 / 421.96 ms │     no change │
│ QQuery 15 │         64.49 / 66.05 ±1.40 / 68.69 ms │        57.44 / 58.38 ±0.57 / 59.14 ms │ +1.13x faster │
│ QQuery 16 │            7.36 / 7.46 ±0.08 / 7.60 ms │           6.91 / 7.08 ±0.26 / 7.59 ms │ +1.05x faster │
│ QQuery 17 │         81.89 / 84.03 ±1.75 / 87.22 ms │        80.99 / 83.86 ±2.42 / 88.08 ms │     no change │
│ QQuery 18 │      123.93 / 124.96 ±0.96 / 126.68 ms │     123.48 / 125.73 ±1.62 / 127.64 ms │     no change │
│ QQuery 19 │         41.61 / 42.65 ±1.19 / 44.91 ms │        42.08 / 42.72 ±0.86 / 44.39 ms │     no change │
│ QQuery 20 │         35.69 / 36.60 ±0.70 / 37.52 ms │        36.28 / 37.97 ±1.44 / 40.66 ms │     no change │
│ QQuery 21 │         17.51 / 17.82 ±0.29 / 18.24 ms │        17.91 / 18.03 ±0.10 / 18.19 ms │     no change │
│ QQuery 22 │         61.80 / 63.29 ±1.36 / 65.68 ms │        63.07 / 63.61 ±0.46 / 64.41 ms │     no change │
│ QQuery 23 │      348.82 / 353.45 ±4.00 / 358.68 ms │    346.86 / 389.18 ±25.46 / 416.99 ms │  1.10x slower │
│ QQuery 24 │      227.47 / 229.57 ±2.32 / 233.71 ms │     229.68 / 241.61 ±8.90 / 253.87 ms │  1.05x slower │
│ QQuery 25 │      109.83 / 112.41 ±1.52 / 114.10 ms │     113.70 / 116.78 ±2.25 / 120.07 ms │     no change │
│ QQuery 26 │         57.75 / 58.59 ±0.59 / 59.33 ms │        59.75 / 60.25 ±0.36 / 60.72 ms │     no change │
│ QQuery 27 │            6.48 / 6.66 ±0.17 / 6.99 ms │           6.54 / 6.69 ±0.16 / 6.99 ms │     no change │
│ QQuery 28 │         61.02 / 61.96 ±0.60 / 62.59 ms │        61.13 / 62.04 ±0.55 / 62.79 ms │     no change │
│ QQuery 29 │       97.51 / 101.56 ±4.41 / 109.66 ms │      98.40 / 102.52 ±5.62 / 113.39 ms │     no change │
│ QQuery 30 │         32.78 / 33.19 ±0.45 / 33.96 ms │        31.98 / 32.46 ±0.37 / 33.12 ms │     no change │
│ QQuery 31 │      111.97 / 112.58 ±0.58 / 113.69 ms │     111.90 / 114.66 ±2.61 / 118.92 ms │     no change │
│ QQuery 32 │         20.45 / 22.25 ±3.05 / 28.33 ms │        21.25 / 21.68 ±0.26 / 21.95 ms │     no change │
│ QQuery 33 │         38.45 / 38.87 ±0.27 / 39.20 ms │        38.42 / 38.67 ±0.27 / 39.19 ms │     no change │
│ QQuery 34 │          9.78 / 10.04 ±0.40 / 10.84 ms │        10.04 / 10.38 ±0.51 / 11.39 ms │     no change │
│ QQuery 35 │         73.35 / 73.96 ±0.56 / 74.84 ms │        72.92 / 73.55 ±0.39 / 74.09 ms │     no change │
│ QQuery 36 │            5.99 / 6.13 ±0.19 / 6.52 ms │           6.09 / 6.21 ±0.15 / 6.50 ms │     no change │
│ QQuery 37 │            7.22 / 7.32 ±0.15 / 7.61 ms │           7.09 / 7.26 ±0.11 / 7.42 ms │     no change │
│ QQuery 38 │         61.75 / 62.98 ±1.21 / 65.09 ms │        62.88 / 64.36 ±1.99 / 68.31 ms │     no change │
│ QQuery 39 │         90.42 / 91.32 ±0.71 / 92.54 ms │        90.75 / 91.56 ±0.65 / 92.73 ms │     no change │
│ QQuery 40 │         24.05 / 24.21 ±0.13 / 24.41 ms │        23.54 / 23.91 ±0.29 / 24.34 ms │     no change │
│ QQuery 41 │         11.71 / 11.94 ±0.15 / 12.14 ms │        11.52 / 11.76 ±0.17 / 12.02 ms │     no change │
│ QQuery 42 │         23.86 / 24.39 ±0.36 / 24.91 ms │        24.00 / 24.30 ±0.21 / 24.57 ms │     no change │
│ QQuery 43 │            4.98 / 5.07 ±0.08 / 5.21 ms │           5.12 / 5.20 ±0.12 / 5.45 ms │     no change │
│ QQuery 44 │            9.22 / 9.34 ±0.13 / 9.57 ms │         9.63 / 10.98 ±2.56 / 16.09 ms │  1.18x slower │
