feat: Improved multiple column aggregation performance by using bitmasks rather than Vec<bool>

[huymq1710]

Which issue does this PR close?

• Closes Potential Improved multiple column aggregation performance by using bitmasks rather than Vec<bool> #18676

Rationale for this change

Using bitmasks rather than Vec<bool>

What changes are included in this PR?

• Uses arrow's BooleanBufferBuilder with as_slice_mut() + apply_bitwise_binary_op for the primitive non-nullable path

Are these changes tested?

Are there any user-facing changes?

No

[huymq1710]

Wrong approach
Read https://github.com/apache/arrow-rs/pull/8619/changes#diff-24dc7184f64fe7a484db137ef61b2ec31090bb959fc9013beefea7865cdfea59

[rluvaton]

@huymq1710 thank you for this PR, did you see any improvement in performance?

my original thought was that I can do bulk matching and build mask for every 64 items but because the indexes may not contiguous it will require scatter/gather which is expensive so I did not continue through

[Dandandan]

run benchmarks

[rluvaton]

run benchmark aggregate_query_sql

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4334262172-1856-pbvb8 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 column-aggregation-bitmasks (f3f0def) to bbf67d9 (merge-base) diff using: clickbench_partitioned
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4334262172-1858-7vwfx 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 column-aggregation-bitmasks (f3f0def) to bbf67d9 (merge-base) diff using: tpch
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Criterion benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4334266698-1859-lc64l 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 column-aggregation-bitmasks (f3f0def) to bbf67d9 (merge-base) diff
BENCH_NAME=aggregate_query_sql
BENCH_COMMAND=cargo bench --features=parquet --bench aggregate_query_sql
BENCH_FILTER=
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4334262172-1857-gkv9f 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 column-aggregation-bitmasks (f3f0def) to bbf67d9 (merge-base) diff using: tpcds
Results will be posted here when complete

---

File an issue against this benchmark runner

[Dandandan]

Even if it doesn't improve perf measurable, I think we can take the changes.

I think aggregation is very sensitive to memory / cache usage, so the more we can do to reduce the state the better and at some point it will pay off as aggregation state moves some level up (RAM - > L3, L3 -> L2, etc.)

I think there is more low hanging fruit in this area

• usize -> u32 group ids if possible

• memorizing u64 hash not always needed

• reducing double storage for u128 views

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Details

Comparing HEAD and column-aggregation-bitmasks
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃    column-aggregation-bitmasks ┃    Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 1  │ 40.43 / 41.45 ±1.57 / 44.57 ms │ 40.11 / 41.04 ±1.42 / 43.88 ms │ no change │
│ QQuery 2  │ 21.01 / 21.26 ±0.19 / 21.59 ms │ 20.58 / 21.01 ±0.22 / 21.18 ms │ no change │
│ QQuery 3  │ 38.24 / 40.25 ±1.51 / 42.53 ms │ 39.27 / 41.05 ±1.37 / 43.31 ms │ no change │
│ QQuery 4  │ 18.24 / 18.44 ±0.17 / 18.75 ms │ 18.06 / 18.31 ±0.16 / 18.49 ms │ no change │
│ QQuery 5  │ 47.12 / 49.46 ±1.60 / 51.21 ms │ 48.63 / 50.41 ±0.90 / 50.96 ms │ no change │
│ QQuery 6  │ 17.22 / 17.49 ±0.41 / 18.31 ms │ 17.21 / 17.28 ±0.09 / 17.45 ms │ no change │
│ QQuery 7  │ 54.78 / 56.26 ±1.86 / 59.75 ms │ 53.60 / 54.85 ±0.98 / 56.33 ms │ no change │
│ QQuery 8  │ 48.41 / 48.66 ±0.18 / 48.94 ms │ 48.08 / 48.61 ±0.52 / 49.39 ms │ no change │
│ QQuery 9  │ 53.98 / 54.15 ±0.16 / 54.45 ms │ 53.39 / 54.11 ±0.71 / 55.07 ms │ no change │
│ QQuery 10 │ 65.47 / 66.16 ±0.49 / 66.90 ms │ 65.14 / 66.27 ±1.03 / 67.61 ms │ no change │
│ QQuery 11 │ 13.79 / 14.06 ±0.34 / 14.72 ms │ 13.91 / 14.14 ±0.21 / 14.54 ms │ no change │
│ QQuery 12 │ 27.54 / 28.44 ±1.40 / 31.20 ms │ 27.34 / 27.70 ±0.35 / 28.20 ms │ no change │
│ QQuery 13 │ 37.92 / 38.25 ±0.32 / 38.75 ms │ 36.75 / 38.16 ±0.78 / 38.92 ms │ no change │
│ QQuery 14 │ 27.80 / 27.89 ±0.08 / 28.02 ms │ 27.68 / 28.27 ±0.70 / 29.61 ms │ no change │
│ QQuery 15 │ 33.18 / 33.33 ±0.20 / 33.73 ms │ 33.26 / 33.40 ±0.10 / 33.51 ms │ no change │
│ QQuery 16 │ 15.32 / 15.41 ±0.08 / 15.56 ms │ 15.10 / 15.22 ±0.08 / 15.36 ms │ no change │
│ QQuery 17 │ 78.09 / 79.83 ±1.50 / 82.57 ms │ 78.28 / 78.62 ±0.28 / 78.97 ms │ no change │
│ QQuery 18 │ 75.21 / 76.33 ±0.77 / 77.44 ms │ 74.59 / 75.47 ±0.76 / 76.73 ms │ no change │
│ QQuery 19 │ 37.73 / 38.08 ±0.30 / 38.44 ms │ 37.29 / 39.00 ±2.09 / 42.84 ms │ no change │
│ QQuery 20 │ 39.31 / 40.81 ±2.08 / 44.92 ms │ 38.97 / 39.89 ±0.99 / 41.75 ms │ no change │
│ QQuery 21 │ 62.13 / 64.21 ±1.55 / 66.35 ms │ 62.51 / 64.64 ±1.36 / 66.30 ms │ no change │
│ QQuery 22 │ 23.67 / 23.85 ±0.18 / 24.12 ms │ 23.51 / 23.87 ±0.28 / 24.30 ms │ no change │
└───────────┴────────────────────────────────┴────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                          ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                          │ 894.10ms │
│ Total Time (column-aggregation-bitmasks)   │ 891.33ms │
│ Average Time (HEAD)                        │  40.64ms │
│ Average Time (column-aggregation-bitmasks) │  40.51ms │
│ Queries Faster                             │        0 │
│ Queries Slower                             │        0 │
│ Queries with No Change                     │       22 │
│ Queries with Failure                       │        0 │
└────────────────────────────────────────────┴──────────┘

