Skip to content

Specialize BatchCoalescer for RunArray - #12218

Draft
Rich-T-kid wants to merge 23 commits into
apache:mainfrom
Rich-T-kid:codex/issue-11720
Draft

Rich-T-kid wants to merge 23 commits into
apache:mainfrom
Rich-T-kid:codex/issue-11720

Conversation

@Rich-T-kid

@Rich-T-kid Rich-T-kid commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

Which issue does this PR close?

Rationale for this change

BatchCoalescer currently uses the generic in-progress array implementation for RunArray. A run-aware accumulator can preserve the encoding while rows are copied and batches are combined.

What changes are included in this PR?

Adds InProgressRunArray for run-end encoded columns with Int16, Int32, or Int64 run ends.

The accumulator handles a logical slice in two steps: it appends clamped, rebased run ends, then retains the corresponding physical value range. Adjacent ranges from the same source are combined. Only the first incoming run can continue the buffered tail, so the boundary check is an O(1) source-and-index comparison.

For example, consider source run ends [3, 6, 10] with values [A, B, C], and copy the logical slice [2, 7):

source rows:       AAA | BBB | CCCC
selected rows:       A | BBB | C
relative run ends:   1     4   5

If the output already contains 4 rows, the run ends become [5, 8, 9]. If A is also the buffered tail run, the first existing run end is extended to 5 and only [B, C] is retained. Otherwise [A, B, C] is retained.

Sparse filters use the same accumulator directly. Because filter selections are ordered, the implementation walks the source run ends once instead of materializing a filtered RunArray and searching for every selected range. Denser filters retain the existing materialized path, with a lower cutoff for nullable values.

Equal values in different physical runs remain distinct, matching concat and interleave semantics. finish uses Arrow's existing type-specific concat kernels for the recorded value ranges.

Are these changes tested?

The shared multi-column coalescer tests include a nullable RunArray and exercise sparse and dense filters, empty batches, non-null variants, and the single-column path. The full arrow-select test suite and strict Clippy checks pass.

Benchmark Result Runs
coalesce_kernels / single_ree All 52 reported cases improve: filters by 1.07-11.89x, take by 1.06-1.27x, and interleave by 1.04-1.18x ARM64, AMD64

Are there any significant user-facing changes?

No.

@boring-cyborg boring-cyborg Bot added the arrow-kernels Crates that transform arrow data (e.g. compute, cast) label Nov 12, 2026
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "filter: single_ree"
resources:
arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (9c6602c) to cd91f89 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (9c6602c) to cd91f89 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1
Details
group                                                       base                                   changed
-----                                                       ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001      8.41    342.1±5.44ms        ? ?/sec    1.00     40.7±0.52ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005      4.07     58.7±1.04ms        ? ?/sec    1.00     14.4±0.35ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01       2.44     26.9±0.43ms        ? ?/sec    1.00     11.0±0.20ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1        1.16      2.6±0.03ms        ? ?/sec    1.00      2.3±0.04ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8        1.17    321.1±5.45µs        ? ?/sec    1.00    273.8±4.87µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001    8.23    352.5±4.83ms        ? ?/sec    1.00     42.9±0.70ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005    3.91     60.3±1.05ms        ? ?/sec    1.00     15.4±0.25ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01     2.37     27.9±0.80ms        ? ?/sec    1.00     11.8±0.23ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.19      2.8±0.05ms        ? ?/sec    1.00      2.3±0.03ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.17    333.5±5.62µs        ? ?/sec    1.00    284.3±5.59µs        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-f0lk0bbd-7e69c37a-5rsc

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6837855954-4856-kc6m4 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 170.0s
Peak memory 35.2 MiB
Avg memory 16.4 MiB
CPU user 167.7s
CPU sys 0.3s
Peak spill 0 B

branch

Metric Value
Wall time 120.0s
Peak memory 27.6 MiB
Avg memory 15.4 MiB
CPU user 115.8s
CPU sys 0.3s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "filter: single_ree"
resources:
arch: amd64

1 similar comment
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "filter: single_ree"
resources:
arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (9c6602c) to cd91f89 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (9c6602c) to cd91f89 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (9c6602c) to cd91f89 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1
Details
group                                                       base                                   changed
-----                                                       ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001      8.66    347.9±3.72ms        ? ?/sec    1.00     40.2±0.65ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005      4.10     58.7±0.75ms        ? ?/sec    1.00     14.3±0.29ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01       2.42     26.8±0.61ms        ? ?/sec    1.00     11.1±0.31ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1        1.19      2.7±0.08ms        ? ?/sec    1.00      2.2±0.03ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8        1.20    323.2±4.97µs        ? ?/sec    1.00    269.4±2.28µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001    8.38    349.3±4.59ms        ? ?/sec    1.00     41.7±0.32ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005    3.94     60.1±1.24ms        ? ?/sec    1.00     15.2±0.13ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01     2.39     27.5±1.14ms        ? ?/sec    1.00     11.5±0.14ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.19      2.8±0.04ms        ? ?/sec    1.00      2.3±0.03ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.19    330.6±5.08µs        ? ?/sec    1.00    278.5±3.34µs        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-f0lk0bbd-7e69c37a-5rsc

