aggregate blocked impl by rluvaton · Pull Request #24928 · apache/datafusion

rluvaton · 2026-09-03T16:55:44Z

Huge blocked impl that is backwards compatible
And just see the performance cost

Currently it only contain blocked impl for single group by and some aggregate expression

Which issue does this PR close?

Closes #.

Rationale for this change

What changes are included in this PR?

What is the testing strategy for this PR?

Are there any user-facing changes?

…so all usages will be migrated

…warded

…ockedGroupColumn`

…not `usize`

…e grouped_hash_stream code

# Conflicts: # datafusion/expr-common/src/groups_accumulator.rs

… often

rluvaton · 2026-09-06T18:22:09Z

run benchmark h2o_small h2o_medium external_aggr

rluvaton · 2026-09-06T18:22:22Z

run benchmarks

adriangbot · 2026-09-06T18:24:54Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5561218687-2175-tcqw8 6.12.94+ #1 SMP Fri Jul 17 09:42:57 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 add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark h2o_small

Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:25:03Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5561219942-2179-qz98x 6.12.94+ #1 SMP Fri Jul 17 09:42:57 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 add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark tpcds

Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:25:12Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5561219942-2180-qmds9 6.12.94+ #1 SMP Fri Jul 17 09:42:57 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 add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark tpch

Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:25:20Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5561218687-2176-zrdnc 6.12.94+ #1 SMP Fri Jul 17 09:42:57 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 add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark h2o_medium

Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:25:29Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5561218687-2177-bhrp8 6.12.94+ #1 SMP Fri Jul 17 09:42:57 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 add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark external_aggr

Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:25:55Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5561219942-2178-hzwzl 6.12.94+ #1 SMP Fri Jul 17 09:42:57 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 add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark clickbench_partitioned

Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:40:02Z

🤖 Benchmark completed (GKE) | trigger

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

Comparing add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark tpch

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 add-blocks-impl-from-scratch
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃     HEAD ┃ add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 1  │ 38.98 ms │                     40.65 ms │    no change │
│ QQuery 2  │ 18.86 ms │                     19.53 ms │    no change │
│ QQuery 3  │ 28.19 ms │                     29.70 ms │ 1.05x slower │
│ QQuery 4  │ 17.26 ms │                     17.94 ms │    no change │
│ QQuery 5  │ 34.90 ms │                     35.66 ms │    no change │
│ QQuery 6  │ 15.82 ms │                     15.75 ms │    no change │
│ QQuery 7  │ 40.34 ms │                     42.12 ms │    no change │
│ QQuery 8  │ 40.73 ms │                     41.47 ms │    no change │
│ QQuery 9  │ 49.36 ms │                     50.21 ms │    no change │
│ QQuery 10 │ 42.05 ms │                     45.50 ms │ 1.08x slower │
│ QQuery 11 │ 13.22 ms │                     13.85 ms │    no change │
│ QQuery 12 │ 23.56 ms │                     25.24 ms │ 1.07x slower │
│ QQuery 13 │ 38.60 ms │                     42.63 ms │ 1.10x slower │
│ QQuery 14 │ 24.41 ms │                     24.22 ms │    no change │
│ QQuery 15 │ 30.54 ms │                     31.68 ms │    no change │
│ QQuery 16 │ 13.76 ms │                     18.61 ms │ 1.35x slower │
│ QQuery 17 │ 68.57 ms │                     79.48 ms │ 1.16x slower │
│ QQuery 18 │ 57.81 ms │                     69.51 ms │ 1.20x slower │
│ QQuery 19 │ 32.46 ms │                     32.64 ms │    no change │
│ QQuery 20 │ 30.99 ms │                     36.21 ms │ 1.17x slower │
│ QQuery 21 │ 54.69 ms │                     55.86 ms │    no change │
│ QQuery 22 │ 13.83 ms │                     14.26 ms │    no change │
└───────────┴──────────┴──────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                           ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 728.94ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 782.72ms │
│ Average Time (HEAD)                         │  33.13ms │
│ Average Time (add-blocks-impl-from-scratch) │  35.58ms │
│ Queries Faster                              │        0 │
│ Queries Slower                              │        8 │
│ Queries with No Change                      │       14 │
│ Queries with Failure                        │        0 │
└─────────────────────────────────────────────┴──────────┘

Distribution per query (min / mean ±stddev / max):

Comparing HEAD and add-blocks-impl-from-scratch
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃   add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 1  │ 38.98 / 40.17 ±1.64 / 43.22 ms │ 40.65 / 41.23 ±1.02 / 43.26 ms │    no change │
│ QQuery 2  │ 18.86 / 19.01 ±0.13 / 19.19 ms │ 19.53 / 19.71 ±0.17 / 19.94 ms │    no change │
│ QQuery 3  │ 28.19 / 29.03 ±0.90 / 30.74 ms │ 29.70 / 29.92 ±0.17 / 30.12 ms │    no change │
│ QQuery 4  │ 17.26 / 17.82 ±0.57 / 18.92 ms │ 17.94 / 18.32 ±0.53 / 19.37 ms │    no change │
│ QQuery 5  │ 34.90 / 35.21 ±0.28 / 35.57 ms │ 35.66 / 36.02 ±0.33 / 36.62 ms │    no change │
│ QQuery 6  │ 15.82 / 16.00 ±0.11 / 16.13 ms │ 15.75 / 16.17 ±0.38 / 16.87 ms │    no change │
│ QQuery 7  │ 40.34 / 41.45 ±1.34 / 44.04 ms │ 42.12 / 42.95 ±1.17 / 45.24 ms │    no change │
│ QQuery 8  │ 40.73 / 42.13 ±2.39 / 46.89 ms │ 41.47 / 42.33 ±0.68 / 43.30 ms │    no change │
│ QQuery 9  │ 49.36 / 49.68 ±0.25 / 50.07 ms │ 50.21 / 51.04 ±0.79 / 52.31 ms │    no change │
│ QQuery 10 │ 42.05 / 42.15 ±0.07 / 42.21 ms │ 45.50 / 46.16 ±0.83 / 47.63 ms │ 1.10x slower │
│ QQuery 11 │ 13.22 / 13.39 ±0.17 / 13.69 ms │ 13.85 / 14.08 ±0.27 / 14.62 ms │ 1.05x slower │
│ QQuery 12 │ 23.56 / 23.89 ±0.23 / 24.12 ms │ 25.24 / 25.67 ±0.36 / 26.26 ms │ 1.07x slower │
│ QQuery 13 │ 38.60 / 40.28 ±2.03 / 44.04 ms │ 42.63 / 43.77 ±1.56 / 46.83 ms │ 1.09x slower │
│ QQuery 14 │ 24.41 / 24.59 ±0.21 / 25.00 ms │ 24.22 / 24.46 ±0.17 / 24.68 ms │    no change │
│ QQuery 15 │ 30.54 / 30.66 ±0.13 / 30.87 ms │ 31.68 / 31.88 ±0.17 / 32.18 ms │    no change │
│ QQuery 16 │ 13.76 / 13.88 ±0.10 / 14.05 ms │ 18.61 / 19.08 ±0.38 / 19.58 ms │ 1.37x slower │
│ QQuery 17 │ 68.57 / 69.39 ±0.47 / 69.84 ms │ 79.48 / 79.97 ±0.47 / 80.83 ms │ 1.15x slower │
│ QQuery 18 │ 57.81 / 61.22 ±3.10 / 67.03 ms │ 69.51 / 72.20 ±3.38 / 78.87 ms │ 1.18x slower │
│ QQuery 19 │ 32.46 / 33.43 ±1.31 / 35.97 ms │ 32.64 / 33.52 ±0.79 / 34.71 ms │    no change │
│ QQuery 20 │ 30.99 / 31.42 ±0.34 / 32.01 ms │ 36.21 / 36.68 ±0.41 / 37.40 ms │ 1.17x slower │
│ QQuery 21 │ 54.69 / 55.19 ±0.42 / 55.82 ms │ 55.86 / 56.41 ±0.53 / 57.10 ms │    no change │
│ QQuery 22 │ 13.83 / 14.22 ±0.37 / 14.92 ms │ 14.26 / 14.64 ±0.24 / 14.99 ms │    no change │
└───────────┴────────────────────────────────┴────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                           ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 744.21ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 796.22ms │
│ Average Time (HEAD)                         │  33.83ms │
│ Average Time (add-blocks-impl-from-scratch) │  36.19ms │
│ Queries Faster                              │        0 │
│ Queries Slower                              │        8 │
│ Queries with No Change                      │       14 │
│ Queries with Failure                        │        0 │
└─────────────────────────────────────────────┴──────────┘

Resource Usage

tpch — base (merge-base)

Metric	Value
Wall time	5.0s
Peak memory	1.2 GiB
Avg memory	499.0 MiB
CPU user	20.7s
CPU sys	1.6s
Peak spill	0 B

tpch — branch

Metric	Value
Wall time	5.0s
Peak memory	1.1 GiB
Avg memory	502.4 MiB
CPU user	21.3s
CPU sys	2.5s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:40:28Z

🤖 Benchmark completed (GKE) | trigger

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

Comparing add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark h2o_small

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 add-blocks-impl-from-scratch
--------------------
Benchmark h2o.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃      HEAD ┃ add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 1  │  98.30 ms │                     99.51 ms │    no change │
│ QQuery 2  │ 206.90 ms │                    248.84 ms │ 1.20x slower │
│ QQuery 3  │ 172.02 ms │                    193.28 ms │ 1.12x slower │
│ QQuery 4  │ 130.99 ms │                    133.58 ms │    no change │
│ QQuery 5  │ 167.53 ms │                    173.37 ms │    no change │
│ QQuery 6  │ 155.28 ms │                    166.78 ms │ 1.07x slower │
│ QQuery 7  │ 162.68 ms │                    179.96 ms │ 1.11x slower │
│ QQuery 8  │ 311.78 ms │                    326.78 ms │    no change │
│ QQuery 9  │ 254.12 ms │                    324.26 ms │ 1.28x slower │
│ QQuery 10 │ 248.61 ms │                    283.39 ms │ 1.14x slower │
└───────────┴───────────┴──────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 1908.22ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 2129.75ms │
│ Average Time (HEAD)                         │  190.82ms │
│ Average Time (add-blocks-impl-from-scratch) │  212.97ms │
│ Queries Faster                              │         0 │
│ Queries Slower                              │         6 │
│ Queries with No Change                      │         4 │
│ Queries with Failure                        │         0 │
└─────────────────────────────────────────────┴───────────┘

Distribution per query (min / mean ±stddev / max):