Resource Usage

tpch — base (merge-base)

Metric	Value
Wall time	5.0s
Peak memory	5.5 GiB
Avg memory	4.9 GiB
CPU user	33.7s
CPU sys	2.3s
Peak spill	0 B

tpch — branch

Metric	Value
Wall time	5.0s
Peak memory	5.3 GiB
Avg memory	4.8 GiB
CPU user	33.5s
CPU sys	2.4s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Details

Comparing HEAD and column-aggregation-bitmasks
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                     HEAD ┃              column-aggregation-bitmasks ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │              7.20 / 7.66 ±0.78 / 9.22 ms │              6.93 / 7.48 ±0.86 / 9.19 ms │     no change │
│ QQuery 2  │        146.37 / 146.57 ±0.15 / 146.74 ms │        143.51 / 143.95 ±0.41 / 144.67 ms │     no change │
│ QQuery 3  │        114.62 / 115.75 ±1.09 / 117.20 ms │        112.45 / 113.55 ±0.79 / 114.49 ms │     no change │
│ QQuery 4  │     1320.69 / 1327.14 ±4.01 / 1331.43 ms │     1316.50 / 1323.49 ±4.14 / 1328.37 ms │     no change │
│ QQuery 5  │        174.85 / 176.82 ±1.79 / 179.86 ms │        173.38 / 174.68 ±1.24 / 176.80 ms │     no change │
│ QQuery 6  │        144.21 / 147.78 ±2.29 / 150.17 ms │        140.53 / 142.91 ±1.54 / 144.93 ms │     no change │
│ QQuery 7  │        336.37 / 338.83 ±1.64 / 341.09 ms │        331.38 / 334.42 ±2.29 / 337.21 ms │     no change │
│ QQuery 8  │        114.18 / 114.89 ±0.54 / 115.49 ms │        113.17 / 113.67 ±0.32 / 114.06 ms │     no change │
│ QQuery 9  │          97.05 / 98.59 ±1.78 / 101.94 ms │          96.06 / 99.76 ±2.90 / 103.21 ms │     no change │
│ QQuery 10 │        103.86 / 104.53 ±0.45 / 105.22 ms │        101.84 / 102.83 ±0.60 / 103.47 ms │     no change │
│ QQuery 11 │        921.64 / 931.52 ±5.59 / 938.98 ms │        902.90 / 915.32 ±7.06 / 923.08 ms │     no change │
│ QQuery 12 │           43.55 / 43.88 ±0.17 / 44.03 ms │           43.35 / 43.58 ±0.21 / 43.93 ms │     no change │
│ QQuery 13 │        392.54 / 394.38 ±2.19 / 397.87 ms │        388.58 / 390.67 ±2.26 / 394.93 ms │     no change │
│ QQuery 14 │     1039.23 / 1041.18 ±1.35 / 1042.32 ms │     1029.93 / 1036.83 ±4.67 / 1042.29 ms │     no change │
│ QQuery 15 │           14.89 / 15.21 ±0.31 / 15.68 ms │           14.97 / 15.21 ±0.23 / 15.62 ms │     no change │
│ QQuery 16 │             7.50 / 9.07 ±2.72 / 14.51 ms │              7.42 / 7.56 ±0.14 / 7.83 ms │ +1.20x faster │
│ QQuery 17 │        222.85 / 224.80 ±1.62 / 227.01 ms │        221.21 / 222.43 ±1.11 / 224.47 ms │     no change │
│ QQuery 18 │        126.58 / 127.44 ±0.52 / 127.91 ms │        125.21 / 126.79 ±1.06 / 128.10 ms │     no change │
│ QQuery 19 │        155.00 / 156.57 ±1.73 / 159.81 ms │        153.27 / 155.59 ±1.45 / 157.60 ms │     no change │
│ QQuery 20 │           13.49 / 13.59 ±0.08 / 13.69 ms │           13.11 / 13.60 ±0.33 / 14.00 ms │     no change │
│ QQuery 21 │           19.76 / 19.97 ±0.16 / 20.17 ms │           19.24 / 19.46 ±0.19 / 19.72 ms │     no change │
│ QQuery 22 │        478.81 / 482.65 ±3.55 / 486.95 ms │        484.79 / 488.58 ±2.58 / 492.01 ms │     no change │
│ QQuery 23 │     1107.60 / 1113.12 ±6.17 / 1123.75 ms │     1096.11 / 1100.69 ±7.39 / 1115.41 ms │     no change │
│ QQuery 24 │        716.27 / 719.12 ±2.96 / 723.64 ms │        705.75 / 710.20 ±3.95 / 716.42 ms │     no change │
│ QQuery 25 │        335.19 / 336.54 ±1.60 / 339.56 ms │        331.24 / 333.46 ±1.84 / 335.28 ms │     no change │
│ QQuery 26 │           78.57 / 79.14 ±0.60 / 80.28 ms │           77.14 / 77.65 ±0.37 / 78.25 ms │     no change │
│ QQuery 27 │              7.23 / 7.33 ±0.11 / 7.55 ms │              6.98 / 7.08 ±0.15 / 7.38 ms │     no change │
│ QQuery 28 │        149.33 / 150.95 ±2.52 / 155.96 ms │        147.73 / 149.12 ±0.85 / 150.40 ms │     no change │
│ QQuery 29 │        272.55 / 276.25 ±3.39 / 282.54 ms │        270.40 / 273.53 ±3.47 / 280.28 ms │     no change │
│ QQuery 30 │           42.24 / 42.60 ±0.19 / 42.82 ms │           42.56 / 44.91 ±3.30 / 51.43 ms │  1.05x slower │
│ QQuery 31 │        166.78 / 169.50 ±2.05 / 172.98 ms │        166.00 / 168.48 ±1.98 / 171.78 ms │     no change │
│ QQuery 32 │           13.28 / 13.77 ±0.27 / 14.01 ms │           13.51 / 13.66 ±0.12 / 13.85 ms │     no change │