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6838051194-4859-mj62g 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 175.0s
Peak memory 19.1 MiB
Avg memory 15.9 MiB
CPU user 168.8s
CPU sys 0.3s
Peak spill 0 B

branch

Metric Value
Wall time 120.0s
Peak memory 22.9 MiB
Avg memory 15.0 MiB
CPU user 113.9s
CPU sys 0.2s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (9c6602c) to cd91f89 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1
Details
group                                                       base                                   changed
-----                                                       ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001      7.49    389.1±2.99ms        ? ?/sec    1.00     51.9±1.20ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005      3.81     66.5±1.30ms        ? ?/sec    1.00     17.5±0.24ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01       2.23     30.2±0.80ms        ? ?/sec    1.00     13.6±0.15ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1        1.01      3.0±0.04ms        ? ?/sec    1.00      3.0±0.04ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8        1.01    361.2±7.10µs        ? ?/sec    1.00    356.6±8.80µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001    7.34    399.4±4.32ms        ? ?/sec    1.00     54.4±1.05ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005    3.60     68.5±1.48ms        ? ?/sec    1.00     19.0±0.32ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01     2.15     31.3±0.86ms        ? ?/sec    1.00     14.6±0.18ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.02      3.1±0.08ms        ? ?/sec    1.00      3.1±0.06ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.01    371.5±7.71µs        ? ?/sec    1.00   366.2±13.21µs        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-f0lk0bbd-28ec8992-mpmz

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6838051798-4860-s9k4h 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 180.0s
Peak memory 22.3 MiB
Avg memory 15.9 MiB
CPU user 177.5s
CPU sys 0.3s
Peak spill 0 B

branch

Metric Value
Wall time 105.0s
Peak memory 30.3 MiB
Avg memory 15.5 MiB
CPU user 100.8s
CPU sys 0.3s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid Rich-T-kid changed the title Improve sparse filtering for run-end encoded arrays Specialize BatchCoalescer for RunArray Nov 13, 2026
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "filter: single_ree"
resources:
arch: amd64

1 similar comment
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "filter: single_ree"
resources:
arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "filter: single_ree"
resources:
arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1
Details
group                                                       base                                   changed
-----                                                       ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001      10.01   348.5±4.78ms        ? ?/sec    1.00     34.8±0.54ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005      4.98     59.5±1.16ms        ? ?/sec    1.00     12.0±0.17ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01       3.02     27.1±0.42ms        ? ?/sec    1.00      9.0±0.12ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1        1.00      2.7±0.04ms        ? ?/sec    1.02      2.7±0.06ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8        1.00    324.6±7.75µs        ? ?/sec    1.01    327.3±5.74µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001    10.18   351.3±4.68ms        ? ?/sec    1.00     34.5±0.63ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005    4.73     60.6±0.77ms        ? ?/sec    1.00     12.8±0.17ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01     2.91     28.1±0.65ms        ? ?/sec    1.00      9.7±0.16ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.01      2.8±0.04ms        ? ?/sec    1.00      2.8±0.04ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.00    333.5±5.25µs        ? ?/sec    1.01    337.2±3.91µs        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-1ey2y0e2-efdd6ef4-4xr5

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6841281482-4868-62tvh 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 175.0s
Peak memory 26.7 MiB
Avg memory 15.7 MiB
CPU user 168.6s
CPU sys 0.3s
Peak spill 0 B

branch

Metric Value
Wall time 115.0s
Peak memory 24.0 MiB
Avg memory 14.9 MiB
CPU user 108.8s
CPU sys 0.3s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "filter: single_ree"
resources:
arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1
Details
group                                                       base                                   changed
-----                                                       ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001      10.26   338.3±1.74ms        ? ?/sec    1.00     33.0±0.45ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005      4.96     57.5±0.75ms        ? ?/sec    1.00     11.6±0.17ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01       3.06     26.3±0.16ms        ? ?/sec    1.00      8.6±0.12ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1        1.00      2.6±0.07ms        ? ?/sec    1.02      2.7±0.02ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8        1.00    315.8±1.52µs        ? ?/sec    1.02    323.2±2.98µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001    10.23   342.4±3.06ms        ? ?/sec    1.00     33.5±0.39ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005    4.62     58.8±0.33ms        ? ?/sec    1.00     12.7±0.24ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01     2.83     27.2±0.13ms        ? ?/sec    1.00      9.6±0.23ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.00      2.7±0.02ms        ? ?/sec    1.01      2.8±0.05ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.00    325.5±2.89µs        ? ?/sec    1.02    333.2±2.25µs        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-1ey2y0e2-c6aa1932-89l4

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6841282854-4869-pwx6q 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 170.0s
Peak memory 22.7 MiB
Avg memory 16.0 MiB
CPU user 165.8s
CPU sys 0.2s
Peak spill 0 B

branch

Metric Value
Wall time 115.0s
Peak memory 30.1 MiB
Avg memory 15.6 MiB
CPU user 111.9s
CPU sys 0.3s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1
Details
group                                                       base                                   changed
-----                                                       ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001      9.90    342.5±5.08ms        ? ?/sec    1.00     34.6±0.69ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005      4.87     58.1±1.09ms        ? ?/sec    1.00     11.9±0.25ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01       3.01     26.6±0.47ms        ? ?/sec    1.00      8.9±0.18ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1        1.00      2.6±0.05ms        ? ?/sec    1.03      2.7±0.05ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8        1.00    321.0±4.81µs        ? ?/sec    1.03    329.9±5.89µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001    10.03   346.5±4.75ms        ? ?/sec    1.00     34.6±0.64ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005    4.60     60.0±1.23ms        ? ?/sec    1.00     13.1±0.24ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01     2.78     27.5±0.48ms        ? ?/sec    1.00      9.9±0.15ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.00      2.8±0.05ms        ? ?/sec    1.01      2.8±0.05ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.00    326.1±3.64µs        ? ?/sec    1.04    340.0±7.53µs        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-1ey2y0e2-efdd6ef4-5vcj