│ QQuery 45 │         37.67 / 40.03 ±2.89 / 45.59 ms │        40.52 / 41.03 ±0.38 / 41.42 ms │     no change │
│ QQuery 46 │         12.68 / 12.91 ±0.26 / 13.30 ms │        11.99 / 12.48 ±0.32 / 12.86 ms │     no change │
│ QQuery 47 │      228.57 / 231.67 ±2.14 / 234.84 ms │     229.16 / 234.02 ±4.51 / 240.95 ms │     no change │
│ QQuery 48 │         96.75 / 97.72 ±1.02 / 99.63 ms │        96.31 / 97.29 ±0.66 / 98.27 ms │     no change │
│ QQuery 49 │         76.79 / 77.44 ±0.50 / 77.94 ms │        76.82 / 77.73 ±0.81 / 78.93 ms │     no change │
│ QQuery 50 │         59.13 / 59.87 ±1.09 / 62.00 ms │        59.14 / 59.46 ±0.24 / 59.83 ms │     no change │
│ QQuery 51 │         90.87 / 93.20 ±1.73 / 95.49 ms │        93.40 / 95.57 ±2.38 / 98.78 ms │     no change │
│ QQuery 52 │         24.25 / 26.32 ±3.73 / 33.77 ms │        24.68 / 25.06 ±0.58 / 26.20 ms │     no change │
│ QQuery 53 │         30.19 / 30.64 ±0.34 / 31.13 ms │        30.47 / 30.61 ±0.18 / 30.94 ms │     no change │
│ QQuery 54 │         55.98 / 56.53 ±0.49 / 57.28 ms │        56.27 / 56.42 ±0.12 / 56.60 ms │     no change │
│ QQuery 55 │         23.79 / 23.93 ±0.11 / 24.12 ms │        23.82 / 23.89 ±0.08 / 24.05 ms │     no change │
│ QQuery 56 │         39.35 / 40.14 ±0.69 / 41.17 ms │        38.96 / 40.87 ±3.17 / 47.20 ms │     no change │
│ QQuery 57 │      175.55 / 178.41 ±3.33 / 184.92 ms │     178.96 / 180.10 ±0.74 / 181.15 ms │     no change │
│ QQuery 58 │      115.95 / 118.18 ±2.11 / 121.49 ms │     117.11 / 119.01 ±1.30 / 120.97 ms │     no change │
│ QQuery 59 │      117.66 / 118.25 ±0.45 / 118.97 ms │     118.11 / 119.90 ±2.07 / 123.82 ms │     no change │
│ QQuery 60 │         39.70 / 41.07 ±1.04 / 42.92 ms │        39.90 / 40.87 ±0.61 / 41.82 ms │     no change │
│ QQuery 61 │         13.54 / 13.68 ±0.09 / 13.79 ms │        12.36 / 12.65 ±0.23 / 13.06 ms │ +1.08x faster │
│ QQuery 62 │         49.07 / 49.55 ±0.48 / 50.36 ms │        46.76 / 46.97 ±0.24 / 47.42 ms │ +1.05x faster │
│ QQuery 63 │         31.83 / 32.14 ±0.38 / 32.88 ms │        29.84 / 30.53 ±0.58 / 31.61 ms │ +1.05x faster │
│ QQuery 64 │     425.13 / 443.76 ±11.52 / 460.80 ms │     413.28 / 419.54 ±7.40 / 432.12 ms │ +1.06x faster │
│ QQuery 65 │      150.06 / 155.45 ±7.30 / 169.65 ms │     143.39 / 146.51 ±2.45 / 150.75 ms │ +1.06x faster │
│ QQuery 66 │         79.23 / 80.61 ±1.73 / 83.85 ms │        81.42 / 82.45 ±1.12 / 84.61 ms │     no change │
│ QQuery 67 │      243.17 / 247.42 ±6.20 / 259.71 ms │    249.06 / 272.61 ±17.84 / 294.32 ms │  1.10x slower │
│ QQuery 68 │         12.36 / 12.71 ±0.53 / 13.75 ms │        13.47 / 13.91 ±0.62 / 15.12 ms │  1.09x slower │
│ QQuery 69 │         58.33 / 58.83 ±0.28 / 59.13 ms │        61.70 / 62.59 ±0.58 / 63.44 ms │  1.06x slower │
│ QQuery 70 │      105.25 / 108.38 ±3.20 / 114.33 ms │     115.05 / 121.78 ±7.84 / 134.55 ms │  1.12x slower │
│ QQuery 71 │         35.44 / 38.24 ±4.39 / 46.94 ms │        39.03 / 39.49 ±0.38 / 40.04 ms │     no change │
│ QQuery 72 │ 2055.37 / 2224.86 ±192.13 / 2562.07 ms │ 2096.88 / 2166.51 ±63.71 / 2270.18 ms │     no change │
│ QQuery 73 │          9.80 / 10.04 ±0.15 / 10.26 ms │        10.52 / 10.76 ±0.18 / 11.04 ms │  1.07x slower │
│ QQuery 74 │      172.99 / 176.34 ±2.75 / 180.15 ms │     211.23 / 213.22 ±1.89 / 216.76 ms │  1.21x slower │