Comparing HEAD and add-blocks-impl-from-scratch
--------------------
Benchmark h2o.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃                               HEAD ┃       add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 1  │   98.30 / 100.12 ±1.34 / 101.49 ms │   99.51 / 100.17 ±0.55 / 100.86 ms │    no change │
│ QQuery 2  │  206.90 / 211.05 ±3.98 / 216.41 ms │  248.84 / 249.91 ±0.84 / 250.88 ms │ 1.18x slower │
│ QQuery 3  │  172.02 / 172.47 ±0.41 / 173.01 ms │  193.28 / 194.68 ±1.25 / 196.30 ms │ 1.13x slower │
│ QQuery 4  │  130.99 / 131.30 ±0.36 / 131.81 ms │  133.58 / 134.72 ±1.41 / 136.71 ms │    no change │
│ QQuery 5  │  167.53 / 168.71 ±1.06 / 170.09 ms │  173.37 / 174.87 ±1.11 / 176.03 ms │    no change │
│ QQuery 6  │  155.28 / 155.96 ±0.76 / 157.03 ms │  166.78 / 168.47 ±2.36 / 171.80 ms │ 1.08x slower │
│ QQuery 7  │  162.68 / 165.29 ±2.21 / 168.08 ms │  179.96 / 180.95 ±1.04 / 182.39 ms │ 1.09x slower │
│ QQuery 8  │  311.78 / 323.72 ±9.85 / 335.89 ms │ 326.78 / 351.82 ±27.08 / 389.44 ms │ 1.09x slower │
│ QQuery 9  │  254.12 / 259.26 ±4.17 / 264.33 ms │ 324.26 / 338.44 ±10.12 / 347.23 ms │ 1.31x slower │
│ QQuery 10 │ 248.61 / 259.76 ±13.77 / 279.16 ms │  283.39 / 294.71 ±9.13 / 305.76 ms │ 1.13x slower │
└───────────┴────────────────────────────────────┴────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 1947.63ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 2188.74ms │
│ Average Time (HEAD)                         │  194.76ms │
│ Average Time (add-blocks-impl-from-scratch) │  218.87ms │
│ Queries Faster                              │         0 │
│ Queries Slower                              │         7 │
│ Queries with No Change                      │         3 │
│ Queries with Failure                        │         0 │
└─────────────────────────────────────────────┴───────────┘

Resource Usage

h2o_small — base (merge-base)

Metric	Value
Wall time	10.0s
Peak memory	1.7 GiB
Avg memory	532.3 MiB
CPU user	52.1s
CPU sys	3.8s
Peak spill	0 B

h2o_small — branch

Metric	Value
Wall time	10.0s
Peak memory	974.0 MiB
Avg memory	438.5 MiB
CPU user	55.3s
CPU sys	6.1s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:41:53Z

🤖 Benchmark completed (GKE) | trigger

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

Comparing add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark tpcds

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 add-blocks-impl-from-scratch
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 1  │    5.74 ms │                      6.74 ms │ 1.17x slower │
│ QQuery 2  │   80.08 ms │                     86.41 ms │ 1.08x slower │
│ QQuery 3  │   28.60 ms │                     30.02 ms │    no change │
│ QQuery 4  │  477.71 ms │                    508.26 ms │ 1.06x slower │
│ QQuery 5  │   51.74 ms │                     55.63 ms │ 1.08x slower │
│ QQuery 6  │   35.61 ms │                     37.68 ms │ 1.06x slower │
│ QQuery 7  │   74.17 ms │                     77.51 ms │    no change │
│ QQuery 8  │   36.92 ms │                     37.92 ms │    no change │
│ QQuery 9  │   52.95 ms │                     52.94 ms │    no change │
│ QQuery 10 │   62.48 ms │                     64.66 ms │    no change │
│ QQuery 11 │  296.17 ms │                    323.82 ms │ 1.09x slower │
│ QQuery 12 │   28.87 ms │                     30.11 ms │    no change │
│ QQuery 13 │  117.50 ms │                    118.68 ms │    no change │
│ QQuery 14 │  413.95 ms │                    428.71 ms │    no change │
│ QQuery 15 │   56.38 ms │                     58.07 ms │    no change │
│ QQuery 16 │    6.80 ms │                      7.18 ms │ 1.06x slower │
│ QQuery 17 │   78.68 ms │                     81.46 ms │    no change │
│ QQuery 18 │  101.81 ms │                    107.92 ms │ 1.06x slower │
│ QQuery 19 │   41.25 ms │                     43.16 ms │    no change │
│ QQuery 20 │   35.12 ms │                     37.86 ms │ 1.08x slower │
│ QQuery 21 │   17.10 ms │                     18.27 ms │ 1.07x slower │
│ QQuery 22 │   63.29 ms │                     64.94 ms │    no change │
│ QQuery 23 │  309.58 ms │                    356.43 ms │ 1.15x slower │
│ QQuery 24 │  194.05 ms │                    207.06 ms │ 1.07x slower │
│ QQuery 25 │  109.94 ms │                    110.35 ms │    no change │
│ QQuery 26 │   47.82 ms │                     50.95 ms │ 1.07x slower │
│ QQuery 27 │    6.17 ms │                      6.89 ms │ 1.12x slower │
│ QQuery 28 │   59.88 ms │                     58.63 ms │    no change │
│ QQuery 29 │   97.07 ms │                     97.81 ms │    no change │
│ QQuery 30 │   32.17 ms │                     36.32 ms │ 1.13x slower │
│ QQuery 31 │  110.95 ms │                    119.25 ms │ 1.07x slower │
│ QQuery 32 │   20.27 ms │                     21.39 ms │ 1.06x slower │
│ QQuery 33 │   37.53 ms │                     39.53 ms │ 1.05x slower │
│ QQuery 34 │    9.79 ms │                     10.46 ms │ 1.07x slower │
│ QQuery 35 │   72.13 ms │                     76.37 ms │ 1.06x slower │
│ QQuery 36 │    5.75 ms │                      6.29 ms │ 1.09x slower │
│ QQuery 37 │    6.81 ms │                      7.00 ms │    no change │
│ QQuery 38 │   61.60 ms │                     66.68 ms │ 1.08x slower │
│ QQuery 39 │   89.42 ms │                    100.98 ms │ 1.13x slower │
│ QQuery 40 │   23.71 ms │                     25.22 ms │ 1.06x slower │
│ QQuery 41 │   11.26 ms │                     11.59 ms │    no change │
│ QQuery 42 │   23.25 ms │                     24.45 ms │ 1.05x slower │
│ QQuery 43 │    5.34 ms │                      6.40 ms │ 1.20x slower │
│ QQuery 44 │    9.35 ms │                     10.33 ms │ 1.10x slower │
│ QQuery 45 │   37.98 ms │                     40.25 ms │ 1.06x slower │
│ QQuery 46 │   11.67 ms │                     12.79 ms │ 1.10x slower │
│ QQuery 47 │  225.04 ms │                    248.04 ms │ 1.10x slower │
│ QQuery 48 │   95.31 ms │                     96.49 ms │    no change │
│ QQuery 49 │   70.90 ms │                     76.76 ms │ 1.08x slower │
│ QQuery 50 │   59.22 ms │                     63.91 ms │ 1.08x slower │
│ QQuery 51 │   91.49 ms │                    100.92 ms │ 1.10x slower │
│ QQuery 52 │   24.53 ms │                     25.63 ms │    no change │
│ QQuery 53 │   29.12 ms │                     30.24 ms │    no change │
│ QQuery 54 │   54.48 ms │                     56.99 ms │    no change │
│ QQuery 55 │   23.24 ms │                     24.86 ms │ 1.07x slower │
│ QQuery 56 │   39.17 ms │                     40.79 ms │    no change │
│ QQuery 57 │  174.96 ms │                    188.12 ms │ 1.08x slower │
│ QQuery 58 │  112.71 ms │                    114.56 ms │    no change │
│ QQuery 59 │  117.39 ms │                    127.33 ms │ 1.08x slower │
│ QQuery 60 │   39.42 ms │                     40.90 ms │    no change │
│ QQuery 61 │   12.22 ms │                     12.78 ms │    no change │
│ QQuery 62 │   46.26 ms │                     48.99 ms │ 1.06x slower │
│ QQuery 63 │   29.34 ms │                     30.42 ms │    no change │
│ QQuery 64 │  363.76 ms │                    387.15 ms │ 1.06x slower │
│ QQuery 65 │  123.11 ms │                    129.36 ms │ 1.05x slower │
│ QQuery 66 │   81.47 ms │                     89.47 ms │ 1.10x slower │
│ QQuery 67 │  243.51 ms │                    257.07 ms │ 1.06x slower │
│ QQuery 68 │   11.87 ms │                     12.64 ms │ 1.07x slower │
│ QQuery 69 │   56.59 ms │                     59.36 ms │    no change │
│ QQuery 70 │  107.15 ms │                    107.39 ms │    no change │
│ QQuery 71 │   35.25 ms │                     37.59 ms │ 1.07x slower │
│ QQuery 72 │ 1885.68 ms │                   1822.47 ms │    no change │
│ QQuery 73 │    9.73 ms │                     10.89 ms │ 1.12x slower │
│ QQuery 74 │  167.65 ms │                    207.35 ms │ 1.24x slower │
│ QQuery 75 │  145.27 ms │                    173.80 ms │ 1.20x slower │
│ QQuery 76 │   35.21 ms │                     43.07 ms │ 1.22x slower │
│ QQuery 77 │   62.66 ms │                     69.68 ms │ 1.11x slower │
│ QQuery 78 │  221.25 ms │                    255.53 ms │ 1.15x slower │
│ QQuery 79 │   66.72 ms │                     74.84 ms │ 1.12x slower │
│ QQuery 80 │  100.36 ms │                    111.56 ms │ 1.11x slower │
│ QQuery 81 │   26.12 ms │                     32.30 ms │ 1.24x slower │
│ QQuery 82 │   16.27 ms │                     17.72 ms │ 1.09x slower │
│ QQuery 83 │   33.78 ms │                     37.45 ms │ 1.11x slower │
│ QQuery 84 │   29.24 ms │                     30.36 ms │    no change │
│ QQuery 85 │  102.88 ms │                    106.19 ms │    no change │
│ QQuery 86 │   25.09 ms │                     26.91 ms │ 1.07x slower │
│ QQuery 87 │   61.48 ms │                     68.54 ms │ 1.11x slower │
│ QQuery 88 │   63.70 ms │                     64.47 ms │    no change │
│ QQuery 89 │   35.93 ms │                     37.86 ms │ 1.05x slower │
│ QQuery 90 │   16.77 ms │                     17.55 ms │    no change │
│ QQuery 91 │   45.05 ms │                     47.72 ms │ 1.06x slower │
│ QQuery 92 │   29.83 ms │                     31.53 ms │ 1.06x slower │
│ QQuery 93 │   49.39 ms │                     52.85 ms │ 1.07x slower │
│ QQuery 94 │   38.43 ms │                     41.36 ms │ 1.08x slower │
│ QQuery 95 │   80.15 ms │                     85.68 ms │ 1.07x slower │
│ QQuery 96 │   23.76 ms │                     25.22 ms │ 1.06x slower │
│ QQuery 97 │   52.39 ms │                     59.22 ms │ 1.13x slower │
│ QQuery 98 │   42.69 ms │                     48.48 ms │ 1.14x slower │
│ QQuery 99 │   70.19 ms │                     76.42 ms │ 1.09x slower │
└───────────┴────────────┴──────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 9361.64ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 9866.16ms │
│ Average Time (HEAD)                         │   94.56ms │
│ Average Time (add-blocks-impl-from-scratch) │   99.66ms │
│ Queries Faster                              │         0 │
│ Queries Slower                              │        66 │
│ Queries with No Change                      │        33 │
│ Queries with Failure                        │         0 │
└─────────────────────────────────────────────┴───────────┘