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6841282192-4870-nzkr9 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 170.0s
Peak memory 52.9 MiB
Avg memory 16.2 MiB
CPU user 166.5s
CPU sys 0.3s
Peak spill 0 B

branch

Metric Value
Wall time 115.0s
Peak memory 22.9 MiB
Avg memory 15.3 MiB
CPU user 110.9s
CPU sys 0.2s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1
Details
group                                                       base                                   changed
-----                                                       ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001      10.09   344.9±4.39ms        ? ?/sec    1.00     34.2±0.74ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005      4.84     58.6±1.09ms        ? ?/sec    1.00     12.1±0.21ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01       2.99     26.7±0.38ms        ? ?/sec    1.00      8.9±0.19ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1        1.00      2.6±0.05ms        ? ?/sec    1.03      2.7±0.04ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8        1.00    316.9±3.40µs        ? ?/sec    1.04    329.0±8.04µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001    9.98    345.1±4.10ms        ? ?/sec    1.00     34.6±0.81ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005    4.65     60.1±1.24ms        ? ?/sec    1.00     12.9±0.26ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01     2.79     27.6±0.44ms        ? ?/sec    1.00      9.9±0.20ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.00      2.8±0.09ms        ? ?/sec    1.00      2.8±0.04ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.00    331.6±7.68µs        ? ?/sec    1.01    336.4±4.15µs        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-1ey2y0e2-efdd6ef4-5vcj

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6841425014-4873-fxpgk 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 170.0s
Peak memory 20.5 MiB
Avg memory 16.0 MiB
CPU user 166.7s
CPU sys 0.3s
Peak spill 0 B

branch

Metric Value
Wall time 115.0s
Peak memory 24.5 MiB
Avg memory 15.4 MiB
CPU user 111.9s
CPU sys 0.3s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "filter: single_ree"
resources:
arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (0207495) to 4c03e90 (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1
Details
group                                                       base                                   changed
-----                                                       ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001      10.11   346.7±4.81ms        ? ?/sec    1.00     34.3±0.92ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005      4.93     58.6±1.37ms        ? ?/sec    1.00     11.9±0.22ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01       3.02     26.8±0.37ms        ? ?/sec    1.00      8.9±0.16ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1        1.00      2.7±0.04ms        ? ?/sec    1.01      2.7±0.04ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8        1.00    319.9±7.15µs        ? ?/sec    1.03    329.6±5.37µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001    10.24   350.2±5.73ms        ? ?/sec    1.00     34.2±0.82ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005    4.68     60.8±1.05ms        ? ?/sec    1.00     13.0±0.24ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01     2.90     28.2±0.96ms        ? ?/sec    1.00      9.7±0.15ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.00      2.8±0.09ms        ? ?/sec    1.00      2.8±0.06ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.00   333.8±12.28µs        ? ?/sec    1.01    338.5±7.13µs        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-1ey2y0e2-efdd6ef4-5vcj

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6841579175-4874-5l87p 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 175.0s
Peak memory 29.4 MiB
Avg memory 15.8 MiB
CPU user 169.7s
CPU sys 0.3s
Peak spill 0 B

branch

Metric Value
Wall time 115.0s
Peak memory 26.3 MiB
Avg memory 15.2 MiB
CPU user 108.9s
CPU sys 0.2s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels

env:
  BENCH_FILTER: "filter: single_ree"
resources:
  arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (a3e3cb5) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: 'filter: single_ree'
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

Hi @Rich-T-kid, thanks for the request (#12218 (comment)).

Usage:

run benchmark <name>           # run specific benchmark(s)
run benchmarks                 # run default suite
run benchmarks <name1> <name2> # run specific benchmarks

Any benchmark name is accepted: bench.sh suite names (e.g. tpch, clickbench_partitioned, wide_schema) and Criterion bench targets (e.g. sql_planner) are resolved automatically. A name that matches neither fails on the runner.

Per-side configuration (run benchmark tpch followed by):

env:
# shared env is inherited by BOTH the build and the run, so build
# flags go here. Builds default to no debuginfo for speed; opt back
# in for hung-job gdb dumps and cap jobs to stay within memory:
CARGO_PROFILE_RELEASE_DEBUG: "1"
CARGO_BUILD_JOBS: "1"
baseline:
ref: v45.0.0
env:
# per-side env only reaches the benchmark run, not the build
DATAFUSION_RUNTIME_MEMORY_LIMIT: 1G
changed:
ref: v46.0.0
env:
DATAFUSION_RUNTIME_MEMORY_LIMIT: 2G

Pod size and CPU architecture (each key is optional and keeps the default when absent; requests and limits are set equal):

resources:
cpu: "16"
memory: "128Gi"
arch: arm64          # arm64 or amd64

Independent benchmark shards (optional; 1–8, default 1):

run benchmark arrow_writer
shards: 4

Each shard builds both revisions, measures its assigned cases, and posts its own report. Resources and the job deadline apply per shard. DataFusion requests specifying a shard count greater than 1 fail with an error. Omit shards or use shards: 1 for DataFusion.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

Hi @Rich-T-kid, thanks for the request (#12218 (comment)).

Usage:

run benchmark <name>           # run specific benchmark(s)
run benchmarks                 # run default suite
run benchmarks <name1> <name2> # run specific benchmarks

Any benchmark name is accepted: bench.sh suite names (e.g. tpch, clickbench_partitioned, wide_schema) and Criterion bench targets (e.g. sql_planner) are resolved automatically. A name that matches neither fails on the runner.