│ QQuery 33 │        139.64 / 141.93 ±2.61 / 146.95 ms │        138.12 / 140.29 ±1.20 / 141.76 ms │     no change │
│ QQuery 34 │              7.13 / 7.36 ±0.21 / 7.73 ms │              7.20 / 7.33 ±0.18 / 7.68 ms │     no change │
│ QQuery 35 │        103.01 / 103.82 ±0.75 / 105.17 ms │        101.58 / 103.14 ±2.19 / 107.41 ms │     no change │
│ QQuery 36 │              7.23 / 7.29 ±0.04 / 7.34 ms │              6.89 / 6.95 ±0.05 / 7.02 ms │     no change │
│ QQuery 37 │              8.48 / 8.62 ±0.10 / 8.76 ms │              8.25 / 8.37 ±0.08 / 8.50 ms │     no change │
│ QQuery 38 │           87.54 / 89.81 ±2.10 / 92.66 ms │           85.97 / 87.42 ±1.28 / 89.47 ms │     no change │
│ QQuery 39 │        118.29 / 118.86 ±0.56 / 119.88 ms │        117.53 / 121.40 ±5.59 / 132.42 ms │     no change │
│ QQuery 40 │        102.89 / 105.20 ±1.26 / 106.58 ms │        101.12 / 107.71 ±5.15 / 113.78 ms │     no change │
│ QQuery 41 │           14.88 / 15.07 ±0.14 / 15.30 ms │           14.58 / 14.82 ±0.27 / 15.29 ms │     no change │
│ QQuery 42 │        108.60 / 110.69 ±2.24 / 114.65 ms │        106.52 / 107.38 ±0.83 / 108.88 ms │     no change │
│ QQuery 43 │              5.88 / 6.01 ±0.15 / 6.31 ms │              5.64 / 5.94 ±0.28 / 6.34 ms │     no change │
│ QQuery 44 │           11.16 / 11.36 ±0.16 / 11.59 ms │           10.92 / 11.81 ±1.54 / 14.88 ms │     no change │
│ QQuery 45 │           49.12 / 49.51 ±0.29 / 49.97 ms │           48.29 / 48.91 ±0.41 / 49.38 ms │     no change │
│ QQuery 46 │              8.62 / 8.81 ±0.20 / 9.14 ms │              8.48 / 8.64 ±0.14 / 8.90 ms │     no change │
│ QQuery 47 │        710.42 / 719.09 ±7.18 / 731.76 ms │        703.16 / 710.77 ±5.19 / 717.41 ms │     no change │
│ QQuery 48 │        276.50 / 280.68 ±5.21 / 290.92 ms │        273.13 / 277.29 ±4.70 / 286.14 ms │     no change │
│ QQuery 49 │        249.85 / 251.22 ±1.75 / 254.57 ms │        249.00 / 249.90 ±0.71 / 251.15 ms │     no change │
│ QQuery 50 │        204.06 / 206.82 ±3.01 / 212.57 ms │        202.26 / 206.75 ±3.07 / 210.53 ms │     no change │
│ QQuery 51 │        180.12 / 182.50 ±2.63 / 187.23 ms │        172.60 / 176.36 ±2.12 / 179.17 ms │     no change │
│ QQuery 52 │        108.27 / 110.35 ±3.03 / 116.37 ms │        106.85 / 107.33 ±0.33 / 107.63 ms │     no change │
│ QQuery 53 │        103.13 / 103.90 ±0.51 / 104.65 ms │        101.64 / 103.19 ±2.22 / 107.60 ms │     no change │
│ QQuery 54 │        147.38 / 148.71 ±2.03 / 152.68 ms │        144.59 / 145.78 ±0.79 / 147.06 ms │     no change │
│ QQuery 55 │        107.92 / 109.29 ±1.99 / 113.17 ms │        104.99 / 107.31 ±1.57 / 109.91 ms │     no change │
│ QQuery 56 │        141.40 / 141.85 ±0.45 / 142.46 ms │        138.67 / 139.83 ±1.26 / 141.92 ms │     no change │
│ QQuery 57 │        167.57 / 168.51 ±1.26 / 170.94 ms │        163.19 / 165.25 ±1.41 / 166.96 ms │     no change │
│ QQuery 58 │        265.29 / 267.06 ±1.74 / 270.34 ms │        263.39 / 265.44 ±1.45 / 267.70 ms │     no change │
│ QQuery 59 │        195.01 / 196.68 ±1.65 / 199.81 ms │        193.39 / 195.87 ±2.08 / 199.63 ms │     no change │
│ QQuery 60 │        142.37 / 143.86 ±1.22 / 145.53 ms │        140.17 / 141.51 ±0.83 / 142.39 ms │     no change │
│ QQuery 61 │           13.75 / 13.98 ±0.17 / 14.26 ms │           13.77 / 13.89 ±0.15 / 14.19 ms │     no change │
│ QQuery 62 │        870.17 / 878.77 ±5.29 / 885.69 ms │        870.15 / 880.87 ±9.79 / 897.90 ms │     no change │
│ QQuery 63 │        104.08 / 105.60 ±1.62 / 108.36 ms │        102.31 / 103.02 ±0.81 / 104.52 ms │     no change │
│ QQuery 64 │        663.52 / 666.21 ±2.19 / 669.82 ms │        659.60 / 663.61 ±3.75 / 670.53 ms │     no change │
│ QQuery 65 │        250.29 / 254.18 ±4.60 / 263.10 ms │        242.81 / 245.43 ±1.60 / 247.52 ms │     no change │
│ QQuery 66 │       215.67 / 225.59 ±10.75 / 244.89 ms │        213.72 / 219.80 ±9.64 / 238.82 ms │     no change │
│ QQuery 67 │        298.21 / 303.47 ±3.52 / 308.22 ms │       297.78 / 309.06 ±11.81 / 331.26 ms │     no change │
│ QQuery 68 │              8.73 / 8.99 ±0.22 / 9.39 ms │              8.74 / 8.91 ±0.19 / 9.25 ms │     no change │