│ QQuery 75 │      150.08 / 151.32 ±0.92 / 152.83 ms │     161.80 / 165.13 ±5.40 / 175.87 ms │  1.09x slower │
│ QQuery 76 │         35.79 / 39.04 ±3.69 / 45.73 ms │        37.25 / 38.10 ±0.56 / 38.69 ms │     no change │
│ QQuery 77 │         61.72 / 62.32 ±0.41 / 62.72 ms │        63.53 / 65.39 ±1.54 / 68.09 ms │     no change │
│ QQuery 78 │      198.55 / 201.76 ±2.64 / 205.69 ms │     210.76 / 213.11 ±2.01 / 216.58 ms │  1.06x slower │
│ QQuery 79 │         67.25 / 68.00 ±0.73 / 69.38 ms │        70.54 / 73.07 ±3.61 / 80.23 ms │  1.07x slower │
│ QQuery 80 │      100.87 / 102.17 ±2.00 / 106.14 ms │     101.51 / 104.93 ±2.63 / 107.91 ms │     no change │
│ QQuery 81 │         26.05 / 26.53 ±0.33 / 26.94 ms │        26.32 / 26.80 ±0.81 / 28.42 ms │     no change │
│ QQuery 82 │         16.66 / 17.96 ±2.45 / 22.87 ms │        16.80 / 16.93 ±0.08 / 17.04 ms │ +1.06x faster │
│ QQuery 83 │         40.20 / 40.55 ±0.24 / 40.94 ms │        41.11 / 42.96 ±2.71 / 48.31 ms │  1.06x slower │
│ QQuery 84 │         30.90 / 31.06 ±0.13 / 31.24 ms │        30.56 / 31.17 ±0.42 / 31.73 ms │     no change │
│ QQuery 85 │      107.96 / 110.63 ±2.90 / 115.54 ms │     107.52 / 108.93 ±1.31 / 111.43 ms │     no change │
│ QQuery 86 │         25.84 / 26.25 ±0.27 / 26.69 ms │        25.78 / 25.99 ±0.11 / 26.08 ms │     no change │
│ QQuery 87 │         63.22 / 64.56 ±1.22 / 66.75 ms │        63.65 / 65.47 ±2.54 / 70.50 ms │     no change │
│ QQuery 88 │         64.50 / 66.57 ±3.45 / 73.44 ms │        63.75 / 64.13 ±0.31 / 64.61 ms │     no change │
│ QQuery 89 │         36.92 / 37.61 ±0.49 / 38.19 ms │        36.32 / 37.32 ±0.92 / 38.81 ms │     no change │
│ QQuery 90 │         17.66 / 18.03 ±0.29 / 18.43 ms │        17.55 / 17.83 ±0.20 / 18.15 ms │     no change │
│ QQuery 91 │         47.49 / 47.64 ±0.16 / 47.93 ms │        46.93 / 48.23 ±2.06 / 52.31 ms │     no change │
│ QQuery 92 │         30.26 / 30.96 ±0.55 / 31.89 ms │        30.37 / 31.18 ±0.80 / 32.24 ms │     no change │
│ QQuery 93 │         50.83 / 52.10 ±1.44 / 54.69 ms │        50.30 / 50.98 ±0.63 / 52.03 ms │     no change │
│ QQuery 94 │         38.84 / 39.28 ±0.31 / 39.70 ms │        38.63 / 39.51 ±1.00 / 41.42 ms │     no change │
│ QQuery 95 │         81.05 / 82.15 ±0.62 / 82.94 ms │        82.48 / 84.84 ±2.02 / 88.00 ms │     no change │
│ QQuery 96 │         24.39 / 24.75 ±0.35 / 25.34 ms │        24.62 / 24.80 ±0.28 / 25.34 ms │     no change │
│ QQuery 97 │         46.47 / 48.44 ±2.41 / 53.13 ms │        46.86 / 47.46 ±0.57 / 48.54 ms │     no change │
│ QQuery 98 │         42.63 / 43.31 ±0.50 / 43.97 ms │        41.94 / 42.85 ±0.62 / 43.87 ms │     no change │
│ QQuery 99 │         69.91 / 70.53 ±0.69 / 71.77 ms │        71.70 / 73.14 ±2.05 / 77.15 ms │     no change │
└───────────┴────────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                   ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                   │ 10147.32ms │
│ Total Time (fast-batch-wise-skip)   │ 10307.09ms │
│ Average Time (HEAD)                 │   102.50ms │
│ Average Time (fast-batch-wise-skip) │   104.11ms │
│ Queries Faster                      │          8 │
│ Queries Slower                      │         14 │
│ Queries with No Change              │         77 │
│ Queries with Failure                │          0 │
└─────────────────────────────────────┴────────────┘