Per-side configuration (run benchmark tpch followed by):

env:
# shared env is inherited by BOTH the build and the run, so build
# flags go here. Builds default to no debuginfo for speed; opt back
# in for hung-job gdb dumps and cap jobs to stay within memory:
CARGO_PROFILE_RELEASE_DEBUG: "1"
CARGO_BUILD_JOBS: "1"
baseline:
ref: v45.0.0
env:
# per-side env only reaches the benchmark run, not the build
DATAFUSION_RUNTIME_MEMORY_LIMIT: 1G
changed:
ref: v46.0.0
env:
DATAFUSION_RUNTIME_MEMORY_LIMIT: 2G

Pod size and CPU architecture (each key is optional and keeps the default when absent; requests and limits are set equal):

resources:
cpu: "16"
memory: "128Gi"
arch: arm64          # arm64 or amd64

Independent benchmark shards (optional; 1–8, default 1):

run benchmark arrow_writer
shards: 4

Each shard builds both revisions, measures its assigned cases, and posts its own report. Resources and the job deadline apply per shard. DataFusion requests specifying a shard count greater than 1 fail with an error. Omit shards or use shards: 1 for DataFusion.


File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "single_ree"
resources:
arch: arm64

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "single_ree"
resources:
arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (b65ff90) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: arm64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (b65ff90) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (b65ff90) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: arm64
shards: 1
Details
group                                                           base                                   changed
-----                                                           ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001          11.68   370.2±1.17ms        ? ?/sec    1.00     31.7±0.17ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005          2.42     60.3±0.04ms        ? ?/sec    1.00     25.0±0.19ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01           1.37     27.4±0.02ms        ? ?/sec    1.00     20.0±0.11ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1            1.11      2.7±0.00ms        ? ?/sec    1.00      2.4±0.01ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8            1.11    328.7±0.74µs        ? ?/sec    1.00    297.0±0.73µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001        9.91    376.2±0.51ms        ? ?/sec    1.00     38.0±0.11ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005        2.00     61.9±0.05ms        ? ?/sec    1.00     31.0±0.07ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01         1.13     28.1±0.01ms        ? ?/sec    1.00     24.9±0.06ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1          1.11      2.8±0.00ms        ? ?/sec    1.00      2.5±0.01ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8          1.10    338.1±0.71µs        ? ?/sec    1.00    308.3±0.97µs        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.001      1.08     73.5±0.12ms        ? ?/sec    1.00     68.1±0.10ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.01       1.10     16.1±0.03ms        ? ?/sec    1.00     14.6±0.05ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.1        1.07      6.6±0.02ms        ? ?/sec    1.00      6.1±0.02ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.8        1.08      5.2±0.01ms        ? ?/sec    1.00      4.8±0.01ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.001    1.09     78.4±0.15ms        ? ?/sec    1.00     72.2±0.15ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.01     1.08     17.2±0.02ms        ? ?/sec    1.00     16.0±0.03ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.05      7.0±0.02ms        ? ?/sec    1.00      6.7±0.02ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.07      5.6±0.01ms        ? ?/sec    1.00      5.2±0.01ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.001            1.19     33.2±0.12ms        ? ?/sec    1.00     27.9±0.06ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.01             1.22      8.0±0.02ms        ? ?/sec    1.00      6.6±0.01ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.1              1.17      4.8±0.01ms        ? ?/sec    1.00      4.1±0.00ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.8              1.10      4.7±0.00ms        ? ?/sec    1.00      4.3±0.01ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.001          1.19     42.2±0.36ms        ? ?/sec    1.00     35.5±0.18ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.01           1.18     10.2±0.02ms        ? ?/sec    1.00      8.7±0.04ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.1            1.10      6.0±0.01ms        ? ?/sec    1.00      5.5±0.01ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.8            1.08      5.9±0.00ms        ? ?/sec    1.00      5.5±0.00ms        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-60ma4f95-6c11b3b1-72z8

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

uname:

Linux bench-c6952363957-5305-l8724 6.12.105+ #1 SMP Sat Sep 19 07:55:25 UTC 2026 aarch64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
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
Resource Usage

base (merge-base)

Metric Value
Wall time 350.1s
Peak memory 29.5 MiB
Avg memory 15.7 MiB
CPU user 346.7s
CPU sys 0.3s
Peak spill 0 B

branch

Metric Value
Wall time 280.1s
Peak memory 23.8 MiB
Avg memory 14.7 MiB
CPU user 277.0s
CPU sys 0.3s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (b65ff90) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1
Details
group                                                           base                                   changed
-----                                                           ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001          10.57   345.2±7.34ms        ? ?/sec    1.00     32.7±1.48ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005          2.53     57.5±2.23ms        ? ?/sec    1.00     22.7±0.80ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01           1.46     26.5±1.02ms        ? ?/sec    1.00     18.1±0.69ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1            1.16      2.6±0.10ms        ? ?/sec    1.00      2.3±0.10ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8            1.14   320.3±14.04µs        ? ?/sec    1.00   280.3±13.76µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001        9.21    351.8±6.74ms        ? ?/sec    1.00     38.2±1.45ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005        2.08     60.1±2.29ms        ? ?/sec    1.00     28.9±1.11ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01         1.20     27.8±1.12ms        ? ?/sec    1.00     23.2±0.84ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1          1.16      2.8±0.11ms        ? ?/sec    1.00      2.4±0.13ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8          1.02   333.1±15.83µs        ? ?/sec    1.00    326.8±5.45µs        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.001      1.04     62.9±3.24ms        ? ?/sec    1.00     60.3±2.11ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.01       1.16     17.7±0.69ms        ? ?/sec    1.00     15.3±0.58ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.1        1.06      9.6±0.33ms        ? ?/sec    1.00      9.1±0.29ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.8        1.05      7.7±0.27ms        ? ?/sec    1.00      7.3±0.28ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.001    1.04     67.3±2.52ms        ? ?/sec    1.00     64.6±2.39ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.01     1.10     18.7±0.68ms        ? ?/sec    1.00     17.0±0.60ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.07     10.4±0.39ms        ? ?/sec    1.00      9.7±0.31ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.04      8.2±0.34ms        ? ?/sec    1.00      7.9±0.31ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.001            1.00     32.0±1.07ms        ? ?/sec    1.03     32.8±0.46ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.01             1.05      9.4±0.35ms        ? ?/sec    1.00      9.0±0.12ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.1              1.04      7.9±0.30ms        ? ?/sec    1.00      7.6±0.31ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.8              1.03      7.3±0.28ms        ? ?/sec    1.00      7.1±0.25ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.001          1.09     39.0±1.44ms        ? ?/sec    1.00     35.9±1.26ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.01           1.24     11.7±0.43ms        ? ?/sec    1.00      9.4±0.41ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.1            1.10      9.0±0.32ms        ? ?/sec    1.00      8.2±0.31ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.8            1.09      8.3±0.44ms        ? ?/sec    1.00      7.6±0.24ms        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-1n18gwqc-38f4eb21-k2tl