│ QQuery 69 │         99.50 / 100.80 ±1.45 / 103.59 ms │           97.86 / 98.44 ±0.48 / 99.26 ms │     no change │
│ QQuery 70 │        317.90 / 321.41 ±4.56 / 330.32 ms │        321.29 / 329.50 ±5.51 / 337.51 ms │     no change │
│ QQuery 71 │        134.67 / 136.74 ±3.25 / 143.18 ms │        134.00 / 135.18 ±1.48 / 137.97 ms │     no change │
│ QQuery 72 │       579.56 / 596.43 ±10.10 / 611.28 ms │        579.93 / 589.57 ±6.54 / 600.02 ms │     no change │
│ QQuery 73 │              6.77 / 6.92 ±0.16 / 7.23 ms │              6.77 / 6.90 ±0.19 / 7.27 ms │     no change │
│ QQuery 74 │        587.43 / 592.33 ±3.89 / 599.08 ms │        582.17 / 586.47 ±4.00 / 592.72 ms │     no change │
│ QQuery 75 │        268.90 / 270.31 ±1.33 / 272.68 ms │        267.82 / 270.37 ±3.95 / 278.24 ms │     no change │
│ QQuery 76 │        130.14 / 130.91 ±0.43 / 131.38 ms │        129.79 / 131.57 ±1.67 / 134.70 ms │     no change │
│ QQuery 77 │        188.69 / 190.05 ±0.99 / 191.30 ms │        187.22 / 190.19 ±3.65 / 197.40 ms │     no change │
│ QQuery 78 │        331.76 / 335.06 ±3.82 / 342.43 ms │        329.29 / 331.39 ±1.96 / 334.38 ms │     no change │
│ QQuery 79 │        231.27 / 236.00 ±4.07 / 243.60 ms │        228.47 / 230.52 ±1.50 / 232.71 ms │     no change │
│ QQuery 80 │        320.03 / 322.18 ±1.72 / 324.76 ms │        317.50 / 319.36 ±1.17 / 321.17 ms │     no change │
│ QQuery 81 │           26.49 / 26.86 ±0.31 / 27.37 ms │           26.23 / 26.90 ±1.22 / 29.35 ms │     no change │
│ QQuery 82 │           40.30 / 41.31 ±1.78 / 44.85 ms │           39.23 / 39.48 ±0.23 / 39.81 ms │     no change │
│ QQuery 83 │           37.53 / 37.89 ±0.37 / 38.54 ms │           37.41 / 37.66 ±0.42 / 38.50 ms │     no change │
│ QQuery 84 │           46.52 / 46.85 ±0.40 / 47.62 ms │           46.15 / 46.52 ±0.26 / 46.83 ms │     no change │
│ QQuery 85 │        141.93 / 143.31 ±1.29 / 145.70 ms │        140.66 / 142.48 ±1.80 / 145.17 ms │     no change │
│ QQuery 86 │           37.88 / 38.29 ±0.28 / 38.70 ms │           37.81 / 38.08 ±0.25 / 38.49 ms │     no change │
│ QQuery 87 │              3.63 / 3.74 ±0.16 / 4.06 ms │              3.55 / 3.67 ±0.19 / 4.04 ms │     no change │
│ QQuery 88 │        101.26 / 101.75 ±0.38 / 102.35 ms │         99.20 / 105.21 ±9.81 / 124.69 ms │     no change │
│ QQuery 89 │        119.27 / 121.32 ±3.09 / 127.48 ms │        115.37 / 116.52 ±0.87 / 117.62 ms │     no change │
│ QQuery 90 │           22.95 / 23.50 ±0.34 / 23.89 ms │           22.47 / 22.68 ±0.18 / 22.96 ms │     no change │
│ QQuery 91 │           58.86 / 59.70 ±0.60 / 60.51 ms │           58.05 / 59.22 ±1.27 / 61.60 ms │     no change │
│ QQuery 92 │           57.54 / 58.73 ±0.97 / 60.45 ms │           56.39 / 57.17 ±0.54 / 57.71 ms │     no change │
│ QQuery 93 │        180.91 / 183.57 ±2.75 / 188.63 ms │        179.15 / 180.56 ±0.88 / 181.62 ms │     no change │
│ QQuery 94 │           60.89 / 61.70 ±0.52 / 62.37 ms │           60.65 / 61.05 ±0.24 / 61.37 ms │     no change │
│ QQuery 95 │        126.33 / 127.54 ±1.27 / 129.96 ms │        125.66 / 126.10 ±0.58 / 127.23 ms │     no change │
│ QQuery 96 │           69.75 / 70.77 ±0.76 / 72.05 ms │           67.87 / 68.86 ±0.63 / 69.56 ms │     no change │
│ QQuery 97 │        117.76 / 119.23 ±2.35 / 123.90 ms │        114.79 / 116.89 ±1.54 / 119.24 ms │     no change │
│ QQuery 98 │        149.48 / 151.08 ±1.78 / 154.33 ms │        149.99 / 151.55 ±1.32 / 153.58 ms │     no change │
│ QQuery 99 │ 10744.52 / 10801.72 ±45.96 / 10863.54 ms │ 10748.45 / 10802.67 ±48.73 / 10878.07 ms │     no change │
└───────────┴──────────────────────────────────────────┴──────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                          │ 30676.77ms │
│ Total Time (column-aggregation-bitmasks)   │ 30511.12ms │
│ Average Time (HEAD)                        │   309.87ms │
│ Average Time (column-aggregation-bitmasks) │   308.19ms │
│ Queries Faster                             │          1 │
│ Queries Slower                             │          1 │
│ Queries with No Change                     │         97 │
│ Queries with Failure                       │          0 │
└────────────────────────────────────────────┴────────────┘