Distribution per query (min / mean ±stddev / max):

Comparing HEAD and add-blocks-impl-from-scratch
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃          add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 1  │           5.74 / 6.21 ±0.91 / 8.04 ms │           6.74 / 7.23 ±0.87 / 8.97 ms │ 1.16x slower │
│ QQuery 2  │        80.08 / 80.42 ±0.31 / 80.87 ms │        86.41 / 86.79 ±0.33 / 87.23 ms │ 1.08x slower │
│ QQuery 3  │        28.60 / 28.84 ±0.18 / 29.16 ms │        30.02 / 30.46 ±0.36 / 31.09 ms │ 1.06x slower │
│ QQuery 4  │     477.71 / 485.83 ±4.56 / 491.31 ms │     508.26 / 519.33 ±6.49 / 526.35 ms │ 1.07x slower │
│ QQuery 5  │        51.74 / 54.64 ±4.46 / 63.52 ms │        55.63 / 56.30 ±0.71 / 57.56 ms │    no change │
│ QQuery 6  │        35.61 / 36.44 ±0.48 / 37.11 ms │        37.68 / 38.09 ±0.33 / 38.66 ms │    no change │
│ QQuery 7  │        74.17 / 75.05 ±0.87 / 76.11 ms │        77.51 / 80.17 ±4.17 / 88.49 ms │ 1.07x slower │
│ QQuery 8  │        36.92 / 39.17 ±4.33 / 47.83 ms │        37.92 / 38.18 ±0.15 / 38.36 ms │    no change │
│ QQuery 9  │        52.95 / 54.46 ±1.04 / 55.96 ms │        52.94 / 54.82 ±2.43 / 59.63 ms │    no change │
│ QQuery 10 │        62.48 / 62.75 ±0.25 / 63.19 ms │        64.66 / 65.05 ±0.29 / 65.54 ms │    no change │
│ QQuery 11 │     296.17 / 300.46 ±4.85 / 306.91 ms │     323.82 / 329.21 ±3.27 / 333.27 ms │ 1.10x slower │
│ QQuery 12 │        28.87 / 29.06 ±0.13 / 29.27 ms │        30.11 / 30.52 ±0.30 / 30.99 ms │ 1.05x slower │
│ QQuery 13 │     117.50 / 118.53 ±0.82 / 119.84 ms │     118.68 / 120.62 ±1.93 / 124.01 ms │    no change │
│ QQuery 14 │     413.95 / 418.20 ±3.68 / 424.16 ms │     428.71 / 433.87 ±4.57 / 441.08 ms │    no change │
│ QQuery 15 │        56.38 / 56.99 ±0.51 / 57.85 ms │        58.07 / 58.87 ±0.58 / 59.56 ms │    no change │
│ QQuery 16 │           6.80 / 7.02 ±0.27 / 7.53 ms │           7.18 / 7.34 ±0.20 / 7.74 ms │    no change │
│ QQuery 17 │        78.68 / 79.95 ±1.24 / 81.95 ms │        81.46 / 83.49 ±2.02 / 87.21 ms │    no change │
│ QQuery 18 │     101.81 / 104.54 ±1.44 / 106.12 ms │     107.92 / 108.42 ±0.35 / 108.88 ms │    no change │
│ QQuery 19 │        41.25 / 41.83 ±0.38 / 42.35 ms │        43.16 / 44.25 ±1.32 / 46.78 ms │ 1.06x slower │
│ QQuery 20 │        35.12 / 35.72 ±0.40 / 36.21 ms │        37.86 / 38.60 ±0.52 / 39.45 ms │ 1.08x slower │
│ QQuery 21 │        17.10 / 17.55 ±0.36 / 18.20 ms │        18.27 / 18.52 ±0.21 / 18.88 ms │ 1.06x slower │
│ QQuery 22 │        63.29 / 64.68 ±1.14 / 66.47 ms │        64.94 / 65.90 ±0.53 / 66.46 ms │    no change │
│ QQuery 23 │     309.58 / 313.23 ±2.32 / 315.80 ms │     356.43 / 358.02 ±1.54 / 360.80 ms │ 1.14x slower │
│ QQuery 24 │     194.05 / 196.20 ±1.70 / 198.91 ms │     207.06 / 209.97 ±2.81 / 213.92 ms │ 1.07x slower │
│ QQuery 25 │     109.94 / 111.32 ±0.92 / 112.71 ms │     110.35 / 111.20 ±0.76 / 112.29 ms │    no change │
│ QQuery 26 │        47.82 / 48.42 ±0.51 / 49.20 ms │        50.95 / 51.96 ±0.79 / 53.12 ms │ 1.07x slower │
│ QQuery 27 │           6.17 / 6.31 ±0.16 / 6.59 ms │           6.89 / 7.05 ±0.21 / 7.47 ms │ 1.12x slower │
│ QQuery 28 │        59.88 / 60.56 ±0.66 / 61.68 ms │        58.63 / 60.63 ±1.04 / 61.62 ms │    no change │
│ QQuery 29 │        97.07 / 97.92 ±0.73 / 98.99 ms │      97.81 / 100.49 ±3.29 / 106.87 ms │    no change │
│ QQuery 30 │        32.17 / 32.46 ±0.21 / 32.80 ms │        36.32 / 36.52 ±0.17 / 36.70 ms │ 1.13x slower │
│ QQuery 31 │     110.95 / 112.49 ±1.22 / 114.43 ms │     119.25 / 119.71 ±0.33 / 120.26 ms │ 1.06x slower │
│ QQuery 32 │        20.27 / 20.65 ±0.20 / 20.82 ms │        21.39 / 21.57 ±0.11 / 21.71 ms │    no change │
│ QQuery 33 │        37.53 / 38.30 ±0.71 / 39.50 ms │        39.53 / 39.92 ±0.27 / 40.32 ms │    no change │
│ QQuery 34 │         9.79 / 10.20 ±0.37 / 10.67 ms │        10.46 / 10.67 ±0.23 / 11.09 ms │    no change │
│ QQuery 35 │        72.13 / 72.53 ±0.43 / 73.19 ms │        76.37 / 76.84 ±0.38 / 77.48 ms │ 1.06x slower │
│ QQuery 36 │           5.75 / 5.85 ±0.13 / 6.11 ms │           6.29 / 6.43 ±0.19 / 6.80 ms │ 1.10x slower │
│ QQuery 37 │           6.81 / 6.92 ±0.09 / 7.09 ms │           7.00 / 7.15 ±0.09 / 7.27 ms │    no change │
│ QQuery 38 │        61.60 / 63.98 ±1.40 / 65.97 ms │        66.68 / 67.13 ±0.45 / 67.92 ms │    no change │
│ QQuery 39 │        89.42 / 90.40 ±1.01 / 92.29 ms │     100.98 / 102.34 ±1.36 / 104.87 ms │ 1.13x slower │
│ QQuery 40 │        23.71 / 23.94 ±0.18 / 24.15 ms │        25.22 / 25.72 ±0.31 / 26.15 ms │ 1.07x slower │
│ QQuery 41 │        11.26 / 11.41 ±0.13 / 11.65 ms │        11.59 / 11.75 ±0.15 / 12.01 ms │    no change │
│ QQuery 42 │        23.25 / 23.72 ±0.35 / 24.34 ms │        24.45 / 24.72 ±0.25 / 25.18 ms │    no change │
│ QQuery 43 │           5.34 / 5.45 ±0.15 / 5.75 ms │           6.40 / 6.53 ±0.17 / 6.86 ms │ 1.20x slower │
│ QQuery 44 │         9.35 / 10.70 ±2.35 / 15.39 ms │        10.33 / 10.47 ±0.10 / 10.63 ms │    no change │
│ QQuery 45 │        37.98 / 39.38 ±0.74 / 40.11 ms │        40.25 / 41.89 ±2.62 / 47.10 ms │ 1.06x slower │
│ QQuery 46 │        11.67 / 12.69 ±0.99 / 14.46 ms │        12.79 / 13.17 ±0.33 / 13.54 ms │    no change │
│ QQuery 47 │     225.04 / 230.81 ±4.08 / 236.25 ms │     248.04 / 253.71 ±5.82 / 263.61 ms │ 1.10x slower │
│ QQuery 48 │        95.31 / 96.89 ±1.57 / 98.92 ms │        96.49 / 97.44 ±0.96 / 99.12 ms │    no change │
│ QQuery 49 │        70.90 / 73.08 ±3.63 / 80.33 ms │        76.76 / 79.58 ±5.08 / 89.72 ms │ 1.09x slower │
│ QQuery 50 │        59.22 / 59.47 ±0.28 / 60.01 ms │        63.91 / 64.44 ±0.41 / 64.95 ms │ 1.08x slower │
│ QQuery 51 │        91.49 / 93.56 ±1.54 / 95.92 ms │     100.92 / 103.61 ±3.13 / 109.71 ms │ 1.11x slower │
│ QQuery 52 │        24.53 / 24.80 ±0.37 / 25.52 ms │        25.63 / 25.81 ±0.18 / 26.07 ms │    no change │
│ QQuery 53 │        29.12 / 30.10 ±1.46 / 33.00 ms │        30.24 / 30.47 ±0.13 / 30.60 ms │    no change │
│ QQuery 54 │        54.48 / 54.63 ±0.22 / 55.07 ms │        56.99 / 57.21 ±0.22 / 57.58 ms │    no change │
│ QQuery 55 │        23.24 / 23.44 ±0.14 / 23.57 ms │        24.86 / 24.96 ±0.08 / 25.09 ms │ 1.06x slower │
│ QQuery 56 │        39.17 / 39.34 ±0.18 / 39.67 ms │        40.79 / 43.09 ±3.16 / 49.31 ms │ 1.10x slower │
│ QQuery 57 │     174.96 / 177.86 ±2.67 / 181.32 ms │     188.12 / 190.71 ±1.30 / 191.50 ms │ 1.07x slower │
│ QQuery 58 │     112.71 / 114.13 ±1.77 / 117.48 ms │     114.56 / 116.22 ±1.67 / 119.42 ms │    no change │
│ QQuery 59 │     117.39 / 117.90 ±0.42 / 118.61 ms │     127.33 / 128.78 ±2.50 / 133.78 ms │ 1.09x slower │
│ QQuery 60 │        39.42 / 40.71 ±2.18 / 45.06 ms │        40.90 / 41.83 ±0.86 / 43.32 ms │    no change │
│ QQuery 61 │        12.22 / 12.42 ±0.19 / 12.70 ms │        12.78 / 12.91 ±0.11 / 13.07 ms │    no change │
│ QQuery 62 │        46.26 / 47.14 ±1.02 / 49.06 ms │        48.99 / 49.49 ±0.44 / 50.25 ms │    no change │
│ QQuery 63 │        29.34 / 29.50 ±0.14 / 29.76 ms │        30.42 / 30.91 ±0.38 / 31.49 ms │    no change │
│ QQuery 64 │     363.76 / 369.85 ±5.13 / 377.61 ms │     387.15 / 392.73 ±4.17 / 397.62 ms │ 1.06x slower │
│ QQuery 65 │     123.11 / 124.93 ±2.22 / 129.04 ms │     129.36 / 131.67 ±2.19 / 135.56 ms │ 1.05x slower │
│ QQuery 66 │        81.47 / 82.33 ±0.78 / 83.56 ms │        89.47 / 90.08 ±0.53 / 90.76 ms │ 1.09x slower │
│ QQuery 67 │     243.51 / 252.09 ±5.31 / 258.71 ms │     257.07 / 261.65 ±4.58 / 267.81 ms │    no change │
│ QQuery 68 │        11.87 / 12.13 ±0.23 / 12.49 ms │        12.64 / 12.94 ±0.26 / 13.37 ms │ 1.07x slower │
│ QQuery 69 │        56.59 / 59.27 ±4.49 / 68.19 ms │        59.36 / 59.93 ±0.81 / 61.54 ms │    no change │
│ QQuery 70 │     107.15 / 111.88 ±4.47 / 118.53 ms │     107.39 / 111.17 ±3.45 / 116.33 ms │    no change │
│ QQuery 71 │        35.25 / 35.70 ±0.35 / 36.23 ms │        37.59 / 39.05 ±1.16 / 40.82 ms │ 1.09x slower │
│ QQuery 72 │ 1885.68 / 1902.57 ±16.10 / 1932.54 ms │ 1822.47 / 1924.21 ±90.74 / 2033.48 ms │    no change │
│ QQuery 73 │         9.73 / 10.19 ±0.52 / 11.18 ms │        10.89 / 12.56 ±1.75 / 15.60 ms │ 1.23x slower │
│ QQuery 74 │     167.65 / 170.70 ±3.42 / 177.06 ms │     207.35 / 210.27 ±1.97 / 212.68 ms │ 1.23x slower │
│ QQuery 75 │     145.27 / 146.70 ±0.81 / 147.72 ms │     173.80 / 174.99 ±0.80 / 175.93 ms │ 1.19x slower │
│ QQuery 76 │        35.21 / 37.47 ±3.54 / 44.47 ms │        43.07 / 44.41 ±2.26 / 48.92 ms │ 1.19x slower │
│ QQuery 77 │        62.66 / 63.83 ±1.59 / 66.96 ms │        69.68 / 70.65 ±0.76 / 71.72 ms │ 1.11x slower │
│ QQuery 78 │     221.25 / 225.86 ±2.97 / 229.31 ms │     255.53 / 258.36 ±2.87 / 262.15 ms │ 1.14x slower │
│ QQuery 79 │        66.72 / 66.97 ±0.18 / 67.27 ms │        74.84 / 75.37 ±0.55 / 76.11 ms │ 1.13x slower │
│ QQuery 80 │     100.36 / 104.28 ±5.17 / 114.25 ms │     111.56 / 114.39 ±3.03 / 119.85 ms │ 1.10x slower │
│ QQuery 81 │        26.12 / 26.27 ±0.12 / 26.41 ms │        32.30 / 32.58 ±0.30 / 32.97 ms │ 1.24x slower │
│ QQuery 82 │        16.27 / 16.66 ±0.36 / 17.32 ms │        17.72 / 17.96 ±0.27 / 18.44 ms │ 1.08x slower │
│ QQuery 83 │        33.78 / 34.09 ±0.20 / 34.36 ms │        37.45 / 40.58 ±5.57 / 51.71 ms │ 1.19x slower │
│ QQuery 84 │        29.24 / 29.55 ±0.25 / 29.96 ms │        30.36 / 30.79 ±0.23 / 31.03 ms │    no change │
│ QQuery 85 │     102.88 / 107.12 ±3.72 / 113.76 ms │     106.19 / 107.35 ±1.11 / 109.24 ms │    no change │
│ QQuery 86 │        25.09 / 25.31 ±0.12 / 25.44 ms │        26.91 / 28.48 ±2.86 / 34.20 ms │ 1.13x slower │
│ QQuery 87 │        61.48 / 61.85 ±0.52 / 62.86 ms │        68.54 / 70.03 ±1.21 / 72.00 ms │ 1.13x slower │
│ QQuery 88 │        63.70 / 65.22 ±2.67 / 70.55 ms │        64.47 / 64.88 ±0.33 / 65.29 ms │    no change │
│ QQuery 89 │        35.93 / 36.39 ±0.65 / 37.65 ms │        37.86 / 38.28 ±0.28 / 38.52 ms │ 1.05x slower │
│ QQuery 90 │        16.77 / 17.12 ±0.21 / 17.43 ms │        17.55 / 17.62 ±0.05 / 17.69 ms │    no change │
│ QQuery 91 │        45.05 / 45.73 ±0.38 / 46.12 ms │        47.72 / 49.81 ±2.95 / 55.60 ms │ 1.09x slower │
│ QQuery 92 │        29.83 / 30.18 ±0.29 / 30.61 ms │        31.53 / 32.32 ±0.44 / 32.80 ms │ 1.07x slower │
│ QQuery 93 │        49.39 / 51.12 ±1.61 / 53.96 ms │        52.85 / 53.98 ±0.80 / 55.34 ms │ 1.06x slower │
│ QQuery 94 │        38.43 / 39.99 ±2.42 / 44.78 ms │        41.36 / 41.61 ±0.31 / 42.21 ms │    no change │
│ QQuery 95 │        80.15 / 81.48 ±1.01 / 83.26 ms │        85.68 / 88.51 ±2.64 / 93.14 ms │ 1.09x slower │
│ QQuery 96 │        23.76 / 24.11 ±0.21 / 24.42 ms │        25.22 / 25.43 ±0.15 / 25.70 ms │ 1.05x slower │
│ QQuery 97 │        52.39 / 53.86 ±1.30 / 55.73 ms │        59.22 / 60.38 ±1.26 / 62.82 ms │ 1.12x slower │
│ QQuery 98 │        42.69 / 43.62 ±0.70 / 44.41 ms │        48.48 / 49.47 ±0.61 / 50.28 ms │ 1.13x slower │
│ QQuery 99 │        70.19 / 70.83 ±0.44 / 71.57 ms │        76.42 / 78.45 ±2.94 / 84.23 ms │ 1.11x slower │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │  9516.38ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 10101.99ms │
│ Average Time (HEAD)                         │    96.13ms │
│ Average Time (add-blocks-impl-from-scratch) │   102.04ms │
│ Queries Faster                              │          0 │
│ Queries Slower                              │         57 │
│ Queries with No Change                      │         42 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Resource Usage