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6952364530-5306-4n6vq 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 340.1s
Peak memory 23.7 MiB
Avg memory 17.1 MiB
CPU user 335.7s
CPU sys 0.3s
Peak spill 0 B

branch

Metric Value
Wall time 285.1s
Peak memory 25.8 MiB
Avg memory 16.2 MiB
CPU user 282.0s
CPU sys 0.4s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "single_ree"
resources:
arch: amd64

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
baseline:
ref: b3c1346
env:
BENCH_FILTER: "single_ree"
resources:
arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (b65ff90) to b3c1346 diff

Run configuration
run benchmark coalesce_kernels
baseline:
  ref: b3c134666cb6afc59eba9918e761faf2e9945840
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (b65ff90) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (b65ff90) to b3c1346 diff

Run configuration
run benchmark coalesce_kernels
baseline:
  ref: b3c134666cb6afc59eba9918e761faf2e9945840
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1
Details
group                                                           base                                   changed
-----                                                           ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001          1.00     35.7±1.08ms        ? ?/sec    1.01     36.0±0.50ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005          1.01     25.2±0.67ms        ? ?/sec    1.00     25.0±0.39ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01           1.01     20.4±0.25ms        ? ?/sec    1.00     20.2±0.25ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1            1.10      2.8±0.05ms        ? ?/sec    1.00      2.5±0.03ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8            1.09    336.2±6.27µs        ? ?/sec    1.00    308.2±4.67µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001        1.00     43.1±0.57ms        ? ?/sec    1.01     43.4±0.77ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005        1.00     32.5±0.88ms        ? ?/sec    1.02     33.0±0.49ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01         1.00     26.1±0.53ms        ? ?/sec    1.01     26.4±0.55ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1          1.10      2.9±0.06ms        ? ?/sec    1.00      2.6±0.04ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8          1.10    354.1±5.11µs        ? ?/sec    1.00    323.3±5.14µs        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.001      1.07     73.6±1.59ms        ? ?/sec    1.00     69.0±0.44ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.01       1.02     17.9±0.32ms        ? ?/sec    1.00     17.4±0.24ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.1        1.00      9.9±0.13ms        ? ?/sec    1.02     10.1±0.30ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.8        1.00      8.2±0.13ms        ? ?/sec    1.01      8.3±0.15ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.001    1.08     78.6±1.09ms        ? ?/sec    1.00     73.0±0.66ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.01     1.03     19.7±0.35ms        ? ?/sec    1.00     19.1±0.34ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.00     10.6±0.17ms        ? ?/sec    1.05     11.1±0.14ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.00      8.7±0.19ms        ? ?/sec    1.02      8.9±0.15ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.001            1.00     31.8±0.44ms        ? ?/sec    1.02     32.3±0.46ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.01             1.00      8.6±0.15ms        ? ?/sec    1.05      9.0±0.38ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.1              1.00      8.3±0.07ms        ? ?/sec    1.04      8.7±0.23ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.8              1.00      7.5±0.10ms        ? ?/sec    1.08      8.2±0.22ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.001          1.00     40.7±0.57ms        ? ?/sec    1.02     41.7±0.45ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.01           1.00     10.9±0.45ms        ? ?/sec    1.00     10.9±0.10ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.1            1.00      9.2±0.23ms        ? ?/sec    1.03      9.4±0.11ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.8            1.00      8.7±0.15ms        ? ?/sec    1.01      8.8±0.07ms        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-1odthwem-92f17ebb-szgw

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6952827906-5319-q9b4z 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 295.1s
Peak memory 25.8 MiB
Avg memory 15.8 MiB
CPU user 291.6s
CPU sys 0.4s
Peak spill 0 B