Resource Usage

tpcds — base (merge-base)

Metric	Value
Wall time	155.0s
Peak memory	6.2 GiB
Avg memory	5.6 GiB
CPU user	243.2s
CPU sys	8.0s
Peak spill	0 B

tpcds — branch

Metric	Value
Wall time	155.0s
Peak memory	6.3 GiB
Avg memory	5.6 GiB
CPU user	242.0s
CPU sys	7.4s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Details

Comparing HEAD and column-aggregation-bitmasks
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃           column-aggregation-bitmasks ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │          1.19 / 4.59 ±6.73 / 18.05 ms │          1.21 / 4.59 ±6.70 / 17.99 ms │     no change │
│ QQuery 1  │        12.30 / 12.70 ±0.22 / 12.91 ms │        13.05 / 13.29 ±0.16 / 13.47 ms │     no change │
│ QQuery 2  │        36.75 / 37.15 ±0.30 / 37.64 ms │        36.28 / 36.72 ±0.56 / 37.81 ms │     no change │
│ QQuery 3  │        31.46 / 32.51 ±0.75 / 33.30 ms │        31.64 / 32.05 ±0.30 / 32.55 ms │     no change │
│ QQuery 4  │     237.79 / 243.40 ±3.66 / 247.57 ms │     242.72 / 246.50 ±3.10 / 250.16 ms │     no change │
│ QQuery 5  │     284.31 / 286.21 ±1.15 / 287.62 ms │     285.00 / 287.43 ±1.65 / 289.69 ms │     no change │
│ QQuery 6  │           6.92 / 7.43 ±0.37 / 7.75 ms │           6.19 / 7.27 ±0.81 / 8.32 ms │     no change │
│ QQuery 7  │        13.84 / 13.98 ±0.11 / 14.13 ms │        14.37 / 15.05 ±1.31 / 17.67 ms │  1.08x slower │
│ QQuery 8  │     329.41 / 330.77 ±1.53 / 332.76 ms │     324.22 / 327.95 ±2.72 / 331.77 ms │     no change │
│ QQuery 9  │     442.90 / 455.85 ±9.84 / 471.84 ms │     453.71 / 463.42 ±6.15 / 470.31 ms │     no change │
│ QQuery 10 │        74.42 / 75.22 ±0.87 / 76.89 ms │        76.10 / 80.76 ±7.51 / 95.71 ms │  1.07x slower │
│ QQuery 11 │        85.67 / 87.17 ±2.53 / 92.22 ms │        87.52 / 88.43 ±1.35 / 91.03 ms │     no change │
│ QQuery 12 │     277.49 / 280.52 ±2.28 / 284.35 ms │     281.52 / 286.11 ±4.57 / 294.49 ms │     no change │
│ QQuery 13 │    393.82 / 413.44 ±10.83 / 421.98 ms │     400.00 / 410.89 ±6.28 / 417.70 ms │     no change │
│ QQuery 14 │     285.96 / 290.44 ±2.68 / 294.31 ms │     288.79 / 294.11 ±4.32 / 297.96 ms │     no change │
│ QQuery 15 │     283.12 / 290.19 ±6.40 / 297.92 ms │     288.61 / 292.41 ±2.81 / 295.93 ms │     no change │
│ QQuery 16 │     628.14 / 632.06 ±2.04 / 633.76 ms │     615.96 / 629.17 ±9.80 / 641.80 ms │     no change │
│ QQuery 17 │     634.57 / 642.68 ±6.87 / 654.61 ms │     617.00 / 624.53 ±8.46 / 640.52 ms │     no change │
│ QQuery 18 │ 1342.81 / 1369.57 ±14.97 / 1388.26 ms │ 1231.32 / 1267.16 ±28.27 / 1311.10 ms │ +1.08x faster │
│ QQuery 19 │        29.51 / 29.98 ±0.30 / 30.36 ms │        28.94 / 29.48 ±0.36 / 30.00 ms │     no change │
│ QQuery 20 │     525.98 / 533.75 ±7.52 / 545.33 ms │     522.29 / 530.99 ±6.32 / 538.20 ms │     no change │
│ QQuery 21 │     602.64 / 609.44 ±5.52 / 617.42 ms │     598.51 / 604.61 ±3.34 / 607.93 ms │     no change │