Resource Usage

tpcds — base (merge-base)

Metric	Value
Wall time	55.0s
Peak memory	2.0 GiB
Avg memory	1.5 GiB
CPU user	228.4s
CPU sys	6.2s
Peak spill	0 B

tpcds — branch

Metric	Value
Wall time	55.0s
Peak memory	2.2 GiB
Avg memory	1.5 GiB
CPU user	232.7s
CPU sys	6.3s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-07-06T17:56:59Z

🤖 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

group                                                                       HEAD                                   fast-batch-wise-skip
-----                                                                       ----                                   --------------------
bench_merge_sorted_preserving/multiple_large_string_columns_with_1m_rows    1.00  1788.1±47.71ms        ? ?/sec    1.01  1808.3±63.59ms        ? ?/sec
bench_merge_sorted_preserving/multiple_u64_columns_with_1m_rows             1.10     77.0±3.72ms        ? ?/sec    1.00     69.9±5.47ms        ? ?/sec
bench_merge_sorted_preserving/single_large_string_column_with_1m_rows       1.00    250.5±2.72ms        ? ?/sec    1.03    258.4±7.39ms        ? ?/sec
bench_merge_sorted_preserving/single_u64_column_with_1m_rows                1.00     30.7±0.17ms        ? ?/sec    1.05     32.3±0.22ms        ? ?/sec