branch

Metric Value
Wall time 295.1s
Peak memory 21.9 MiB
Avg memory 16.8 MiB
CPU user 289.1s
CPU sys 0.3s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (b65ff90) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1
Details
group                                                           base                                   changed
-----                                                           ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001          11.36   399.0±4.09ms        ? ?/sec    1.00     35.1±0.58ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005          2.73     67.0±0.52ms        ? ?/sec    1.00     24.5±0.45ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01           1.55     30.6±0.28ms        ? ?/sec    1.00     19.7±0.35ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1            1.23      3.1±0.08ms        ? ?/sec    1.00      2.5±0.03ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8            1.21    365.9±3.79µs        ? ?/sec    1.00    302.5±5.39µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001        9.36    405.0±2.41ms        ? ?/sec    1.00     43.3±1.33ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005        2.04     69.3±0.52ms        ? ?/sec    1.00     34.0±1.02ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01         1.18     31.7±0.16ms        ? ?/sec    1.00     26.9±0.11ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1          1.16      3.2±0.02ms        ? ?/sec    1.00      2.7±0.01ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8          1.15    382.8±2.69µs        ? ?/sec    1.00    333.6±5.59µs        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.001      1.00     66.3±2.30ms        ? ?/sec    1.04     69.2±0.35ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.01       1.02     18.1±0.62ms        ? ?/sec    1.00     17.8±0.14ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.1        1.00     10.0±0.33ms        ? ?/sec    1.07     10.6±0.09ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.8        1.00      8.0±0.32ms        ? ?/sec    1.06      8.5±0.05ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.001    1.00     69.7±2.16ms        ? ?/sec    1.06     73.7±0.44ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.01     1.00     19.6±0.85ms        ? ?/sec    1.00     19.6±0.09ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.00     10.3±0.33ms        ? ?/sec    1.08     11.1±0.28ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.00      8.3±0.30ms        ? ?/sec    1.08      9.0±0.04ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.001            1.10     36.9±0.23ms        ? ?/sec    1.00     33.6±0.15ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.01             1.18     10.9±0.29ms        ? ?/sec    1.00      9.2±0.04ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.1              1.06      9.1±0.09ms        ? ?/sec    1.00      8.7±0.10ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.8              1.01      8.4±0.11ms        ? ?/sec    1.00      8.3±0.05ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.001          1.07     44.8±0.73ms        ? ?/sec    1.00     41.8±0.22ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.01           1.23     13.4±0.08ms        ? ?/sec    1.00     10.9±0.18ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.1            1.09     10.3±0.10ms        ? ?/sec    1.00      9.4±0.11ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.8            1.03      9.2±0.43ms        ? ?/sec    1.00      8.9±0.10ms        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-19rx0sh2-c32e1e2a-n95p

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6952827382-5320-x6sgc 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 360.1s
Peak memory 28.9 MiB
Avg memory 17.0 MiB
CPU user 351.1s
CPU sys 0.4s
Peak spill 0 B

branch

Metric Value
Wall time 300.1s
Peak memory 26.9 MiB
Avg memory 15.6 MiB
CPU user 294.0s
CPU sys 0.4s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "single_ree"
resources:
arch: arm64

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
env:
BENCH_FILTER: "single_ree"
resources:
arch: amd64

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
baseline:
ref: b3c1346
env:
BENCH_FILTER: "single_ree"
resources:
arch: arm64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (ab4370a) to b3c1346 diff

Run configuration
run benchmark coalesce_kernels
baseline:
  ref: b3c134666cb6afc59eba9918e761faf2e9945840
env:
  BENCH_FILTER: single_ree
resources:
  arch: arm64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (ab4370a) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (ab4370a) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: arm64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (ab4370a) to b3c1346 diff

Run configuration
run benchmark coalesce_kernels
baseline:
  ref: b3c134666cb6afc59eba9918e761faf2e9945840
env:
  BENCH_FILTER: single_ree
resources:
  arch: arm64
shards: 1
Details
group                                                           base                                   changed
-----                                                           ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001          1.00     31.9±0.11ms        ? ?/sec    1.00     32.0±0.16ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005          1.00     24.9±0.12ms        ? ?/sec    1.00     24.9±0.12ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01           1.01     20.1±0.13ms        ? ?/sec    1.00     20.0±0.10ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1            1.00      2.4±0.00ms        ? ?/sec    1.00      2.4±0.01ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8            1.01    299.1±0.81µs        ? ?/sec    1.00    297.3±0.81µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001        1.02     38.8±0.22ms        ? ?/sec    1.00     38.2±0.10ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005        1.01     31.7±0.14ms        ? ?/sec    1.00     31.2±0.09ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01         1.01     25.3±0.12ms        ? ?/sec    1.00     25.2±0.07ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1          1.01      2.5±0.01ms        ? ?/sec    1.00      2.5±0.01ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8          1.00    310.0±0.84µs        ? ?/sec    1.00    309.5±0.83µs        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.001      1.01     68.5±0.23ms        ? ?/sec    1.00     68.0±0.08ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.01       1.00     14.6±0.06ms        ? ?/sec    1.00     14.6±0.06ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.1        1.00      6.2±0.02ms        ? ?/sec    1.01      6.2±0.02ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.8        1.00      4.8±0.01ms        ? ?/sec    1.01      4.9±0.02ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.001    1.00     72.4±0.15ms        ? ?/sec    1.00     72.5±0.13ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.01     1.00     15.9±0.05ms        ? ?/sec    1.01     16.0±0.04ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.00      6.7±0.02ms        ? ?/sec    1.01      6.8±0.02ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.00      5.3±0.02ms        ? ?/sec    1.00      5.3±0.02ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.001            1.00     27.8±0.09ms        ? ?/sec    1.01     28.0±0.14ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.01             1.00      6.6±0.01ms        ? ?/sec    1.00      6.6±0.02ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.1              1.00      4.1±0.01ms        ? ?/sec    1.00      4.1±0.01ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.8              1.00      4.3±0.02ms        ? ?/sec    1.00      4.3±0.01ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.001          1.00     35.5±0.11ms        ? ?/sec    1.00     35.4±0.11ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.01           1.00      8.7±0.04ms        ? ?/sec    1.00      8.7±0.03ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.1            1.00      5.5±0.01ms        ? ?/sec    1.00      5.5±0.01ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.8            1.00      5.6±0.01ms        ? ?/sec    1.00      5.6±0.01ms        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-aznwbglv-faa633b0-sxjk

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

uname:

Linux bench-c6953653357-5321-44gl7 6.12.105+ #1 SMP Sat Sep 19 07:55:25 UTC 2026 aarch64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
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
Resource Usage

base (merge-base)