tpcds — base (merge-base)

Metric	Value
Wall time	50.0s
Peak memory	1.9 GiB
Avg memory	1.3 GiB
CPU user	206.3s
CPU sys	5.6s
Peak spill	0 B

tpcds — branch

Metric	Value
Wall time	55.0s
Peak memory	2.1 GiB
Avg memory	1.4 GiB
CPU user	214.0s
CPU sys	12.4s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:46:21Z

🤖 Benchmark completed (GKE) | trigger

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

Comparing add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark clickbench_partitioned

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 add-blocks-impl-from-scratch
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0  │    1.24 ms │                      1.24 ms │    no change │
│ QQuery 1  │   11.91 ms │                     11.61 ms │    no change │
│ QQuery 2  │   36.39 ms │                     36.40 ms │    no change │
│ QQuery 3  │   30.92 ms │                     31.02 ms │    no change │
│ QQuery 4  │  219.61 ms │                    258.35 ms │ 1.18x slower │
│ QQuery 5  │  270.77 ms │                    682.57 ms │ 2.52x slower │
│ QQuery 6  │    1.27 ms │                      1.27 ms │    no change │
│ QQuery 7  │   13.12 ms │                     13.35 ms │    no change │
│ QQuery 8  │  321.83 ms │                    378.70 ms │ 1.18x slower │
│ QQuery 9  │  446.89 ms │                    452.20 ms │    no change │
│ QQuery 10 │   68.56 ms │                     78.52 ms │ 1.15x slower │
│ QQuery 11 │   80.98 ms │                     90.31 ms │ 1.12x slower │
│ QQuery 12 │  265.47 ms │                    679.10 ms │ 2.56x slower │
│ QQuery 13 │  360.34 ms │                    587.75 ms │ 1.63x slower │
│ QQuery 14 │  282.24 ms │                    321.54 ms │ 1.14x slower │
│ QQuery 15 │  264.57 ms │                    350.66 ms │ 1.33x slower │
│ QQuery 16 │  614.97 ms │                    792.60 ms │ 1.29x slower │
│ QQuery 17 │  612.91 ms │                    713.12 ms │ 1.16x slower │
│ QQuery 18 │ 1245.14 ms │                   1647.39 ms │ 1.32x slower │
│ QQuery 19 │   27.41 ms │                     27.38 ms │    no change │
│ QQuery 20 │  515.99 ms │                    519.52 ms │    no change │
│ QQuery 21 │  514.97 ms │                    520.28 ms │    no change │
│ QQuery 22 │  983.83 ms │                    983.35 ms │    no change │
│ QQuery 23 │ 3008.64 ms │                   3068.58 ms │    no change │
│ QQuery 24 │   40.78 ms │                     41.84 ms │    no change │
│ QQuery 25 │  109.95 ms │                    112.24 ms │    no change │
│ QQuery 26 │   40.65 ms │                     41.96 ms │    no change │
│ QQuery 27 │  511.85 ms │                    512.94 ms │    no change │
│ QQuery 28 │ 2923.80 ms │                   3091.30 ms │ 1.06x slower │
│ QQuery 29 │   40.63 ms │                     40.87 ms │    no change │
│ QQuery 30 │  297.78 ms │                    345.10 ms │ 1.16x slower │
│ QQuery 31 │  282.78 ms │                    363.54 ms │ 1.29x slower │
│ QQuery 32 │  910.76 ms │                   1607.40 ms │ 1.76x slower │
│ QQuery 33 │ 1443.13 ms │                   5299.24 ms │ 3.67x slower │
│ QQuery 34 │ 1471.09 ms │                   5277.08 ms │ 3.59x slower │
│ QQuery 35 │  276.73 ms │                    321.98 ms │ 1.16x slower │
│ QQuery 36 │   66.54 ms │                     89.80 ms │ 1.35x slower │
│ QQuery 37 │   36.16 ms │                     38.30 ms │ 1.06x slower │
│ QQuery 38 │   40.09 ms │                     41.26 ms │    no change │
│ QQuery 39 │  131.22 ms │                    140.86 ms │ 1.07x slower │
│ QQuery 40 │   14.06 ms │                     14.11 ms │    no change │
│ QQuery 41 │   13.39 ms │                     13.84 ms │    no change │
│ QQuery 42 │   13.08 ms │                     13.39 ms │    no change │
└───────────┴────────────┴──────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 18864.47ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 29653.84ms │
│ Average Time (HEAD)                         │   438.71ms │
│ Average Time (add-blocks-impl-from-scratch) │   689.62ms │
│ Queries Faster                              │          0 │
│ Queries Slower                              │         22 │
│ Queries with No Change                      │         21 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Distribution per query (min / mean ±stddev / max):