Resource Usage

sort_preserving_merge — base (merge-base)

Metric	Value
Wall time	405.1s
Peak memory	6.3 GiB
Avg memory	2.3 GiB
CPU user	122.8s
CPU sys	122.8s
Peak spill	0 B

sort_preserving_merge — branch

Metric	Value
Wall time	405.1s
Peak memory	6.3 GiB
Avg memory	2.3 GiB
CPU user	123.0s
CPU sys	124.6s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-07-06T17:58:36Z

🤖 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 fast-batch-wise-skip
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃                  fast-batch-wise-skip ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │          1.23 / 4.07 ±5.53 / 15.13 ms │          1.20 / 4.03 ±5.51 / 15.04 ms │     no change │
│ QQuery 1  │        12.77 / 12.87 ±0.11 / 13.07 ms │        12.37 / 12.72 ±0.24 / 13.06 ms │     no change │
│ QQuery 2  │        35.55 / 35.95 ±0.45 / 36.81 ms │        35.78 / 35.98 ±0.27 / 36.52 ms │     no change │
│ QQuery 3  │        30.78 / 31.66 ±0.79 / 32.95 ms │        30.32 / 30.53 ±0.14 / 30.74 ms │     no change │
│ QQuery 4  │     224.62 / 228.73 ±2.97 / 231.78 ms │     223.07 / 227.70 ±3.05 / 231.70 ms │     no change │
│ QQuery 5  │     276.29 / 279.56 ±2.56 / 282.61 ms │     272.24 / 275.62 ±2.09 / 278.05 ms │     no change │
│ QQuery 6  │           1.27 / 1.42 ±0.23 / 1.86 ms │           1.24 / 1.40 ±0.23 / 1.84 ms │     no change │
│ QQuery 7  │        13.77 / 13.90 ±0.09 / 14.03 ms │        13.51 / 13.92 ±0.23 / 14.15 ms │     no change │
│ QQuery 8  │     326.23 / 328.80 ±2.12 / 331.30 ms │     322.59 / 325.98 ±3.78 / 332.95 ms │     no change │
│ QQuery 9  │     463.58 / 469.96 ±4.70 / 477.94 ms │     460.87 / 470.11 ±8.58 / 485.24 ms │     no change │
│ QQuery 10 │        71.37 / 72.02 ±0.50 / 72.60 ms │        70.57 / 71.77 ±1.21 / 73.96 ms │     no change │
│ QQuery 11 │        81.75 / 82.81 ±1.09 / 84.26 ms │        81.69 / 82.77 ±1.35 / 84.92 ms │     no change │
│ QQuery 12 │     268.95 / 274.42 ±6.00 / 285.41 ms │     267.09 / 271.04 ±2.85 / 275.63 ms │     no change │
│ QQuery 13 │     366.91 / 376.31 ±6.78 / 386.61 ms │     372.84 / 378.10 ±2.69 / 380.09 ms │     no change │
│ QQuery 14 │     284.34 / 286.22 ±1.42 / 288.37 ms │     285.25 / 289.83 ±5.38 / 296.59 ms │     no change │
│ QQuery 15 │     276.07 / 279.74 ±2.39 / 282.89 ms │     274.76 / 282.83 ±5.52 / 291.55 ms │     no change │
│ QQuery 16 │     619.51 / 624.30 ±4.39 / 632.45 ms │    618.72 / 659.53 ±50.77 / 759.43 ms │  1.06x slower │
│ QQuery 17 │     626.24 / 633.69 ±7.56 / 642.92 ms │    628.93 / 659.78 ±51.09 / 761.80 ms │     no change │
│ QQuery 18 │ 1256.42 / 1287.88 ±19.04 / 1315.08 ms │  1274.48 / 1286.35 ±7.12 / 1295.86 ms │     no change │
│ QQuery 19 │        27.97 / 31.08 ±4.96 / 40.90 ms │        27.93 / 33.35 ±9.89 / 53.12 ms │  1.07x slower │
│ QQuery 20 │    514.00 / 523.96 ±12.32 / 547.79 ms │    514.94 / 532.56 ±17.45 / 562.20 ms │     no change │
│ QQuery 21 │     517.49 / 522.42 ±3.83 / 528.10 ms │     514.56 / 522.29 ±7.26 / 535.58 ms │     no change │
│ QQuery 22 │   995.42 / 1004.69 ±9.57 / 1022.21 ms │   997.33 / 1003.76 ±3.56 / 1007.13 ms │     no change │
│ QQuery 23 │ 3113.78 / 3138.93 ±23.02 / 3171.14 ms │ 3125.86 / 3146.74 ±14.74 / 3165.05 ms │     no change │
│ QQuery 24 │        41.53 / 44.88 ±5.43 / 55.66 ms │       42.24 / 47.56 ±10.27 / 68.09 ms │  1.06x slower │
│ QQuery 25 │    112.60 / 119.53 ±11.27 / 141.99 ms │     113.70 / 119.15 ±5.28 / 128.00 ms │     no change │
│ QQuery 26 │        41.97 / 42.77 ±0.67 / 43.64 ms │        42.67 / 44.63 ±3.03 / 50.67 ms │     no change │
│ QQuery 27 │     667.39 / 674.92 ±4.73 / 681.77 ms │     682.56 / 691.28 ±8.10 / 702.75 ms │     no change │
│ QQuery 28 │ 3042.94 / 3073.61 ±20.44 / 3101.79 ms │ 3062.10 / 3093.41 ±28.08 / 3129.48 ms │     no change │
│ QQuery 29 │       40.77 / 67.49 ±17.07 / 88.20 ms │       41.51 / 51.65 ±19.49 / 90.64 ms │ +1.31x faster │
│ QQuery 30 │     303.67 / 307.63 ±3.99 / 314.73 ms │     309.20 / 314.09 ±4.17 / 320.61 ms │     no change │
│ QQuery 31 │    282.77 / 296.24 ±14.91 / 325.17 ms │    284.27 / 304.20 ±19.57 / 339.81 ms │     no change │
│ QQuery 32 │   919.34 / 970.38 ±48.43 / 1048.25 ms │   971.31 / 994.30 ±15.14 / 1011.14 ms │     no change │
│ QQuery 33 │ 1457.29 / 1505.26 ±30.34 / 1541.67 ms │ 1535.60 / 1573.03 ±33.81 / 1620.85 ms │     no change │
│ QQuery 34 │ 1476.09 / 1530.91 ±74.90 / 1673.78 ms │ 1581.87 / 1603.11 ±15.91 / 1621.36 ms │     no change │
│ QQuery 35 │    279.74 / 316.35 ±38.71 / 389.13 ms │     321.33 / 326.91 ±7.55 / 341.79 ms │     no change │
│ QQuery 36 │        66.28 / 76.64 ±8.21 / 89.74 ms │        68.20 / 74.90 ±4.45 / 81.10 ms │     no change │
│ QQuery 37 │        35.66 / 36.53 ±0.85 / 37.95 ms │        37.96 / 46.44 ±6.55 / 58.09 ms │  1.27x slower │
│ QQuery 38 │        42.24 / 45.02 ±3.81 / 52.57 ms │        42.19 / 45.40 ±2.87 / 50.53 ms │     no change │
│ QQuery 39 │    134.33 / 150.42 ±10.03 / 162.35 ms │     142.68 / 159.06 ±9.09 / 170.21 ms │  1.06x slower │
│ QQuery 40 │        13.88 / 14.19 ±0.47 / 15.12 ms │        15.28 / 17.45 ±3.52 / 24.46 ms │  1.23x slower │
│ QQuery 41 │        13.44 / 13.63 ±0.12 / 13.79 ms │        14.55 / 15.09 ±0.44 / 15.54 ms │  1.11x slower │
│ QQuery 42 │        13.26 / 14.54 ±2.10 / 18.71 ms │        13.95 / 14.25 ±0.24 / 14.50 ms │     no change │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                   ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                   │ 19846.35ms │
│ Total Time (fast-batch-wise-skip)   │ 20154.61ms │
│ Average Time (HEAD)                 │   461.54ms │
│ Average Time (fast-batch-wise-skip) │   468.71ms │
│ Queries Faster                      │          1 │
│ Queries Slower                      │          7 │
│ Queries with No Change              │         35 │
│ Queries with Failure                │          0 │
└─────────────────────────────────────┴────────────┘