Metric Value
Wall time 285.1s
Peak memory 27.4 MiB
Avg memory 14.5 MiB
CPU user 280.6s
CPU sys 0.4s
Peak spill 0 B

branch

Metric Value
Wall time 280.1s
Peak memory 26.1 MiB
Avg memory 14.6 MiB
CPU user 277.0s
CPU sys 0.3s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (ab4370a) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1
Details
group                                                           base                                   changed
-----                                                           ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001          10.78   398.5±7.46ms        ? ?/sec    1.00     37.0±0.45ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005          2.60     67.0±0.42ms        ? ?/sec    1.00     25.7±0.52ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01           1.48     30.7±0.30ms        ? ?/sec    1.00     20.8±0.30ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1            1.08      3.0±0.02ms        ? ?/sec    1.00      2.8±0.03ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8            1.09    367.5±3.87µs        ? ?/sec    1.00    337.4±8.07µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001        9.14    404.8±2.65ms        ? ?/sec    1.00     44.3±0.60ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005        2.06     69.3±0.69ms        ? ?/sec    1.00     33.6±0.32ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01         1.19     31.8±0.39ms        ? ?/sec    1.00     26.8±0.14ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1          1.07      3.2±0.02ms        ? ?/sec    1.00      2.9±0.05ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8          1.09    384.1±8.17µs        ? ?/sec    1.00    351.3±3.06µs        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.001      1.05     71.9±0.20ms        ? ?/sec    1.00     68.5±1.52ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.01       1.18     20.0±0.57ms        ? ?/sec    1.00     17.0±0.29ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.1        1.11     11.1±0.08ms        ? ?/sec    1.00     10.0±0.20ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.8        1.08      8.8±0.14ms        ? ?/sec    1.00      8.2±0.21ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.001    1.04     77.4±0.83ms        ? ?/sec    1.00     74.3±1.43ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.01     1.12     21.6±0.07ms        ? ?/sec    1.00     19.3±0.44ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.10     11.7±0.10ms        ? ?/sec    1.00     10.7±0.33ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.09      9.3±0.25ms        ? ?/sec    1.00      8.5±0.31ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.001            1.18     36.9±0.21ms        ? ?/sec    1.00     31.2±0.12ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.01             1.27     10.9±0.06ms        ? ?/sec    1.00      8.6±0.05ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.1              1.09      9.2±0.14ms        ? ?/sec    1.00      8.4±0.08ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.8              1.06      8.4±0.14ms        ? ?/sec    1.00      8.0±0.04ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.001          1.16     44.7±0.33ms        ? ?/sec    1.00     38.4±0.60ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.01           1.26     13.5±0.09ms        ? ?/sec    1.00     10.7±0.46ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.1            1.10     10.4±0.08ms        ? ?/sec    1.00      9.5±0.05ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.8            1.12      9.5±0.09ms        ? ?/sec    1.00      8.5±0.25ms        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-19rx0sh2-c32e1e2a-n95p

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6953652848-5322-js52h 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 365.1s
Peak memory 23.8 MiB
Avg memory 17.3 MiB
CPU user 359.5s
CPU sys 0.4s
Peak spill 0 B

branch

Metric Value
Wall time 290.1s
Peak memory 28.7 MiB
Avg memory 16.5 MiB
CPU user 284.9s
CPU sys 0.4s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (ab4370a) to 4e04fea (merge-base) diff

Run configuration
run benchmark coalesce_kernels
env:
  BENCH_FILTER: single_ree
resources:
  arch: arm64
shards: 1
Details
group                                                           base                                   changed
-----                                                           ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001          11.89   380.3±1.85ms        ? ?/sec    1.00     32.0±0.19ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005          2.40     60.7±0.13ms        ? ?/sec    1.00     25.3±0.17ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01           1.36     27.6±0.06ms        ? ?/sec    1.00     20.3±0.12ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1            1.10      2.7±0.01ms        ? ?/sec    1.00      2.4±0.01ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8            1.10    330.2±0.88µs        ? ?/sec    1.00    298.9±0.91µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001        9.63    384.6±1.82ms        ? ?/sec    1.00     39.9±0.48ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005        1.94     62.3±0.14ms        ? ?/sec    1.00     32.2±0.34ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01         1.11     28.3±0.07ms        ? ?/sec    1.00     25.5±0.17ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1          1.11      2.8±0.01ms        ? ?/sec    1.00      2.5±0.01ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8          1.09    340.1±1.12µs        ? ?/sec    1.00    310.7±1.70µs        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.001      1.08     74.8±0.48ms        ? ?/sec    1.00     69.1±0.26ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.01       1.11     16.3±0.06ms        ? ?/sec    1.00     14.7±0.09ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.1        1.04      6.6±0.03ms        ? ?/sec    1.00      6.4±0.04ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.8        1.08      5.4±0.04ms        ? ?/sec    1.00      5.0±0.06ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.001    1.11     80.8±0.70ms        ? ?/sec    1.00     73.0±0.27ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.01     1.08     17.3±0.05ms        ? ?/sec    1.00     16.0±0.06ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.04      7.1±0.03ms        ? ?/sec    1.00      6.8±0.03ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.06      5.8±0.04ms        ? ?/sec    1.00      5.4±0.04ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.001            1.20     33.9±0.24ms        ? ?/sec    1.00     28.3±0.20ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.01             1.22      8.1±0.03ms        ? ?/sec    1.00      6.7±0.04ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.1              1.15      4.8±0.01ms        ? ?/sec    1.00      4.1±0.02ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.8              1.08      4.8±0.02ms        ? ?/sec    1.00      4.4±0.02ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.001          1.20     43.1±0.34ms        ? ?/sec    1.00     36.0±0.21ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.01           1.17     10.3±0.06ms        ? ?/sec    1.00      8.8±0.05ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.1            1.10      6.1±0.02ms        ? ?/sec    1.00      5.5±0.02ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.8            1.07      6.0±0.03ms        ? ?/sec    1.00      5.6±0.02ms        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-aznwbglv-d7f3a719-pl7l

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

uname:

Linux bench-c6953652383-5323-htftc 6.12.105+ #1 SMP Sat Sep 19 07:55:25 UTC 2026 aarch64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
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
Resource Usage

base (merge-base)

Metric Value
Wall time 355.1s
Peak memory 21.9 MiB
Avg memory 15.6 MiB
CPU user 350.6s
CPU sys 0.4s
Peak spill 0 B

branch

Metric Value
Wall time 285.1s
Peak memory 23.2 MiB
Avg memory 14.6 MiB
CPU user 283.0s
CPU sys 0.4s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark coalesce_kernels
baseline:
ref: b3c1346
env:
BENCH_FILTER: "single_ree"
resources:
arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark starting (GKE) | trigger

Target: coalesce_kernels

Sharding factor: 1 (1 workers).