Comparing HEAD and add-blocks-impl-from-scratch
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃           add-blocks-impl-from-scratch ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │          1.24 / 4.05 ±5.55 / 15.15 ms │           1.24 / 3.95 ±5.36 / 14.66 ms │     no change │
│ QQuery 1  │        11.91 / 12.09 ±0.14 / 12.27 ms │         11.61 / 11.89 ±0.19 / 12.13 ms │     no change │
│ QQuery 2  │        36.39 / 36.77 ±0.35 / 37.30 ms │         36.40 / 36.69 ±0.21 / 37.03 ms │     no change │
│ QQuery 3  │        30.92 / 31.38 ±0.55 / 32.43 ms │         31.02 / 31.18 ±0.10 / 31.31 ms │     no change │
│ QQuery 4  │     219.61 / 224.00 ±3.54 / 229.80 ms │      258.35 / 261.32 ±2.16 / 264.85 ms │  1.17x slower │
│ QQuery 5  │     270.77 / 272.94 ±2.44 / 277.01 ms │      682.57 / 689.89 ±7.27 / 702.87 ms │  2.53x slower │
│ QQuery 6  │           1.27 / 1.40 ±0.20 / 1.81 ms │            1.27 / 1.43 ±0.23 / 1.88 ms │     no change │
│ QQuery 7  │        13.12 / 13.20 ±0.08 / 13.34 ms │         13.35 / 13.42 ±0.06 / 13.52 ms │     no change │
│ QQuery 8  │     321.83 / 326.71 ±2.76 / 329.69 ms │      378.70 / 384.00 ±3.02 / 386.50 ms │  1.18x slower │
│ QQuery 9  │     446.89 / 455.71 ±7.12 / 466.85 ms │      452.20 / 463.97 ±6.31 / 470.99 ms │     no change │
│ QQuery 10 │        68.56 / 70.58 ±1.63 / 72.82 ms │         78.52 / 81.77 ±3.25 / 87.16 ms │  1.16x slower │
│ QQuery 11 │        80.98 / 81.55 ±0.48 / 82.24 ms │         90.31 / 91.86 ±1.26 / 94.08 ms │  1.13x slower │
│ QQuery 12 │     265.47 / 268.92 ±2.57 / 272.82 ms │     679.10 / 691.17 ±11.22 / 709.91 ms │  2.57x slower │
│ QQuery 13 │     360.34 / 367.75 ±7.82 / 382.57 ms │     587.75 / 606.86 ±15.34 / 632.57 ms │  1.65x slower │
│ QQuery 14 │     282.24 / 283.76 ±1.17 / 285.45 ms │     321.54 / 334.85 ±19.18 / 372.96 ms │  1.18x slower │
│ QQuery 15 │     264.57 / 272.98 ±8.24 / 284.29 ms │     350.66 / 364.26 ±14.05 / 391.47 ms │  1.33x slower │
│ QQuery 16 │     614.97 / 618.76 ±3.34 / 624.87 ms │     792.60 / 812.21 ±14.98 / 830.44 ms │  1.31x slower │
│ QQuery 17 │     612.91 / 617.75 ±3.20 / 621.60 ms │     713.12 / 762.51 ±53.07 / 838.46 ms │  1.23x slower │
│ QQuery 18 │ 1245.14 / 1271.66 ±15.83 / 1293.99 ms │  1647.39 / 1743.71 ±64.67 / 1819.54 ms │  1.37x slower │
│ QQuery 19 │        27.41 / 28.57 ±1.06 / 30.51 ms │         27.38 / 33.22 ±6.89 / 42.87 ms │  1.16x slower │
│ QQuery 20 │    515.99 / 525.44 ±10.42 / 543.12 ms │      519.52 / 529.59 ±9.93 / 546.01 ms │     no change │
│ QQuery 21 │     514.97 / 519.59 ±3.08 / 522.76 ms │      520.28 / 523.47 ±2.80 / 528.38 ms │     no change │
│ QQuery 22 │     983.83 / 989.22 ±2.83 / 991.42 ms │     983.35 / 991.19 ±9.27 / 1003.46 ms │     no change │
│ QQuery 23 │ 3008.64 / 3044.86 ±28.92 / 3095.84 ms │  3068.58 / 3105.88 ±26.75 / 3151.95 ms │     no change │
│ QQuery 24 │        40.78 / 42.46 ±2.61 / 47.63 ms │        41.84 / 52.62 ±12.81 / 73.14 ms │  1.24x slower │
│ QQuery 25 │     109.95 / 112.58 ±2.02 / 116.14 ms │      112.24 / 113.20 ±1.12 / 115.36 ms │     no change │
│ QQuery 26 │        40.65 / 41.30 ±0.61 / 42.44 ms │         41.96 / 45.52 ±5.74 / 56.89 ms │  1.10x slower │
│ QQuery 27 │     511.85 / 523.09 ±8.76 / 533.04 ms │     512.94 / 533.12 ±16.01 / 559.35 ms │     no change │
│ QQuery 28 │  2923.80 / 2930.65 ±6.46 / 2939.03 ms │  3091.30 / 3124.59 ±18.74 / 3148.37 ms │  1.07x slower │
│ QQuery 29 │      40.63 / 63.06 ±44.01 / 151.08 ms │         40.87 / 41.52 ±0.76 / 42.46 ms │ +1.52x faster │
│ QQuery 30 │     297.78 / 306.49 ±6.85 / 317.31 ms │     345.10 / 366.45 ±18.68 / 398.59 ms │  1.20x slower │
│ QQuery 31 │    282.78 / 302.74 ±21.91 / 345.47 ms │     363.54 / 389.85 ±24.26 / 435.88 ms │  1.29x slower │
│ QQuery 32 │   910.76 / 948.00 ±41.10 / 1025.05 ms │  1607.40 / 1658.37 ±28.39 / 1686.93 ms │  1.75x slower │
│ QQuery 33 │ 1443.13 / 1488.44 ±33.80 / 1533.92 ms │ 5299.24 / 5447.36 ±132.45 / 5673.80 ms │  3.66x slower │
│ QQuery 34 │ 1471.09 / 1540.20 ±69.70 / 1668.43 ms │ 5277.08 / 5441.99 ±142.17 / 5673.90 ms │  3.53x slower │
│ QQuery 35 │    276.73 / 303.35 ±23.32 / 338.83 ms │     321.98 / 341.81 ±17.59 / 363.92 ms │  1.13x slower │
│ QQuery 36 │        66.54 / 72.24 ±5.94 / 81.73 ms │         89.80 / 91.96 ±1.98 / 95.65 ms │  1.27x slower │
│ QQuery 37 │        36.16 / 38.48 ±3.12 / 44.65 ms │         38.30 / 39.10 ±0.66 / 39.89 ms │     no change │
│ QQuery 38 │        40.09 / 42.98 ±2.20 / 46.02 ms │         41.26 / 42.76 ±0.95 / 44.15 ms │     no change │
│ QQuery 39 │     131.22 / 146.33 ±8.83 / 156.47 ms │     140.86 / 154.40 ±13.80 / 174.81 ms │  1.06x slower │
│ QQuery 40 │        14.06 / 14.80 ±1.25 / 17.29 ms │         14.11 / 14.59 ±0.36 / 15.16 ms │     no change │
│ QQuery 41 │        13.39 / 15.20 ±2.89 / 20.94 ms │         13.84 / 14.07 ±0.21 / 14.46 ms │ +1.08x faster │
│ QQuery 42 │        13.08 / 13.31 ±0.24 / 13.74 ms │         13.39 / 13.47 ±0.10 / 13.67 ms │     no change │
└───────────┴───────────────────────────────────────┴────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 19285.33ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 30497.05ms │
│ Average Time (HEAD)                         │   448.50ms │
│ Average Time (add-blocks-impl-from-scratch) │   709.23ms │
│ Queries Faster                              │          2 │
│ Queries Slower                              │         24 │
│ Queries with No Change                      │         17 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Resource Usage

clickbench_partitioned — base (merge-base)

Metric	Value
Wall time	100.0s
Peak memory	10.2 GiB
Avg memory	4.3 GiB
CPU user	986.9s
CPU sys	69.4s
Peak spill	0 B

clickbench_partitioned — branch

Metric	Value
Wall time	155.0s
Peak memory	9.3 GiB
Avg memory	4.6 GiB
CPU user	1430.2s
CPU sys	154.6s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:46:40Z

🤖 Benchmark completed (GKE) | trigger

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

Comparing add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark h2o_medium

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 add-blocks-impl-from-scratch
--------------------
Benchmark h2o.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 1  │  988.45 ms │                    992.59 ms │    no change │
│ QQuery 2  │ 2152.25 ms │                   2596.72 ms │ 1.21x slower │
│ QQuery 3  │ 1864.43 ms │                   2068.10 ms │ 1.11x slower │
│ QQuery 4  │ 1323.83 ms │                   1338.55 ms │    no change │
│ QQuery 5  │ 1771.14 ms │                   1837.23 ms │    no change │
│ QQuery 6  │ 1521.81 ms │                   1614.46 ms │ 1.06x slower │
│ QQuery 7  │ 1763.97 ms │                   1936.35 ms │ 1.10x slower │
│ QQuery 8  │ 3684.43 ms │                   3682.41 ms │    no change │
│ QQuery 9  │ 2657.08 ms │                   3399.74 ms │ 1.28x slower │
│ QQuery 10 │ 2814.31 ms │                   2925.67 ms │    no change │
└───────────┴────────────┴──────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 20541.71ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 22391.81ms │
│ Average Time (HEAD)                         │  2054.17ms │
│ Average Time (add-blocks-impl-from-scratch) │  2239.18ms │
│ Queries Faster                              │          0 │
│ Queries Slower                              │          5 │
│ Queries with No Change                      │          5 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Distribution per query (min / mean ±stddev / max):