Resource Usage

clickbench_partitioned — base (merge-base)

Metric	Value
Wall time	100.0s
Peak memory	12.1 GiB
Avg memory	4.6 GiB
CPU user	1019.6s
CPU sys	69.1s
Peak spill	0 B

clickbench_partitioned — branch

Metric	Value
Wall time	105.0s
Peak memory	12.7 GiB
Avg memory	4.4 GiB
CPU user	1035.3s
CPU sys	73.1s
Peak spill	0 B

File an issue against this benchmark runner

rluvaton · 2026-07-06T18:11:24Z

run benchmark sort_preserving_merge

adriangbot · 2026-07-06T18:13:33Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4896133554-873-8pdqc 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 fast-batch-wise-skip (1d5183b) to 0365d3c (merge-base) diff using: sort_preserving_merge
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-07-06T18:31:55Z

🤖 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

group                                                                       HEAD                                   fast-batch-wise-skip
-----                                                                       ----                                   --------------------
bench_merge_sorted_preserving/multiple_large_string_columns_with_1m_rows    1.00  1753.5±10.15ms        ? ?/sec    1.01  1779.1±10.65ms        ? ?/sec
bench_merge_sorted_preserving/multiple_u64_columns_with_1m_rows             1.00     65.5±0.52ms        ? ?/sec    1.03     67.4±0.33ms        ? ?/sec
bench_merge_sorted_preserving/single_large_string_column_with_1m_rows       1.00    252.0±2.29ms        ? ?/sec    1.02    257.4±3.32ms        ? ?/sec
bench_merge_sorted_preserving/single_u64_column_with_1m_rows                1.00     30.8±0.16ms        ? ?/sec    1.05     32.5±0.27ms        ? ?/sec
spm/nearly_sorted/complex/payload_100                                                                              1.00     48.7±0.18ms        ? ?/sec
spm/nearly_sorted/complex/payload_20                                                                               1.00     18.8±0.09ms        ? ?/sec
spm/nearly_sorted/complex/payload_5                                                                                1.00     13.1±0.02ms        ? ?/sec
spm/nearly_sorted/string/payload_100                                                                               1.00     46.4±0.08ms        ? ?/sec
spm/nearly_sorted/string/payload_20                                                                                1.00     16.5±0.02ms        ? ?/sec
spm/nearly_sorted/string/payload_5                                                                                 1.00     10.9±0.01ms        ? ?/sec
spm/nearly_sorted/u64/payload_100                                                                                  1.00     41.6±0.08ms        ? ?/sec
spm/nearly_sorted/u64/payload_20                                                                                   1.00     11.9±0.02ms        ? ?/sec
spm/nearly_sorted/u64/payload_5                                                                                    1.00      6.3±0.01ms        ? ?/sec
spm/reverse/complex/payload_100                                                                                    1.00     16.0±0.03ms        ? ?/sec
spm/reverse/complex/payload_20                                                                                     1.00      7.4±0.02ms        ? ?/sec
spm/reverse/complex/payload_5                                                                                      1.00      5.8±0.01ms        ? ?/sec
spm/reverse/string/payload_100                                                                                     1.00     12.9±0.03ms        ? ?/sec
spm/reverse/string/payload_20                                                                                      1.00      4.4±0.01ms        ? ?/sec
spm/reverse/string/payload_5                                                                                       1.00      2.7±0.00ms        ? ?/sec
spm/reverse/u64/payload_100                                                                                        1.00     11.7±0.03ms        ? ?/sec
spm/reverse/u64/payload_20                                                                                         1.00      3.2±0.00ms        ? ?/sec
spm/reverse/u64/payload_5                                                                                          1.00   1640.6±7.76µs        ? ?/sec
spm/sorted/complex/payload_100                                                                                     1.00      4.8±0.01ms        ? ?/sec
spm/sorted/complex/payload_20                                                                                      1.00      3.7±0.01ms        ? ?/sec
spm/sorted/complex/payload_5                                                                                       1.00      3.4±0.03ms        ? ?/sec
spm/sorted/string/payload_100                                                                                      1.00   1448.1±7.46µs        ? ?/sec
spm/sorted/string/payload_20                                                                                       1.00    320.1±1.41µs        ? ?/sec
spm/sorted/string/payload_5                                                                                        1.00    116.3±0.38µs        ? ?/sec
spm/sorted/u64/payload_100                                                                                         1.00  1447.3±12.29µs        ? ?/sec
spm/sorted/u64/payload_20                                                                                          1.00    309.2±2.31µs        ? ?/sec
spm/sorted/u64/payload_5                                                                                           1.00    106.5±0.72µs        ? ?/sec
spm/unsorted/complex/payload_100                                                                                   1.00     55.8±0.08ms        ? ?/sec
spm/unsorted/complex/payload_20                                                                                    1.00     26.0±0.17ms        ? ?/sec
spm/unsorted/complex/payload_5                                                                                     1.00     20.2±0.03ms        ? ?/sec
spm/unsorted/string/payload_100                                                                                    1.00     52.2±0.14ms        ? ?/sec
spm/unsorted/string/payload_20                                                                                     1.00     22.4±0.04ms        ? ?/sec
spm/unsorted/string/payload_5                                                                                      1.00     16.6±0.03ms        ? ?/sec
spm/unsorted/u64/payload_100                                                                                       1.00     46.2±0.08ms        ? ?/sec
spm/unsorted/u64/payload_20                                                                                        1.00     16.5±0.03ms        ? ?/sec
spm/unsorted/u64/payload_5                                                                                         1.00     10.9±0.03ms        ? ?/sec

Resource Usage

sort_preserving_merge — base (merge-base)

Metric	Value
Wall time	350.1s
Peak memory	6.3 GiB
Avg memory	2.6 GiB
CPU user	119.3s
CPU sys	121.2s
Peak spill	0 B

sort_preserving_merge — branch

Metric	Value
Wall time	740.2s
Peak memory	6.4 GiB
Avg memory	1.4 GiB
CPU user	594.2s
CPU sys	124.1s
Peak spill	0 B

File an issue against this benchmark runner

rluvaton · 2026-07-06T18:49:25Z

run benchmark sort_preserving_merge

This reverts commit 5a5e1f0.

adriangbot · 2026-07-06T18:52:28Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4896500200-874-bl5dv 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 fast-batch-wise-skip (5a5e1f0) to 0365d3c (merge-base) diff using: sort_preserving_merge
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-07-06T19:11:55Z