│ QQuery 22 │ 1070.53 / 1084.31 ±17.75 / 1117.83 ms │  1064.60 / 1077.70 ±9.19 / 1087.30 ms │     no change │
│ QQuery 23 │ 3379.11 / 3444.48 ±45.05 / 3515.69 ms │ 3330.10 / 3371.51 ±24.51 / 3406.08 ms │     no change │
│ QQuery 24 │        42.89 / 43.79 ±1.40 / 46.55 ms │        42.46 / 46.22 ±4.00 / 53.42 ms │  1.06x slower │
│ QQuery 25 │     114.02 / 117.77 ±3.99 / 125.35 ms │     113.85 / 116.85 ±1.99 / 119.97 ms │     no change │
│ QQuery 26 │        44.22 / 44.73 ±0.34 / 45.29 ms │        42.65 / 43.47 ±0.81 / 44.49 ms │     no change │
│ QQuery 27 │     669.28 / 675.75 ±5.47 / 682.55 ms │     671.93 / 676.24 ±2.19 / 677.79 ms │     no change │
│ QQuery 28 │ 3032.63 / 3052.11 ±11.35 / 3063.13 ms │ 3022.34 / 3042.67 ±14.62 / 3068.03 ms │     no change │
│ QQuery 29 │        43.24 / 52.89 ±7.75 / 60.92 ms │        42.46 / 46.48 ±7.05 / 60.56 ms │ +1.14x faster │
│ QQuery 30 │     312.79 / 319.04 ±4.97 / 326.49 ms │     309.69 / 312.87 ±3.27 / 317.87 ms │     no change │
│ QQuery 31 │     309.30 / 314.56 ±6.18 / 325.91 ms │     302.14 / 307.54 ±5.29 / 316.62 ms │     no change │
│ QQuery 32 │  1016.17 / 1031.05 ±7.59 / 1036.26 ms │    969.80 / 980.80 ±11.36 / 996.19 ms │     no change │
│ QQuery 33 │ 1431.67 / 1463.76 ±16.75 / 1477.28 ms │ 1415.81 / 1445.44 ±17.87 / 1472.12 ms │     no change │
│ QQuery 34 │ 1488.01 / 1517.30 ±16.46 / 1538.13 ms │  1441.95 / 1451.63 ±9.44 / 1463.53 ms │     no change │
│ QQuery 35 │    289.61 / 321.31 ±25.22 / 360.79 ms │    290.81 / 305.45 ±18.42 / 341.36 ms │     no change │
│ QQuery 36 │        67.81 / 71.94 ±2.44 / 75.30 ms │        63.57 / 67.20 ±2.38 / 69.89 ms │ +1.07x faster │
│ QQuery 37 │        36.49 / 39.58 ±2.68 / 43.51 ms │        36.19 / 38.39 ±4.01 / 46.40 ms │     no change │
│ QQuery 38 │        42.54 / 45.40 ±2.37 / 49.75 ms │        42.50 / 43.90 ±0.88 / 44.82 ms │     no change │
│ QQuery 39 │     130.35 / 140.55 ±7.07 / 151.92 ms │     124.90 / 134.48 ±7.45 / 147.77 ms │     no change │
│ QQuery 40 │        15.11 / 16.85 ±2.68 / 22.18 ms │        14.62 / 14.90 ±0.28 / 15.40 ms │ +1.13x faster │
│ QQuery 41 │        14.63 / 15.27 ±1.01 / 17.28 ms │        13.90 / 14.18 ±0.21 / 14.50 ms │ +1.08x faster │
│ QQuery 42 │        13.84 / 14.06 ±0.21 / 14.45 ms │        13.53 / 14.41 ±1.26 / 16.91 ms │     no change │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                          │ 20505.57ms │
│ Total Time (column-aggregation-bitmasks)   │ 20125.23ms │
│ Average Time (HEAD)                        │   476.87ms │
│ Average Time (column-aggregation-bitmasks) │   468.03ms │
│ Queries Faster                             │          5 │
│ Queries Slower                             │          3 │
│ Queries with No Change                     │         35 │
│ Queries with Failure                       │          0 │
└────────────────────────────────────────────┴────────────┘