Comparing HEAD and add-blocks-impl-from-scratch
--------------------
Benchmark h2o.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃          add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 1  │   988.45 / 999.61 ±12.54 / 1017.12 ms │     992.59 / 994.69 ±1.54 / 996.26 ms │    no change │
│ QQuery 2  │ 2152.25 / 2206.02 ±42.74 / 2256.81 ms │ 2596.72 / 2618.05 ±18.18 / 2641.15 ms │ 1.19x slower │
│ QQuery 3  │ 1864.43 / 1896.11 ±22.41 / 1912.27 ms │ 2068.10 / 2083.26 ±14.01 / 2101.88 ms │ 1.10x slower │
│ QQuery 4  │  1323.83 / 1324.95 ±1.17 / 1326.57 ms │  1338.55 / 1338.99 ±0.47 / 1339.64 ms │    no change │
│ QQuery 5  │ 1771.14 / 1808.50 ±35.31 / 1855.88 ms │ 1837.23 / 1895.05 ±40.98 / 1927.18 ms │    no change │
│ QQuery 6  │  1521.81 / 1524.16 ±2.72 / 1527.97 ms │ 1614.46 / 1635.80 ±18.96 / 1660.53 ms │ 1.07x slower │
│ QQuery 7  │  1763.97 / 1773.25 ±6.57 / 1778.47 ms │ 1936.35 / 1955.83 ±14.38 / 1970.64 ms │ 1.10x slower │
│ QQuery 8  │  3684.43 / 3695.85 ±8.08 / 3701.90 ms │ 3682.41 / 3743.34 ±66.99 / 3836.64 ms │    no change │
│ QQuery 9  │ 2657.08 / 2681.64 ±22.26 / 2710.97 ms │ 3399.74 / 3449.14 ±63.32 / 3538.52 ms │ 1.29x slower │
│ QQuery 10 │ 2814.31 / 2828.52 ±12.37 / 2844.46 ms │ 2925.67 / 2951.59 ±19.59 / 2973.01 ms │    no change │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 20738.60ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 22665.75ms │
│ Average Time (HEAD)                         │  2073.86ms │
│ Average Time (add-blocks-impl-from-scratch) │  2266.57ms │
│ Queries Faster                              │          0 │
│ Queries Slower                              │          5 │
│ Queries with No Change                      │          5 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Resource Usage

h2o_medium — base (merge-base)

Metric	Value
Wall time	65.0s
Peak memory	10.0 GiB
Avg memory	2.9 GiB
CPU user	649.4s
CPU sys	55.6s
Peak spill	0 B

h2o_medium — branch

Metric	Value
Wall time	75.0s
Peak memory	11.3 GiB
Avg memory	2.6 GiB
CPU user	701.2s
CPU sys	68.7s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-09-06T18:59:15Z

🤖 Benchmark completed (GKE) | trigger

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

Comparing add-blocks-impl-from-scratch (10864ee) to 5bf6aef (merge-base) diff

Run configuration

run benchmark external_aggr

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 add-blocks-impl-from-scratch
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query        ┃      HEAD ┃ add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ Q1(64.0 MB)  │  51.02 ms │                     96.31 ms │ 1.89x slower │
│ Q1(32.0 MB)  │  50.37 ms │                     70.62 ms │ 1.40x slower │
│ Q1(16.0 MB)  │  46.37 ms │                     65.43 ms │ 1.41x slower │
│ Q2(512.0 MB) │ 275.13 ms │                    708.34 ms │ 2.57x slower │
│ Q2(256.0 MB) │ 283.48 ms │                    550.36 ms │ 1.94x slower │
│ Q2(128.0 MB) │ 244.76 ms │                    438.86 ms │ 1.79x slower │
│ Q2(64.0 MB)  │ 242.06 ms │                    396.89 ms │ 1.64x slower │
│ Q2(32.0 MB)  │ 305.21 ms │                    427.38 ms │ 1.40x slower │
└──────────────┴───────────┴──────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 1498.39ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 2754.20ms │
│ Average Time (HEAD)                         │  187.30ms │
│ Average Time (add-blocks-impl-from-scratch) │  344.27ms │
│ Queries Faster                              │         0 │
│ Queries Slower                              │         8 │
│ Queries with No Change                      │         0 │
│ Queries with Failure                        │         0 │
└─────────────────────────────────────────────┴───────────┘

Distribution per query (min / mean ±stddev / max):

Comparing HEAD and add-blocks-impl-from-scratch
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query        ┃                               HEAD ┃       add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ Q1(64.0 MB)  │     51.02 / 54.20 ±3.78 / 61.51 ms │   96.31 / 103.41 ±9.18 / 121.39 ms │ 1.91x slower │
│ Q1(32.0 MB)  │     50.37 / 52.40 ±1.66 / 54.86 ms │     70.62 / 73.51 ±2.52 / 77.93 ms │ 1.40x slower │
│ Q1(16.0 MB)  │     46.37 / 49.85 ±2.20 / 52.77 ms │     65.43 / 68.14 ±1.97 / 71.18 ms │ 1.37x slower │
│ Q2(512.0 MB) │  275.13 / 291.74 ±9.51 / 301.36 ms │ 708.34 / 721.53 ±16.59 / 753.30 ms │ 2.47x slower │
│ Q2(256.0 MB) │ 283.48 / 298.53 ±10.29 / 311.70 ms │  550.36 / 556.31 ±8.56 / 573.23 ms │ 1.86x slower │
│ Q2(128.0 MB) │  244.76 / 246.20 ±1.49 / 248.89 ms │ 438.86 / 469.35 ±18.89 / 498.07 ms │ 1.91x slower │
│ Q2(64.0 MB)  │  242.06 / 246.27 ±3.67 / 252.36 ms │  396.89 / 400.95 ±3.41 / 405.59 ms │ 1.63x slower │
│ Q2(32.0 MB)  │  305.21 / 310.46 ±3.31 / 314.68 ms │  427.38 / 436.10 ±5.74 / 443.34 ms │ 1.40x slower │
└──────────────┴────────────────────────────────────┴────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 1549.64ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 2829.30ms │
│ Average Time (HEAD)                         │  193.70ms │
│ Average Time (add-blocks-impl-from-scratch) │  353.66ms │
│ Queries Faster                              │         0 │
│ Queries Slower                              │         8 │
│ Queries with No Change                      │         0 │
│ Queries with Failure                        │         0 │
└─────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only when the benchmark runs with DATAFUSION_RUNTIME_MEMORY_LIMIT set.

Base: 5bf6aef (merge-base) | Changed: add-blocks-impl-from-scratch

external_aggr

Query	Base	Changed	Change
1(64.0 MB)	36.8 MiB	47.9 MiB	+30.4%
1(32.0 MB)	17.2 MiB	31.0 MiB	+80.4%
1(16.0 MB)	11.3 MiB	16.1 MiB	+42.7%
2(512.0 MB)	137.3 MiB	369.0 MiB	+168.8%
2(256.0 MB)	111.1 MiB	256.0 MiB	+130.3%
2(128.0 MB)	49.2 MiB	128.0 MiB	+160.4%
2(64.0 MB)	29.1 MiB	64.0 MiB	+120.1%
2(32.0 MB)	30.0 MiB	33.0 MiB	+9.9%

Pool accounting vs. process RSS

Max pool peak is the largest reservation any single query in the run reached; peak RSS covers the whole invocation, including data loading and allocator retention, and the two high-water marks need not coincide in time. The gap is therefore an upper bound on what the pool did not account for, not a measurement of it.

Benchmark	Side	Max pool peak	Peak RSS	Gap	RSS / pool
external_aggr	base (`5bf6aef (merge-base)`)	137.3 MiB	401.6 MiB	264.3 MiB	2.9×
external_aggr	changed (`add-blocks-impl-from-scratch`)	369.0 MiB	538.1 MiB	169.1 MiB	1.5×

Resource Usage

external_aggr — base (merge-base)

Metric	Value
Wall time	545.1s
Peak memory	401.6 MiB
Avg memory	8.1 MiB
CPU user	25.5s
CPU sys	3.8s
Peak spill	0 B

external_aggr — branch

Metric	Value
Wall time	565.1s
Peak memory	538.1 MiB
Avg memory	15.5 MiB
CPU user	50.7s
CPU sys	10.2s
Peak spill	0 B

File an issue against this benchmark runner

…or blocked binary view map

rluvaton · 2026-09-07T09:24:58Z

run benchmark clickbench_partitioned

env:
  DEBUG: true

adriangbot · 2026-09-07T09:27:46Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5568473537-2202-s8n8t 6.12.94+ #1 SMP Fri Jul 17 09:42:57 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 add-blocks-impl-from-scratch (4d3c197) to 5bf6aef (merge-base) diff

Run configuration

run benchmark clickbench_partitioned
env:
  DEBUG: "true"

Results will be posted here when complete

File an issue against this benchmark runner

rluvaton · 2026-09-07T09:28:20Z

run benchmark external_aggr

env:
  DEBUG: true

adriangbot · 2026-09-07T09:31:24Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5568519031-2203-7x9hq 6.12.94+ #1 SMP Fri Jul 17 09:42:57 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 add-blocks-impl-from-scratch (4d3c197) to 5bf6aef (merge-base) diff

Run configuration

run benchmark external_aggr
env:
  DEBUG: "true"

Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-07T09:46:33Z

Benchmark for this request failed before finishing (Kubernetes reason: BackoffLimitExceeded).

Benchmarks requested: clickbench_partitioned

Runner log (last 40 lines)

warning: methods `is_empty` and `start_new_block` are never used
   --> datafusion/physical-plan/src/aggregates_blocked/group_values/multi_group_by/mod.rs:109:8
    |
 73 | pub trait BlockedGroupColumn<const IS_FIXED_GROUP_SIZE: bool>: Send + Sync {
    |           ------------------ methods in this trait
...
109 |     fn is_empty(&self) -> bool {
    |        ^^^^^^^^
...
132 |     fn start_new_block(&mut self);
    |        ^^^^^^^^^^^^^^^

warning: field `batch_size` is never read
   --> datafusion/physical-plan/src/aggregates_blocked/hash_stream.rs:150:5
    |
142 | pub(crate) struct PartialHashAggregateStream {
    |                   -------------------------- field in this struct
...
150 |     batch_size: usize,
    |     ^^^^^^^^^^

warning: `datafusion-physical-plan` (lib) generated 42 warnings (run `cargo fix --lib -p datafusion-physical-plan` to apply 24 suggestions)
    Finished `release` profile [optimized] target(s) in 0.18s
     Running `/workspace/datafusion-branch/target/release/dfbench clickbench --iterations 5 --path /workspace/datafusion-bench/benchmarks/data/hits_partitioned --queries-path /workspace/datafusion-bench/benchmarks/queries/clickbench/queries -o /workspace/datafusion-bench/benchmarks/results/add-blocks-impl-from-scratch/clickbench_partitioned.json`

thread 'tokio-rt-worker' (9937) panicked at datafusion/physical-expr-common/src/blocked_binary_view_map.rs:510:13:
assertion `left == right` failed
  left: 0
 right: 1
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

thread 'tokio-rt-worker' (9943) panicked at datafusion/physical-expr-common/src/blocked_binary_view_map.rs:510:13:
assertion `left == right` failed
  left: 0
 right: 1

thread 'tokio-rt-worker' (9942) panicked at datafusion/physical-expr-common/src/blocked_binary_view_map.rs:510:13:
assertion `left == right` failed
  left: 0
 right: 1

Kubernetes message

Job has reached the specified backoff limit

File an issue against this benchmark runner

rluvaton · 2026-09-07T09:56:39Z

run benchmark clickbench_partitioned

env:
  DEBUG: true

adriangbot · 2026-09-07T09:59:24Z

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5568868456-2204-tbrnh 6.12.94+ #1 SMP Fri Jul 17 09:42:57 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 add-blocks-impl-from-scratch (4b720d7) to 5bf6aef (merge-base) diff

