perf(NullBuffer::Exapand) optimize 8 byte chunks by Rich-T-kid · Pull Request #10976 · apache/arrow-rs

Rich-T-kid · 2026-09-03T19:08:19Z

Which issue does this PR close?

Closes Optimize NullBuffer::expand #10883.

Rationale for this change

NullBuffer::expand repeats each validity bit count times, used whenever a parent null must be propagated to a fixed-size group of child elements. The previous implementation set output bits one at a time regardless of count, which is unnecessarily slow for common sizes.

What changes are included in this PR?

Three tiered fast paths are added to try_expand, checked in order:

count % 8 == 0 : each expanded run is byte-aligned, so contiguous runs of valid bits are filled with memset(0xFF) via BitSliceIterator rather than setting bits individually.
count % 4 == 0 : each bit's range starts and ends on a nibble boundary. Full interior bytes are filled with 0xFF; the one partial boundary byte is set with |= 0x0F or |= 0xF0 depending on alignment. No inner loop.
General case : unchanged bit-by-bit path.

Are these changes tested?

yes, existing test cover this as well as 1 new test

Are there any user-facing changes?

no

Rich-T-kid · 2026-09-03T19:09:08Z

run benchmark boolean_kernels
env:
BENCH_FILTER: null_buffer_expand

adriangbot · 2026-09-03T19:12:35Z

🤖 Arrow criterion benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5530777736-2122-dtsdj 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing rich-T-kid/optimize-NullBuff-expand (704c4c9) to ec21627 (merge-base) diff

Run configuration

run benchmark boolean_kernels
env:
  BENCH_FILTER: "null_buffer_expand"

BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental,object_store --bench boolean_kernels
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-03T19:15:55Z

🤖 Arrow criterion benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-NullBuff-expand (704c4c9) to ec21627 (merge-base) diff

Run configuration

run benchmark boolean_kernels
env:
  BENCH_FILTER: "null_buffer_expand"

CPU Details (lscpu)

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

Details

group                                    main                                    rich-T-kid_optimize-NullBuff-expand
-----                                    ----                                    -----------------------------------
null_buffer_expand/len=1024/count=128    80.07    47.5±0.13µs        ? ?/sec     1.00    592.8±3.30ns        ? ?/sec
null_buffer_expand/len=1024/count=96     61.55    34.2±0.05µs        ? ?/sec     1.00    555.7±0.99ns        ? ?/sec
null_buffer_expand/len=2048/count=192    118.66   138.9±0.27µs        ? ?/sec    1.00   1170.7±6.19ns        ? ?/sec
null_buffer_expand/len=2048/count=256    144.62   184.1±0.15µs        ? ?/sec    1.00   1272.7±3.87ns        ? ?/sec
null_buffer_expand/len=512/count=12      1.00      2.3±0.01µs        ? ?/sec     1.01      2.3±0.00µs        ? ?/sec
null_buffer_expand/len=512/count=16      9.60      3.0±0.01µs        ? ?/sec     1.00    316.2±0.85ns        ? ?/sec

Resource Usage

base (merge-base)

Metric	Value
Wall time	60.0s
Peak memory	10.5 MiB
Avg memory	8.8 MiB
CPU user	55.8s
CPU sys	0.0s
Peak spill	0 B

branch

Metric	Value
Wall time	65.0s
Peak memory	10.5 MiB
Avg memory	8.2 MiB
CPU user	60.1s
CPU sys	0.0s
Peak spill	0 B

File an issue against this benchmark runner

Rich-T-kid · 2026-09-03T19:18:00Z

null_buffer_expand/len=1024/count=128    80.07    47.5±0.13µs        ? ?/sec     1.00    592.8±3.30ns        ? ?/sec
null_buffer_expand/len=1024/count=96     61.55    34.2±0.05µs        ? ?/sec     1.00    555.7±0.99ns        ? ?/sec
null_buffer_expand/len=2048/count=192    118.66   138.9±0.27µs        ? ?/sec    1.00   1170.7±6.19ns        ? ?/sec
null_buffer_expand/len=2048/count=256    144.62   184.1±0.15µs        ? ?/sec    1.00   1272.7±3.87ns        ? ?/sec
null_buffer_expand/len=512/count=12      1.00      2.3±0.01µs        ? ?/sec     1.01      2.3±0.00µs        ? ?/sec
null_buffer_expand/len=512/count=16      9.60      3.0±0.01µs        ? ?/sec     1.00    316.2±0.85ns        ? ?/sec