Resource Usage

clickbench_partitioned — base (merge-base)

Metric	Value
Wall time	105.0s
Peak memory	31.0 GiB
Avg memory	23.4 GiB
CPU user	1087.9s
CPU sys	64.2s
Peak spill	0 B

clickbench_partitioned — branch

Metric	Value
Wall time	105.0s
Peak memory	30.6 GiB
Avg memory	23.1 GiB
CPU user	1063.4s
CPU sys	64.4s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Criterion benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Details

group                                                                         column-aggregation-bitmasks            main
-----                                                                         ---------------------------            ----
aggregate_query_approx_percentile_cont_on_f32                                 1.00  1933.7±34.27µs        ? ?/sec    1.01  1962.6±31.69µs        ? ?/sec
aggregate_query_approx_percentile_cont_on_u64                                 1.00      2.3±0.03ms        ? ?/sec    1.02      2.3±0.05ms        ? ?/sec
aggregate_query_distinct_median                                               1.00      2.5±0.03ms        ? ?/sec    1.01      2.5±0.03ms        ? ?/sec
aggregate_query_group_by                                                      1.00  1274.3±16.65µs        ? ?/sec    1.01  1285.5±21.35µs        ? ?/sec
aggregate_query_group_by_u64 15 12                                            1.00  1226.0±15.99µs        ? ?/sec    1.01  1241.4±20.60µs        ? ?/sec
aggregate_query_group_by_u64_multiple_keys                                    1.00      2.2±0.03ms        ? ?/sec    1.01      2.2±0.02ms        ? ?/sec
aggregate_query_group_by_wide_u64_and_f32_without_aggregate_expressions       1.00  1437.2±19.10µs        ? ?/sec    1.02  1464.6±26.22µs        ? ?/sec
aggregate_query_group_by_wide_u64_and_string_without_aggregate_expressions    1.00  1562.0±16.56µs        ? ?/sec    1.01  1571.3±22.95µs        ? ?/sec
aggregate_query_group_by_with_filter                                          1.00  1391.1±24.32µs        ? ?/sec    1.01  1407.2±19.20µs        ? ?/sec
aggregate_query_group_by_with_filter_u64 15 12                                1.00  1373.8±14.90µs        ? ?/sec    1.02  1402.1±13.78µs        ? ?/sec
aggregate_query_no_group_by 15 12                                             1.00    590.4±6.81µs        ? ?/sec    1.04    613.5±8.19µs        ? ?/sec
aggregate_query_no_group_by_count_distinct_narrow                             1.00  1065.9±15.93µs        ? ?/sec    1.01  1077.5±13.55µs        ? ?/sec
aggregate_query_no_group_by_count_distinct_wide                               1.00  1429.7±20.86µs        ? ?/sec    1.00  1428.8±11.93µs        ? ?/sec
aggregate_query_no_group_by_min_max_f64                                       1.00    548.8±7.72µs        ? ?/sec    1.03   563.4±14.29µs        ? ?/sec
array_agg_query_group_by_few_groups                                           1.00  1305.7±14.94µs        ? ?/sec    1.00  1309.7±18.09µs        ? ?/sec
array_agg_query_group_by_many_groups                                          1.00  1762.4±24.21µs        ? ?/sec    1.01  1785.8±43.64µs        ? ?/sec
array_agg_query_group_by_mid_groups                                           1.00  1479.5±14.64µs        ? ?/sec    1.02  1511.0±22.56µs        ? ?/sec
array_agg_struct_query_group_by_mid_groups                                    1.00  1843.2±25.01µs        ? ?/sec    1.00  1839.9±24.85µs        ? ?/sec
first_last_ignore_nulls                                                       1.00  1624.5±22.83µs        ? ?/sec    1.01  1635.7±16.54µs        ? ?/sec
first_last_many_columns                                                       1.00  1602.6±20.08µs        ? ?/sec    1.01  1622.8±21.72µs        ? ?/sec
first_last_one_column                                                         1.00  1415.2±22.02µs        ? ?/sec    1.01  1430.8±22.19µs        ? ?/sec
string_agg_query_group_by_few_groups                                          1.00  1170.9±14.25µs        ? ?/sec    1.01  1187.6±16.99µs        ? ?/sec
string_agg_query_group_by_many_groups                                         1.00      2.7±0.05ms        ? ?/sec    1.05      2.8±0.05ms        ? ?/sec
string_agg_query_group_by_mid_groups                                          1.00  1671.4±22.75µs        ? ?/sec    1.01  1686.5±20.01µs        ? ?/sec

Resource Usage

base (merge-base)

Metric	Value
Wall time	285.1s
Peak memory	5.6 GiB
Avg memory	5.5 GiB
CPU user	1010.8s
CPU sys	89.9s
Peak spill	0 B

branch

Metric	Value
Wall time	285.1s
Peak memory	5.6 GiB
Avg memory	5.5 GiB
CPU user	1022.2s
CPU sys	86.9s
Peak spill	0 B