Comparing codex/issue-11720 (ab4370a) to b3c1346 diff

Run configuration
run benchmark coalesce_kernels
baseline:
  ref: b3c134666cb6afc59eba9918e761faf2e9945840
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1

Results will be posted when all workers finish.


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

Target: coalesce_kernels

Comparing codex/issue-11720 (ab4370a) to b3c1346 diff

Run configuration
run benchmark coalesce_kernels
baseline:
  ref: b3c134666cb6afc59eba9918e761faf2e9945840
env:
  BENCH_FILTER: single_ree
resources:
  arch: amd64
shards: 1
Details
group                                                           base                                   changed
-----                                                           ----                                   -------
filter: single_ree, 8192, nulls: 0, selectivity: 0.001          1.00     35.6±0.61ms        ? ?/sec    1.00     35.5±0.76ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.005          1.00     24.7±0.47ms        ? ?/sec    1.00     24.6±0.39ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.01           1.00     19.7±0.31ms        ? ?/sec    1.00     19.8±0.30ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.1            1.01      2.7±0.05ms        ? ?/sec    1.00      2.7±0.06ms        ? ?/sec
filter: single_ree, 8192, nulls: 0, selectivity: 0.8            1.02    330.9±8.73µs        ? ?/sec    1.00    324.1±5.69µs        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.001        1.00     42.8±0.89ms        ? ?/sec    1.00     43.0±0.63ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.005        1.00     32.2±0.46ms        ? ?/sec    1.00     32.3±0.65ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.01         1.01     25.8±0.45ms        ? ?/sec    1.00     25.7±0.45ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.1          1.00      2.8±0.05ms        ? ?/sec    1.00      2.8±0.06ms        ? ?/sec
filter: single_ree, 8192, nulls: 0.1, selectivity: 0.8          1.02   347.4±10.93µs        ? ?/sec    1.00    340.6±7.14µs        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.001      1.06     73.1±1.74ms        ? ?/sec    1.00     68.7±1.55ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.01       1.01     17.6±0.33ms        ? ?/sec    1.00     17.5±0.28ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.1        1.03     10.1±0.15ms        ? ?/sec    1.00      9.8±0.16ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0, selectivity: 0.8        1.01      8.2±0.14ms        ? ?/sec    1.00      8.1±0.13ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.001    1.05     78.0±1.59ms        ? ?/sec    1.00     73.9±1.19ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.01     1.00     19.0±0.29ms        ? ?/sec    1.00     18.9±0.36ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.1      1.00     10.6±0.22ms        ? ?/sec    1.00     10.6±0.21ms        ? ?/sec
interleave: single_ree, 8192, nulls: 0.1, selectivity: 0.8      1.02      8.7±0.26ms        ? ?/sec    1.00      8.5±0.12ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.001            1.06     31.4±0.52ms        ? ?/sec    1.00     29.7±0.38ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.01             1.04      8.5±0.14ms        ? ?/sec    1.00      8.2±0.21ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.1              1.00      8.0±0.09ms        ? ?/sec    1.01      8.1±0.14ms        ? ?/sec
take: single_ree, 8192, nulls: 0, selectivity: 0.8              1.00      7.6±0.10ms        ? ?/sec    1.02      7.7±0.23ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.001          1.05     39.6±0.48ms        ? ?/sec    1.00     37.7±0.54ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.01           1.04     10.5±0.41ms        ? ?/sec    1.00     10.2±0.33ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.1            1.00      9.0±0.17ms        ? ?/sec    1.00      9.0±0.18ms        ? ?/sec
take: single_ree, 8192, nulls: 0.1, selectivity: 0.8            1.00      8.6±0.30ms        ? ?/sec    1.00      8.5±0.12ms        ? ?/sec

Per-runner information
coalesce_kernels — shard 1/1

Node: gk3-benchmark-cluster-nap-1uht8jxf-63c9b4e9-pvn8

Instance: c4-highmem-16 (12 vCPU / 65 GiB)

uname:

Linux bench-c6954191004-5331-cph4q 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

BENCH_COMMAND:

cargo bench --features=arrow,async,test_common,experimental,object_store --bench coalesce_kernels
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 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; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Resource Usage

base (merge-base)

Metric Value
Wall time 290.1s
Peak memory 27.2 MiB
Avg memory 15.9 MiB
CPU user 285.7s
CPU sys 0.4s
Peak spill 0 B

branch

Metric Value
Wall time 290.1s
Peak memory 26.7 MiB
Avg memory 15.7 MiB
CPU user 287.0s
CPU sys 0.4s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid
Rich-T-kid marked this pull request as ready for review November 17, 2026 16:48
@Rich-T-kid
Rich-T-kid marked this pull request as draft November 17, 2026 21:19

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

arrow-kernels Crates that transform arrow data (e.g. compute, cast)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Specialize BatchCoalescer for RunArray

2 participants