144x 😄

Rich-T-kid · 2026-09-03T19:28:45Z

run benchmark boolean_kernels
env:
BENCH_FILTER: null_buffer_expand

Rich-T-kid · 2026-09-03T19:28:55Z

run benchmark boolean_kernels
env:
BENCH_FILTER: null_buffer_expand

adriangbot · 2026-09-03T19:31:02Z

🤖 Arrow criterion benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5531008622-2123-hkwrl 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing rich-T-kid/optimize-NullBuff-expand (3395a25) to ec21627 (merge-base) diff

Run configuration

run benchmark boolean_kernels
env:
  BENCH_FILTER: "null_buffer_expand"

BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental,object_store --bench boolean_kernels
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-03T19:31:58Z

🤖 Arrow criterion benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5531010377-2124-f8hr8 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing rich-T-kid/optimize-NullBuff-expand (4272daa) to ec21627 (merge-base) diff

Run configuration

run benchmark boolean_kernels
env:
  BENCH_FILTER: "null_buffer_expand"

BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental,object_store --bench boolean_kernels
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-03T19:34:16Z

🤖 Arrow criterion benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-NullBuff-expand (3395a25) to ec21627 (merge-base) diff

Run configuration

run benchmark boolean_kernels
env:
  BENCH_FILTER: "null_buffer_expand"

CPU Details (lscpu)

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

Details

group                                    main                                    rich-T-kid_optimize-NullBuff-expand
-----                                    ----                                    -----------------------------------
null_buffer_expand/len=1024/count=128    80.16    47.4±0.09µs        ? ?/sec     1.00    591.4±3.68ns        ? ?/sec
null_buffer_expand/len=1024/count=96     61.26    34.2±0.03µs        ? ?/sec     1.00    558.2±1.83ns        ? ?/sec
null_buffer_expand/len=2048/count=192    118.37   138.8±0.17µs        ? ?/sec    1.00   1172.5±8.27ns        ? ?/sec
null_buffer_expand/len=2048/count=256    144.35   184.1±0.11µs        ? ?/sec    1.00   1275.2±6.43ns        ? ?/sec
null_buffer_expand/len=512/count=12      2.00      2.3±0.00µs        ? ?/sec     1.00   1141.7±1.45ns        ? ?/sec
null_buffer_expand/len=512/count=16      9.53      3.0±0.01µs        ? ?/sec     1.00    318.1±2.00ns        ? ?/sec

Resource Usage

base (merge-base)

Metric	Value
Wall time	60.0s
Peak memory	10.5 MiB
Avg memory	8.6 MiB
CPU user	54.8s
CPU sys	0.0s
Peak spill	0 B

branch

Metric	Value
Wall time	65.0s
Peak memory	10.5 MiB
Avg memory	8.1 MiB
CPU user	60.1s
CPU sys	0.0s
Peak spill	0 B

File an issue against this benchmark runner

adriangbot · 2026-09-03T19:35:13Z

🤖 Arrow criterion benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-NullBuff-expand (4272daa) to ec21627 (merge-base) diff

Run configuration

run benchmark boolean_kernels
env:
  BENCH_FILTER: "null_buffer_expand"

CPU Details (lscpu)

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

Details

group                                    main                                    rich-T-kid_optimize-NullBuff-expand
-----                                    ----                                    -----------------------------------
null_buffer_expand/len=1024/count=128    80.45    47.4±0.08µs        ? ?/sec     1.00    589.4±2.36ns        ? ?/sec
null_buffer_expand/len=1024/count=96     61.45    34.2±0.03µs        ? ?/sec     1.00    556.3±1.32ns        ? ?/sec
null_buffer_expand/len=2048/count=192    117.53   138.9±0.26µs        ? ?/sec    1.00   1181.9±5.35ns        ? ?/sec
null_buffer_expand/len=2048/count=256    143.13   184.0±0.15µs        ? ?/sec    1.00   1285.7±7.63ns        ? ?/sec
null_buffer_expand/len=512/count=12      2.00      2.3±0.01µs        ? ?/sec     1.00   1141.0±0.76ns        ? ?/sec
null_buffer_expand/len=512/count=16      9.57      3.0±0.00µs        ? ?/sec     1.00    317.2±1.84ns        ? ?/sec

Resource Usage

base (merge-base)

Metric	Value
Wall time	60.0s
Peak memory	10.5 MiB
Avg memory	8.6 MiB
CPU user	54.8s
CPU sys	0.0s
Peak spill	0 B

branch

Metric	Value
Wall time	65.0s
Peak memory	10.5 MiB
Avg memory	8.1 MiB
CPU user	60.1s
CPU sys	0.0s
Peak spill	0 B