🤖 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

group                                                                       HEAD                                   fast-batch-wise-skip
-----                                                                       ----                                   --------------------
bench_merge_sorted_preserving/multiple_large_string_columns_with_1m_rows    1.00  1814.1±22.63ms        ? ?/sec    1.00  1813.9±23.47ms        ? ?/sec
bench_merge_sorted_preserving/multiple_u64_columns_with_1m_rows             1.00     65.4±0.38ms        ? ?/sec    1.03     67.3±3.50ms        ? ?/sec
bench_merge_sorted_preserving/single_large_string_column_with_1m_rows       1.00    255.0±2.10ms        ? ?/sec    1.00    255.6±3.78ms        ? ?/sec
bench_merge_sorted_preserving/single_u64_column_with_1m_rows                1.00     30.6±0.19ms        ? ?/sec    1.01     30.9±0.28ms        ? ?/sec
spm/nearly_sorted/complex/payload_100                                                                              1.00     48.2±0.09ms        ? ?/sec
spm/nearly_sorted/complex/payload_20                                                                               1.00     18.7±0.07ms        ? ?/sec
spm/nearly_sorted/complex/payload_5                                                                                1.00     13.1±0.11ms        ? ?/sec
spm/nearly_sorted/string/payload_100                                                                               1.00     45.5±0.11ms        ? ?/sec
spm/nearly_sorted/string/payload_20                                                                                1.00     16.0±0.05ms        ? ?/sec
spm/nearly_sorted/string/payload_5                                                                                 1.00     10.4±0.07ms        ? ?/sec
spm/nearly_sorted/u64/payload_100                                                                                  1.00     41.4±0.08ms        ? ?/sec
spm/nearly_sorted/u64/payload_20                                                                                   1.00     11.8±0.08ms        ? ?/sec
spm/nearly_sorted/u64/payload_5                                                                                    1.00      6.1±0.02ms        ? ?/sec
spm/reverse/complex/payload_100                                                                                    1.00     48.1±0.13ms        ? ?/sec
spm/reverse/complex/payload_20                                                                                     1.00     18.2±0.04ms        ? ?/sec
spm/reverse/complex/payload_5                                                                                      1.00     12.7±0.15ms        ? ?/sec
spm/reverse/string/payload_100                                                                                     1.00     45.4±0.15ms        ? ?/sec
spm/reverse/string/payload_20                                                                                      1.00     15.7±0.05ms        ? ?/sec
spm/reverse/string/payload_5                                                                                       1.00     10.1±0.02ms        ? ?/sec
spm/reverse/u64/payload_100                                                                                        1.00     41.3±0.14ms        ? ?/sec
spm/reverse/u64/payload_20                                                                                         1.00     11.7±0.08ms        ? ?/sec
spm/reverse/u64/payload_5                                                                                          1.00      6.1±0.07ms        ? ?/sec
spm/sorted/complex/payload_100                                                                                     1.00     48.0±0.12ms        ? ?/sec
spm/sorted/complex/payload_20                                                                                      1.00     18.3±0.16ms        ? ?/sec
spm/sorted/complex/payload_5                                                                                       1.00     12.8±0.12ms        ? ?/sec
spm/sorted/string/payload_100                                                                                      1.00     45.5±0.10ms        ? ?/sec
spm/sorted/string/payload_20                                                                                       1.00     15.7±0.03ms        ? ?/sec
spm/sorted/string/payload_5                                                                                        1.00     10.2±0.07ms        ? ?/sec
spm/sorted/u64/payload_100                                                                                         1.00     41.2±0.12ms        ? ?/sec
spm/sorted/u64/payload_20                                                                                          1.00     11.6±0.07ms        ? ?/sec
spm/sorted/u64/payload_5                                                                                           1.00      6.0±0.02ms        ? ?/sec
spm/unsorted/complex/payload_100                                                                                   1.00     55.9±0.16ms        ? ?/sec
spm/unsorted/complex/payload_20                                                                                    1.00     25.8±0.04ms        ? ?/sec
spm/unsorted/complex/payload_5                                                                                     1.00     20.0±0.10ms        ? ?/sec
spm/unsorted/string/payload_100                                                                                    1.00     51.6±0.12ms        ? ?/sec
spm/unsorted/string/payload_20                                                                                     1.00     21.9±0.03ms        ? ?/sec
spm/unsorted/string/payload_5                                                                                      1.00     16.2±0.04ms        ? ?/sec
spm/unsorted/u64/payload_100                                                                                       1.00     45.8±0.06ms        ? ?/sec
spm/unsorted/u64/payload_20                                                                                        1.00     16.1±0.05ms        ? ?/sec
spm/unsorted/u64/payload_5                                                                                         1.00     10.5±0.06ms        ? ?/sec

Resource Usage

sort_preserving_merge — base (merge-base)

Metric	Value
Wall time	355.1s
Peak memory	6.3 GiB
Avg memory	2.6 GiB
CPU user	120.2s
CPU sys	127.4s
Peak spill	0 B

sort_preserving_merge — branch

Metric	Value
Wall time	800.2s
Peak memory	6.3 GiB
Avg memory	1.4 GiB
CPU user	608.1s
CPU sys	127.4s
Peak spill	0 B

File an issue against this benchmark runner

rluvaton · 2026-07-06T19:21:57Z

run benchmark clickbench_sorted

adriangbot · 2026-07-06T19:24:42Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4896780949-875-vww69 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 fast-batch-wise-skip (80d85bf) to 0365d3c (merge-base) diff using: clickbench_sorted
Results will be posted here when complete

File an issue against this benchmark runner

rluvaton · 2026-07-06T19:28:22Z

run benchmark sort_tpch10

adriangbot · 2026-07-06T19:31:16Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4896828325-876-rt6zg 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 fast-batch-wise-skip (80d85bf) to 0365d3c (merge-base) diff using: sort_tpch10
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-07-06T19:48:26Z