Run configuration

run benchmark clickbench_partitioned
env:
  DEBUG: "true"

Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-07T10:02:05Z

🤖 Benchmark completed (GKE) | trigger

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

Comparing add-blocks-impl-from-scratch (4d3c197) to 5bf6aef (merge-base) diff

Run configuration

run benchmark external_aggr
env:
  DEBUG: "true"

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 add-blocks-impl-from-scratch
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query        ┃      HEAD ┃ add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ Q1(64.0 MB)  │  52.83 ms │                     95.23 ms │ 1.80x slower │
│ Q1(32.0 MB)  │  50.21 ms │                     73.29 ms │ 1.46x slower │
│ Q1(16.0 MB)  │  48.65 ms │                     64.57 ms │ 1.33x slower │
│ Q2(512.0 MB) │ 274.93 ms │                    718.04 ms │ 2.61x slower │
│ Q2(256.0 MB) │ 272.88 ms │                    553.05 ms │ 2.03x slower │
│ Q2(128.0 MB) │ 248.15 ms │                    427.39 ms │ 1.72x slower │
│ Q2(64.0 MB)  │ 246.61 ms │                    376.08 ms │ 1.52x slower │
│ Q2(32.0 MB)  │ 312.59 ms │                    415.77 ms │ 1.33x slower │
└──────────────┴───────────┴──────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 1506.85ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 2723.42ms │
│ Average Time (HEAD)                         │  188.36ms │
│ Average Time (add-blocks-impl-from-scratch) │  340.43ms │
│ Queries Faster                              │         0 │
│ Queries Slower                              │         8 │
│ Queries with No Change                      │         0 │
│ Queries with Failure                        │         0 │
└─────────────────────────────────────────────┴───────────┘

Distribution per query (min / mean ±stddev / max):

Comparing HEAD and add-blocks-impl-from-scratch
--------------------
Benchmark external_aggr.json
--------------------
┏━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query        ┃                               HEAD ┃       add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ Q1(64.0 MB)  │     52.83 / 57.20 ±5.01 / 66.91 ms │   95.23 / 103.10 ±7.45 / 116.28 ms │ 1.80x slower │
│ Q1(32.0 MB)  │     50.21 / 52.23 ±1.61 / 53.83 ms │     73.29 / 77.77 ±2.92 / 81.01 ms │ 1.49x slower │
│ Q1(16.0 MB)  │     48.65 / 50.95 ±1.67 / 53.80 ms │     64.57 / 67.36 ±2.33 / 70.62 ms │ 1.32x slower │
│ Q2(512.0 MB) │  274.93 / 289.70 ±7.89 / 297.30 ms │ 718.04 / 738.55 ±19.90 / 774.39 ms │ 2.55x slower │
│ Q2(256.0 MB) │ 272.88 / 291.83 ±18.67 / 323.81 ms │ 553.05 / 589.21 ±19.62 / 606.45 ms │ 2.02x slower │
│ Q2(128.0 MB) │  248.15 / 256.50 ±8.19 / 268.83 ms │  427.39 / 438.52 ±7.33 / 446.10 ms │ 1.71x slower │
│ Q2(64.0 MB)  │  246.61 / 247.70 ±0.83 / 248.76 ms │  376.08 / 382.63 ±4.84 / 390.94 ms │ 1.54x slower │
│ Q2(32.0 MB)  │  312.59 / 316.12 ±2.43 / 319.30 ms │  415.77 / 423.41 ±5.72 / 430.99 ms │ 1.34x slower │
└──────────────┴────────────────────────────────────┴────────────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 1562.23ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 2820.54ms │
│ Average Time (HEAD)                         │  195.28ms │
│ Average Time (add-blocks-impl-from-scratch) │  352.57ms │
│ Queries Faster                              │         0 │
│ Queries Slower                              │         8 │
│ Queries with No Change                      │         0 │
│ Queries with Failure                        │         0 │
└─────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only when the benchmark runs with DATAFUSION_RUNTIME_MEMORY_LIMIT set.

Base: 5bf6aef (merge-base) | Changed: add-blocks-impl-from-scratch

external_aggr

Query	Base	Changed	Change
1(64.0 MB)	36.8 MiB	48.0 MiB	+30.6%
1(32.0 MB)	18.6 MiB	31.0 MiB	+66.4%
1(16.0 MB)	11.4 MiB	16.1 MiB	+41.1%
2(512.0 MB)	136.8 MiB	368.8 MiB	+169.7%
2(256.0 MB)	97.4 MiB	256.0 MiB	+162.9%
2(128.0 MB)	49.1 MiB	128.0 MiB	+160.5%
2(64.0 MB)	29.2 MiB	64.0 MiB	+119.2%
2(32.0 MB)	30.0 MiB	33.0 MiB	+9.9%

Pool accounting vs. process RSS

Max pool peak is the largest reservation any single query in the run reached; peak RSS covers the whole invocation, including data loading and allocator retention, and the two high-water marks need not coincide in time. The gap is therefore an upper bound on what the pool did not account for, not a measurement of it.

Benchmark	Side	Max pool peak	Peak RSS	Gap	RSS / pool
external_aggr	base (`5bf6aef (merge-base)`)	136.8 MiB	507.2 MiB	370.4 MiB	3.7×
external_aggr	changed (`add-blocks-impl-from-scratch`)	368.8 MiB	653.8 MiB	284.9 MiB	1.8×

Resource Usage

external_aggr — base (merge-base)

Metric	Value
Wall time	540.1s
Peak memory	507.2 MiB
Avg memory	8.8 MiB
CPU user	25.6s
CPU sys	3.8s
Peak spill	0 B

external_aggr — branch

Metric	Value
Wall time	565.1s
Peak memory	653.8 MiB
Avg memory	15.0 MiB
CPU user	46.9s
CPU sys	9.1s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-09-07T10:19:42Z

🤖 Benchmark completed (GKE) | trigger

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

Comparing add-blocks-impl-from-scratch (4b720d7) to 5bf6aef (merge-base) diff

Run configuration

run benchmark clickbench_partitioned
env:
  DEBUG: "true"

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 add-blocks-impl-from-scratch
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ add-blocks-impl-from-scratch ┃       Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━┩
│ QQuery 0  │    1.27 ms │                      1.25 ms │    no change │
│ QQuery 1  │   12.33 ms │                     11.88 ms │    no change │
│ QQuery 2  │   36.72 ms │                     37.23 ms │    no change │
│ QQuery 3  │   31.21 ms │                     31.16 ms │    no change │
│ QQuery 4  │  228.73 ms │                    270.75 ms │ 1.18x slower │
│ QQuery 5  │  279.62 ms │                    369.84 ms │ 1.32x slower │
│ QQuery 6  │    1.29 ms │                      1.31 ms │    no change │
│ QQuery 7  │   13.47 ms │                     13.85 ms │    no change │
│ QQuery 8  │  328.77 ms │                    390.30 ms │ 1.19x slower │
│ QQuery 9  │  471.73 ms │                    478.13 ms │    no change │
│ QQuery 10 │   70.61 ms │                     80.63 ms │ 1.14x slower │
│ QQuery 11 │   82.70 ms │                     92.97 ms │ 1.12x slower │
│ QQuery 12 │  274.67 ms │                    365.99 ms │ 1.33x slower │
│ QQuery 13 │  370.54 ms │                    451.63 ms │ 1.22x slower │
│ QQuery 14 │  287.76 ms │                    333.27 ms │ 1.16x slower │
│ QQuery 15 │  275.73 ms │                    361.15 ms │ 1.31x slower │
│ QQuery 16 │  619.23 ms │                    814.71 ms │ 1.32x slower │
│ QQuery 17 │  625.40 ms │                    717.74 ms │ 1.15x slower │
│ QQuery 18 │ 1298.36 ms │                   1732.43 ms │ 1.33x slower │
│ QQuery 19 │   27.57 ms │                     27.94 ms │    no change │
│ QQuery 20 │  522.30 ms │                    515.47 ms │    no change │
│ QQuery 21 │  515.17 ms │                    522.78 ms │    no change │
│ QQuery 22 │  999.02 ms │                    995.81 ms │    no change │
│ QQuery 23 │ 3060.50 ms │                   3103.58 ms │    no change │
│ QQuery 24 │   41.30 ms │                     41.21 ms │    no change │
│ QQuery 25 │  111.23 ms │                    112.25 ms │    no change │
│ QQuery 26 │   41.34 ms │                     41.98 ms │    no change │
│ QQuery 27 │  515.76 ms │                    523.49 ms │    no change │
│ QQuery 28 │ 2932.13 ms │                   2999.71 ms │    no change │
│ QQuery 29 │   41.25 ms │                     41.38 ms │    no change │
│ QQuery 30 │  305.39 ms │                    348.72 ms │ 1.14x slower │
│ QQuery 31 │  281.09 ms │                    376.43 ms │ 1.34x slower │
│ QQuery 32 │  941.89 ms │                   1631.16 ms │ 1.73x slower │
│ QQuery 33 │ 1482.44 ms │                   2102.63 ms │ 1.42x slower │
│ QQuery 34 │ 1507.73 ms │                   1984.67 ms │ 1.32x slower │
│ QQuery 35 │  288.65 ms │                    323.69 ms │ 1.12x slower │
│ QQuery 36 │   67.21 ms │                     79.15 ms │ 1.18x slower │
│ QQuery 37 │   35.86 ms │                     38.32 ms │ 1.07x slower │
│ QQuery 38 │   39.96 ms │                     41.96 ms │ 1.05x slower │
│ QQuery 39 │  140.21 ms │                    152.30 ms │ 1.09x slower │
│ QQuery 40 │   14.36 ms │                     14.67 ms │    no change │
│ QQuery 41 │   13.85 ms │                     13.79 ms │    no change │
│ QQuery 42 │   13.70 ms │                     14.09 ms │    no change │
└───────────┴────────────┴──────────────────────────────┴──────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 19250.09ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 22603.39ms │
│ Average Time (HEAD)                         │   447.68ms │
│ Average Time (add-blocks-impl-from-scratch) │   525.66ms │
│ Queries Faster                              │          0 │
│ Queries Slower                              │         22 │
│ Queries with No Change                      │         21 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Distribution per query (min / mean ±stddev / max):