File an issue against this benchmark runner

Rich-T-kid · 2026-09-03T19:44:13Z

run benchmark take_kernels
env:
BENCH_FILTER: fixed_size_list

adriangbot · 2026-09-03T19:46:58Z

🤖 Arrow criterion benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c5531187008-2125-lp7t5 6.12.94+ #1 SMP Fri Jul 17 09:42:57 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing rich-T-kid/optimize-NullBuff-expand (4272daa) to ec21627 (merge-base) diff

Run configuration

run benchmark take_kernels
env:
  BENCH_FILTER: "fixed_size_list"

BENCH_COMMAND=cargo bench --features=arrow,async,test_common,experimental,object_store --bench take_kernels
Results will be posted here when complete

File an issue against this benchmark runner

adriangbot · 2026-09-03T19:49:34Z

🤖 Arrow criterion benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-NullBuff-expand (4272daa) to ec21627 (merge-base) diff

Run configuration

run benchmark take_kernels
env:
  BENCH_FILTER: "fixed_size_list"

CPU Details (lscpu)

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

Details

group                                              main                                   rich-T-kid_optimize-NullBuff-expand
-----                                              ----                                   -----------------------------------
take fixed_size_list<i32>[22] 1024                 1.00      3.5±0.00µs        ? ?/sec    1.00      3.4±0.01µs        ? ?/sec
take fixed_size_list<i32>[22] null indices 1024    1.05     20.8±0.02µs        ? ?/sec    1.00     19.7±0.01µs        ? ?/sec
take fixed_size_list<i32>[22] null values 1024     1.05     21.7±0.01µs        ? ?/sec    1.00     20.7±0.01µs        ? ?/sec
take fixed_size_list<i32>[8] 1024                  1.01      2.8±0.01µs        ? ?/sec    1.00      2.8±0.00µs        ? ?/sec
take fixed_size_list<i32>[8] null indices 1024     2.32      9.4±0.01µs        ? ?/sec    1.00      4.1±0.00µs        ? ?/sec
take fixed_size_list<i32>[8] null values 1024      1.90     10.5±0.01µs        ? ?/sec    1.00      5.5±0.01µs        ? ?/sec

Resource Usage

base (merge-base)

Metric	Value
Wall time	65.0s
Peak memory	16.4 MiB
Avg memory	13.9 MiB
CPU user	61.8s
CPU sys	0.0s
Peak spill	0 B

branch

Metric	Value
Wall time	65.0s
Peak memory	16.4 MiB
Avg memory	13.8 MiB
CPU user	60.1s
CPU sys	0.0s
Peak spill	0 B

File an issue against this benchmark runner

Jefffrey · 2026-09-04T11:33:31Z

thanks for this @Rich-T-kid

good numbers on the benchmarks 🚀

Rich-T-kid added 2 commits September 3, 2026 14:57

perf(NullBuffer::exampnd); fill on 8 multiple

cedb7e4

linear iteration

704c4c9

github-actions Bot added arrow Changes to the arrow crate arrow-buffer labels Sep 3, 2026

optimize 4 branch case

3395a25

Rich-T-kid commented Sep 3, 2026

View reviewed changes

Comment thread arrow-buffer/src/buffer/null.rs Outdated

trim

4272daa

Rich-T-kid marked this pull request as ready for review September 3, 2026 19:31

fix clippy

287971f

Jefffrey added the performance label Sep 4, 2026

Jefffrey approved these changes Sep 4, 2026

View reviewed changes

Jefffrey merged commit 128d200 into apache:main Sep 4, 2026
41 checks passed

Rich-T-kid mentioned this pull request Sep 4, 2026

perf(nullbuf::expand) non aligned counts #10980

Open

Uh oh!

Conversation

Rich-T-kid commented Sep 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Which issue does this PR close?

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

Uh oh!

Rich-T-kid commented Sep 3, 2026

Uh oh!

adriangbot commented Sep 3, 2026

Uh oh!

adriangbot commented Sep 3, 2026

Uh oh!

Rich-T-kid commented Sep 3, 2026

Uh oh!

Rich-T-kid commented Sep 3, 2026

Uh oh!

Rich-T-kid commented Sep 3, 2026

Uh oh!

Uh oh!

adriangbot commented Sep 3, 2026

Uh oh!

adriangbot commented Sep 3, 2026

Uh oh!

adriangbot commented Sep 3, 2026

Uh oh!

adriangbot commented Sep 3, 2026

Uh oh!

Rich-T-kid commented Sep 3, 2026

Uh oh!

adriangbot commented Sep 3, 2026

Uh oh!

adriangbot commented Sep 3, 2026

Uh oh!

Uh oh!

Jefffrey commented Sep 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Rich-T-kid commented Sep 3, 2026 •

edited

Loading