🤖 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 fast-batch-wise-skip
--------------------
Benchmark sort_tpch10.json
--------------------
┏━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃                                   HEAD ┃                   fast-batch-wise-skip ┃        Change ┃
┡━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ Q1    │   1631.98 / 1643.06 ±8.49 / 1658.03 ms │  1396.34 / 1418.26 ±17.19 / 1437.64 ms │ +1.16x faster │
│ Q2    │  1421.45 / 1431.48 ±10.65 / 1451.01 ms │      201.29 / 208.62 ±6.77 / 217.57 ms │ +6.86x faster │
│ Q3    │ 8893.78 / 9189.49 ±218.44 / 9425.68 ms │ 8880.60 / 9062.28 ±184.88 / 9413.73 ms │     no change │
│ Q4    │  2741.65 / 2770.88 ±16.58 / 2789.94 ms │     392.40 / 423.10 ±20.71 / 453.79 ms │ +6.55x faster │
│ Q5    │  2644.71 / 2680.47 ±22.80 / 2708.24 ms │  2700.25 / 2725.43 ±19.05 / 2745.08 ms │     no change │
│ Q6    │  2819.27 / 2853.20 ±33.86 / 2901.49 ms │  2827.71 / 2850.04 ±17.91 / 2878.97 ms │     no change │
│ Q7    │  5602.34 / 5670.57 ±63.47 / 5780.55 ms │  5569.72 / 5676.77 ±82.71 / 5824.64 ms │     no change │
│ Q8    │  3978.51 / 4065.05 ±75.19 / 4176.24 ms │  3456.58 / 3553.36 ±82.71 / 3672.65 ms │ +1.14x faster │
│ Q9    │  4369.52 / 4443.83 ±58.81 / 4508.15 ms │  3719.71 / 3864.12 ±88.71 / 3970.66 ms │ +1.15x faster │
│ Q10   │ 7035.70 / 7185.30 ±128.68 / 7353.68 ms │ 5843.73 / 6066.70 ±184.81 / 6299.54 ms │ +1.18x faster │
│ Q11   │  3674.51 / 3745.29 ±64.70 / 3854.17 ms │ 3834.03 / 3984.65 ±202.64 / 4361.24 ms │  1.06x slower │
└───────┴────────────────────────────────────────┴────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                   ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                   │ 45678.61ms │
│ Total Time (fast-batch-wise-skip)   │ 39833.34ms │
│ Average Time (HEAD)                 │  4152.60ms │
│ Average Time (fast-batch-wise-skip) │  3621.21ms │
│ Queries Faster                      │          6 │
│ Queries Slower                      │          1 │
│ Queries with No Change              │          4 │
│ Queries with Failure                │          0 │
└─────────────────────────────────────┴────────────┘

Resource Usage

sort_tpch10 — base (merge-base)

Metric	Value
Wall time	230.1s
Peak memory	19.0 GiB
Avg memory	7.5 GiB
CPU user	909.4s
CPU sys	70.8s
Peak spill	0 B

sort_tpch10 — branch

Metric	Value
Wall time	200.0s
Peak memory	20.8 GiB
Avg memory	8.9 GiB
CPU user	828.9s
CPU sys	51.8s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-07-06T20:07:22Z

🤖 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 fast-batch-wise-skip
--------------------
Benchmark clickbench_sorted.json
--------------------
┏━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query    ┃                               HEAD ┃             fast-batch-wise-skip ┃        Change ┃
┡━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 79.41 / 160.38 ±115.25 / 380.87 ms │ 79.02 / 95.76 ±27.65 / 150.45 ms │ +1.67x faster │
└──────────┴────────────────────────────────────┴──────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                   ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                   │ 160.38ms │
│ Total Time (fast-batch-wise-skip)   │  95.76ms │
│ Average Time (HEAD)                 │ 160.38ms │
│ Average Time (fast-batch-wise-skip) │  95.76ms │
│ Queries Faster                      │        1 │
│ Queries Slower                      │        0 │
│ Queries with No Change              │        0 │
│ Queries with Failure                │        0 │
└─────────────────────────────────────┴──────────┘

Resource Usage

clickbench_sorted — base (merge-base)

Metric	Value
Wall time	5.0s
Peak memory	285.7 MiB
Avg memory	47.7 MiB
CPU user	0.0s
CPU sys	0.0s
Peak spill	0 B

clickbench_sorted — branch

Metric	Value
Wall time	5.0s
Peak memory	1.5 GiB
Avg memory	249.4 MiB
CPU user	0.0s
CPU sys	0.0s
Peak spill	0 B

File an issue against this benchmark runner

rluvaton · 2026-07-06T20:52:53Z

Extracted to:

bench: add sort benchmarks for various data profile #23346

rluvaton added 3 commits July 6, 2026 18:35

perf(sort): fast path for full batch when it sorts before all other s…

2947e23

…treams

added tests

c2bb469

added test

f3cffeb

github-actions Bot added the physical-plan Changes to the physical-plan crate label Jul 6, 2026

rluvaton added 3 commits July 6, 2026 21:05

added benches

6dc6b2d

change collect to drain to avoid holding the memory

55ece4b

Merge branch 'main' into fast-batch-wise-skip

1d5183b

revert change and see perf

5a5e1f0

Revert "revert change and see perf"

80d85bf

This reverts commit 5a5e1f0.

fix tests and bugs

17bafed

github-actions Bot added core Core DataFusion crate ffi Changes to the ffi crate labels Jul 6, 2026

rluvaton commented Jul 6, 2026

View reviewed changes

add comment

92505cf

rluvaton marked this pull request as ready for review July 6, 2026 21:01

rluvaton added the performance Make DataFusion faster label Jul 6, 2026

rluvaton changed the title ~~perf(sort): fast path for full batch when it sorts before all other streams in SortPreservingMergeStream~~ perf(sort): fast path for full batch when it sorts before all other streams in SortPreservingMergeStream Jul 6, 2026

Uh oh!

Conversation

rluvaton commented Jul 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Which issue does this PR close?

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

Uh oh!

rluvaton commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

rluvaton commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

rluvaton commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

rluvaton commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

rluvaton commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

rluvaton commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

adriangbot commented Jul 6, 2026

Uh oh!

rluvaton Jul 6, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

rluvaton commented Jul 6, 2026 •

edited

Loading