Comparing HEAD and add-blocks-impl-from-scratch
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃          add-blocks-impl-from-scratch ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │          1.27 / 4.15 ±5.58 / 15.31 ms │          1.25 / 4.13 ±5.65 / 15.42 ms │     no change │
│ QQuery 1  │        12.33 / 12.56 ±0.16 / 12.74 ms │        11.88 / 12.36 ±0.29 / 12.65 ms │     no change │
│ QQuery 2  │        36.72 / 37.02 ±0.25 / 37.35 ms │        37.23 / 37.50 ±0.23 / 37.83 ms │     no change │
│ QQuery 3  │        31.21 / 31.80 ±0.66 / 33.08 ms │        31.16 / 31.68 ±0.31 / 32.08 ms │     no change │
│ QQuery 4  │     228.73 / 234.97 ±4.38 / 240.34 ms │     270.75 / 271.83 ±1.68 / 275.17 ms │  1.16x slower │
│ QQuery 5  │     279.62 / 283.25 ±2.28 / 286.48 ms │     369.84 / 377.73 ±5.31 / 384.48 ms │  1.33x slower │
│ QQuery 6  │           1.29 / 1.34 ±0.06 / 1.46 ms │           1.31 / 1.36 ±0.07 / 1.49 ms │     no change │
│ QQuery 7  │        13.47 / 13.68 ±0.13 / 13.81 ms │        13.85 / 13.92 ±0.07 / 14.01 ms │     no change │
│ QQuery 8  │    328.77 / 347.26 ±10.16 / 359.32 ms │     390.30 / 393.68 ±3.19 / 399.36 ms │  1.13x slower │
│ QQuery 9  │     471.73 / 477.92 ±5.65 / 486.55 ms │     478.13 / 489.26 ±8.02 / 498.52 ms │     no change │
│ QQuery 10 │        70.61 / 71.73 ±1.09 / 73.72 ms │        80.63 / 81.25 ±0.76 / 82.71 ms │  1.13x slower │
│ QQuery 11 │        82.70 / 83.11 ±0.36 / 83.61 ms │        92.97 / 93.47 ±0.49 / 94.33 ms │  1.12x slower │
│ QQuery 12 │     274.67 / 281.63 ±3.88 / 285.76 ms │    365.99 / 376.08 ±11.79 / 397.89 ms │  1.34x slower │
│ QQuery 13 │    370.54 / 390.96 ±17.11 / 414.71 ms │     451.63 / 462.28 ±9.66 / 479.46 ms │  1.18x slower │
│ QQuery 14 │     287.76 / 294.83 ±5.32 / 304.02 ms │     333.27 / 335.52 ±1.84 / 338.65 ms │  1.14x slower │
│ QQuery 15 │    275.73 / 288.59 ±13.35 / 313.56 ms │     361.15 / 366.62 ±5.75 / 376.78 ms │  1.27x slower │
│ QQuery 16 │     619.23 / 628.07 ±7.20 / 639.26 ms │    814.71 / 839.79 ±23.34 / 883.85 ms │  1.34x slower │
│ QQuery 17 │     625.40 / 633.73 ±9.25 / 648.79 ms │    717.74 / 763.13 ±34.60 / 806.19 ms │  1.20x slower │
│ QQuery 18 │ 1298.36 / 1315.70 ±17.42 / 1346.37 ms │ 1732.43 / 1787.33 ±39.52 / 1826.54 ms │  1.36x slower │
│ QQuery 19 │        27.57 / 31.50 ±4.63 / 37.72 ms │        27.94 / 28.35 ±0.32 / 28.78 ms │ +1.11x faster │
│ QQuery 20 │    522.30 / 531.53 ±10.54 / 551.53 ms │    515.47 / 522.50 ±11.57 / 545.60 ms │     no change │
│ QQuery 21 │    515.17 / 527.20 ±10.11 / 543.68 ms │     522.78 / 526.22 ±2.03 / 528.54 ms │     no change │
│ QQuery 22 │   999.02 / 1015.60 ±8.75 / 1025.03 ms │    995.81 / 998.86 ±2.97 / 1004.32 ms │     no change │
│ QQuery 23 │ 3060.50 / 3122.21 ±34.88 / 3161.13 ms │ 3103.58 / 3118.73 ±10.15 / 3135.05 ms │     no change │
│ QQuery 24 │        41.30 / 41.95 ±0.67 / 43.23 ms │        41.21 / 44.15 ±3.96 / 52.00 ms │  1.05x slower │
│ QQuery 25 │     111.23 / 112.69 ±1.14 / 114.20 ms │    112.25 / 122.73 ±13.02 / 143.46 ms │  1.09x slower │
│ QQuery 26 │        41.34 / 42.57 ±0.92 / 44.14 ms │        41.98 / 43.06 ±1.25 / 45.51 ms │     no change │
│ QQuery 27 │     515.76 / 521.99 ±4.67 / 529.52 ms │    523.49 / 533.42 ±11.69 / 555.74 ms │     no change │
│ QQuery 28 │ 2932.13 / 2951.08 ±12.49 / 2969.23 ms │ 2999.71 / 3026.18 ±18.74 / 3054.99 ms │     no change │
│ QQuery 29 │        41.25 / 46.70 ±7.89 / 61.61 ms │        41.38 / 46.32 ±9.05 / 64.41 ms │     no change │
│ QQuery 30 │     305.39 / 312.89 ±7.46 / 327.20 ms │    348.72 / 361.78 ±10.53 / 376.11 ms │  1.16x slower │
│ QQuery 31 │    281.09 / 291.17 ±10.06 / 309.35 ms │    376.43 / 386.29 ±13.30 / 411.80 ms │  1.33x slower │
│ QQuery 32 │    941.89 / 967.87 ±13.98 / 982.92 ms │ 1631.16 / 1685.72 ±35.35 / 1729.73 ms │  1.74x slower │
│ QQuery 33 │  1482.44 / 1491.65 ±7.10 / 1498.58 ms │ 2102.63 / 2153.42 ±34.60 / 2203.10 ms │  1.44x slower │
│ QQuery 34 │ 1507.73 / 1561.02 ±54.07 / 1654.61 ms │ 1984.67 / 2148.28 ±82.55 / 2209.90 ms │  1.38x slower │
│ QQuery 35 │     288.65 / 305.60 ±9.85 / 316.66 ms │     323.69 / 334.96 ±8.67 / 345.66 ms │  1.10x slower │
│ QQuery 36 │        67.21 / 74.29 ±7.63 / 83.97 ms │     79.15 / 100.42 ±33.10 / 166.20 ms │  1.35x slower │
│ QQuery 37 │        35.86 / 39.16 ±2.65 / 42.48 ms │        38.32 / 39.35 ±0.89 / 40.69 ms │     no change │
│ QQuery 38 │        39.96 / 43.42 ±2.84 / 48.54 ms │        41.96 / 43.15 ±1.01 / 44.56 ms │     no change │
│ QQuery 39 │     140.21 / 152.74 ±7.73 / 163.06 ms │     152.30 / 157.71 ±5.24 / 166.90 ms │     no change │
│ QQuery 40 │        14.36 / 16.81 ±3.85 / 24.37 ms │        14.67 / 15.07 ±0.29 / 15.50 ms │ +1.12x faster │
│ QQuery 41 │        13.85 / 14.22 ±0.19 / 14.41 ms │        13.79 / 14.58 ±0.47 / 15.25 ms │     no change │
│ QQuery 42 │        13.70 / 14.59 ±1.37 / 17.30 ms │        14.09 / 14.32 ±0.17 / 14.62 ms │     no change │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                           ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                           │ 19662.75ms │
│ Total Time (add-blocks-impl-from-scratch)   │ 23204.49ms │
│ Average Time (HEAD)                         │   457.27ms │
│ Average Time (add-blocks-impl-from-scratch) │   539.64ms │
│ Queries Faster                              │          2 │
│ Queries Slower                              │         21 │
│ Queries with No Change                      │         20 │
│ Queries with Failure                        │          0 │
└─────────────────────────────────────────────┴────────────┘

Resource Usage

clickbench_partitioned — base (merge-base)

Metric	Value
Wall time	100.0s
Peak memory	11.1 GiB
Avg memory	4.7 GiB
CPU user	1004.2s
CPU sys	72.5s
Peak spill	0 B

clickbench_partitioned — branch

Metric	Value
Wall time	120.0s
Peak memory	10.8 GiB
Avg memory	3.8 GiB
CPU user	1026.3s
CPU sys	149.4s
Peak spill	0 B

File an issue against this benchmark runner

rluvaton added 30 commits August 31, 2026 10:54

copy paste aggregates as is to aggregates_blocked

5cdf7d5

update imports and make it compile

cc0e79a

add steps

5979c90

use existing structs from aggregate to avoid full duplication

16e3848

rename AggregateExec in blocked to BlockedAggregateExec

1a2e531

update AggregateExec::try_new to return the BlockedAggregateExec …

f0f7b8d

…so all usages will be migrated

add fallback to BlockedAggregateExec so unsupported logic will be for…

672ebb5

…warded

Rename GroupValues to BlockedGroupValues and GroupColumn to `Bl…

41b926c

…ockedGroupColumn`

Add BlockedIndex and BlockedEmitTo

358d618

Change in group indices to be stored internally as BlocksIndex and …

464f8fe

…not `usize`

Remove all BlockedGroupValues impl and replace with adapter

e5d3895

Add batch_size to BlockedGroupValues for later emit change

1e9c6d1

fallback to non blocked in grouped_hash_stream and removed duplicat…

2ae5d2c

…e grouped_hash_stream code

add blocked emit to

6970f8c

changed intern to get BlocksIndex and not usize

a605809

create blocked groups accumulator

e6cdab2

remove print

d6a5479

fix compile for tests

50796d5

add group indices to ordering

f76ba2a

Merge branch 'main' into add-blocks-impl-from-scratch

22e0f32

# Conflicts: # datafusion/expr-common/src/groups_accumulator.rs

fix emit too much

9082071

add blocked group by boolean

0c67d68

add state_preserving to blocked group accumulator adapter

923ee62

add evaluate_preserving to blocked group accumulator adapter

0aeb359

fix compile for BlockedGroupsSelection

7fb5b7a

impl blocked group accumulator for sum

f511677

fix group by boolean

11216d2

add support for blocked min max

ee5f265

add more blocked helpers and tests

98e7c1a

re-export and fix clippy

e34750f

rluvaton added 2 commits September 6, 2026 20:14

let claude increase mmap address space so we don't need to reallocate…

0074517

… often

change to flat indexes

10864ee

rluvaton added 3 commits September 7, 2026 11:06

add idea about into iter on group values

e20af72

implement dedicated take all rather than using take block in a loop f…

9396a63

…or blocked binary view map

build null once

4d3c197

fix not all blocks for empty blocks

4b720d7

let claude improve perf a little

38c8d88

Uh oh!

Conversation

rluvaton commented Sep 3, 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?

What is the testing strategy for this PR?

Are there any user-facing changes?

Uh oh!

rluvaton commented Sep 6, 2026

Uh oh!

rluvaton commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

adriangbot commented Sep 6, 2026

Uh oh!

rluvaton commented Sep 7, 2026

Uh oh!

adriangbot commented Sep 7, 2026

Uh oh!

rluvaton commented Sep 7, 2026

Uh oh!

adriangbot commented Sep 7, 2026

Uh oh!

adriangbot commented Sep 7, 2026

Uh oh!

rluvaton commented Sep 7, 2026

Uh oh!

adriangbot commented Sep 7, 2026

Uh oh!

adriangbot commented Sep 7, 2026

Uh oh!

adriangbot commented Sep 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

rluvaton commented Sep 3, 2026 •

edited

Loading