---

File an issue against this benchmark runner

[Dandandan]

run benchmark aggregate_query_sql

[adriangbot]

🤖 Criterion benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4334593628-1863-mtwh7 6.12.55+ #1 SMP Sun Feb 1 08:59:41 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 column-aggregation-bitmasks (f878431) to bbf67d9 (merge-base) diff
BENCH_NAME=aggregate_query_sql
BENCH_COMMAND=cargo bench --features=parquet --bench aggregate_query_sql
BENCH_FILTER=
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Criterion benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Details

group                                                                         column-aggregation-bitmasks            main
-----                                                                         ---------------------------            ----
aggregate_query_approx_percentile_cont_on_f32                                 1.00  1950.5±22.57µs        ? ?/sec    1.00  1948.0±29.72µs        ? ?/sec
aggregate_query_approx_percentile_cont_on_u64                                 1.00      2.3±0.03ms        ? ?/sec    1.00      2.3±0.02ms        ? ?/sec
aggregate_query_distinct_median                                               1.01      2.5±0.02ms        ? ?/sec    1.00      2.4±0.02ms        ? ?/sec
aggregate_query_group_by                                                      1.01  1302.9±13.75µs        ? ?/sec    1.00  1295.6±15.16µs        ? ?/sec
aggregate_query_group_by_u64 15 12                                            1.00  1247.0±14.55µs        ? ?/sec    1.00  1244.5±17.45µs        ? ?/sec
aggregate_query_group_by_u64_multiple_keys                                    1.00      2.2±0.03ms        ? ?/sec    1.00      2.2±0.03ms        ? ?/sec
aggregate_query_group_by_wide_u64_and_f32_without_aggregate_expressions       1.00  1466.3±27.12µs        ? ?/sec    1.00  1464.3±38.52µs        ? ?/sec
aggregate_query_group_by_wide_u64_and_string_without_aggregate_expressions    1.01  1576.4±15.72µs        ? ?/sec    1.00  1568.5±15.97µs        ? ?/sec
aggregate_query_group_by_with_filter                                          1.00  1410.6±19.52µs        ? ?/sec    1.00  1406.9±14.70µs        ? ?/sec
aggregate_query_group_by_with_filter_u64 15 12                                1.02  1413.6±15.23µs        ? ?/sec    1.00  1392.0±15.30µs        ? ?/sec
aggregate_query_no_group_by 15 12                                             1.02    614.5±9.34µs        ? ?/sec    1.00    602.0±6.51µs        ? ?/sec
aggregate_query_no_group_by_count_distinct_narrow                             1.00  1086.5±13.77µs        ? ?/sec    1.00  1088.4±15.64µs        ? ?/sec
aggregate_query_no_group_by_count_distinct_wide                               1.00  1441.4±13.67µs        ? ?/sec    1.00  1444.2±39.08µs        ? ?/sec
aggregate_query_no_group_by_min_max_f64                                       1.03    575.7±5.52µs        ? ?/sec    1.00    559.8±7.13µs        ? ?/sec
array_agg_query_group_by_few_groups                                           1.01  1330.6±13.60µs        ? ?/sec    1.00  1320.0±14.48µs        ? ?/sec
array_agg_query_group_by_many_groups                                          1.01  1788.2±21.24µs        ? ?/sec    1.00  1776.2±28.52µs        ? ?/sec
array_agg_query_group_by_mid_groups                                           1.01  1507.0±20.96µs        ? ?/sec    1.00  1490.4±17.66µs        ? ?/sec
array_agg_struct_query_group_by_mid_groups                                    1.00  1844.4±20.81µs        ? ?/sec    1.00  1853.5±130.83µs        ? ?/sec
first_last_ignore_nulls                                                       1.00  1657.3±15.41µs        ? ?/sec    1.00  1652.4±23.47µs        ? ?/sec
first_last_many_columns                                                       1.01  1643.5±18.06µs        ? ?/sec    1.00  1625.4±17.28µs        ? ?/sec
first_last_one_column                                                         1.01  1442.5±14.79µs        ? ?/sec    1.00  1425.4±13.69µs        ? ?/sec
string_agg_query_group_by_few_groups                                          1.00  1193.3±12.04µs        ? ?/sec    1.00  1191.0±13.71µs        ? ?/sec
string_agg_query_group_by_many_groups                                         1.00      2.8±0.04ms        ? ?/sec    1.00      2.8±0.05ms        ? ?/sec
string_agg_query_group_by_mid_groups                                          1.01  1699.9±21.78µs        ? ?/sec    1.00  1687.8±19.46µs        ? ?/sec

Resource Usage

base (merge-base)

Metric	Value
Wall time	290.1s
Peak memory	5.6 GiB
Avg memory	5.5 GiB
CPU user	1029.7s
CPU sys	88.7s
Peak spill	0 B

branch

Metric	Value
Wall time	290.1s
Peak memory	5.6 GiB
Avg memory	5.5 GiB
CPU user	1033.7s
CPU sys	88.7s
Peak spill	0 B

---

File an issue against this benchmark runner

[huymq1710]

Thanks for the reviews and running the benchmarks

Performance appears neutral (1.00x - 1.03x), so the benefit is primarily the state reduction and memory efficiency mentioned by @Dandandan

My goal is to learn DataFusion's internals, so I'm happy to either merge or close it if you'd prefer to keep the current implementation