Skip to content

Pipeline non-null dictionary hash scatter on AArch64 - #26849

Closed
Rich-T-kid wants to merge 1 commit into
apache:mainfrom
Rich-T-kid:rich-T-kid/optimize-dictionary-key-scatter
Closed

Rich-T-kid wants to merge 1 commit into
apache:mainfrom
Rich-T-kid:rich-T-kid/optimize-dictionary-key-scatter

Conversation

@Rich-T-kid

@Rich-T-kid Rich-T-kid commented Oct 28, 2026 •

Copy link
Copy Markdown
Contributor

Which issue does this PR close?

Related to #26780.

Rationale for this change

Scattering cached dictionary hashes performs an indexed load for each key. On AArch64, the scalar composite-hash loop does not expose enough independent loads to consistently hide that latency.

What changes are included in this PR?

For non-null composite dictionary hashes on AArch64, load eight dictionary hashes before combining them with the row hashes. This gives the CPU independent loads to overlap. The tail remains scalar.

Fresh hashes, nullable dictionaries, and other architectures keep the existing loop. There is no dictionary-size threshold or explicit prefetching.

Results

The upstream Neoverse-V2 run improved the existing multiple, no nulls benchmark from 10.8 us to 10.4 us. The single-column and nullable cases were unchanged.

The upstream bot also tested 4, 16, and 32 lanes. Four lanes measured 10.7 us; 16 and 32 lanes measured 10.4 us. Eight lanes is retained as the smallest best-performing width.

What is the testing strategy for this PR?

The hash utility tests pass in normal and force_hash_collisions configurations. Workspace formatting, Clippy, and typo checks pass.

Are there any user-facing changes?

No.

@github-actions github-actions Bot added the common Related to common crate label Oct 28, 2026
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes
env:
BENCH_FILTER: dict

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6470394997-4027-hkvhc 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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-dictionary-key-scatter (091664e) to af752e5 (merge-base) diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dict"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-dictionary-key-scatter (091664e) to af752e5 (merge-base) diff

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

group                                        HEAD                                   rich-T-kid_optimize-dictionary-key-scatter
-----                                        ----                                   ------------------------------------------
dictionary_utf8_int32: multiple, no nulls    1.00     10.8±0.01µs        ? ?/sec    1.00     10.8±0.01µs        ? ?/sec
dictionary_utf8_int32: multiple, nulls       1.94     26.5±0.03µs        ? ?/sec    1.00     13.6±0.01µs        ? ?/sec
dictionary_utf8_int32: single, no nulls      1.00      3.2±0.00µs        ? ?/sec    1.01      3.3±0.01µs        ? ?/sec
dictionary_utf8_int32: single, nulls         2.06      8.1±0.01µs        ? ?/sec    1.00      3.9±0.00µs        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 110.0s
Peak memory 33.3 MiB
Avg memory 9.4 MiB
CPU user 46.9s
CPU sys 0.0s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 105.0s
Peak memory 33.4 MiB
Avg memory 9.8 MiB
CPU user 47.6s
CPU sys 0.1s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes

env:
  BENCH_FILTER: dictionary_scatter
baseline:
  ref: "f37a77d8a"
changed:
  ref: "d8ff9f4c4"

@adriangbot

Copy link
Copy Markdown

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

Benchmarks requested: with_hashes

Runner log (last 40 lines)
2026-10-28T15:26:03.574586Z  INFO runner starting benchmark runner bench_type=Datafusion, pr_url=https://github.com/apache/datafusion/pull/26849, benchmarks=with_hashes
2026-10-28T15:26:03.626165Z  INFO benchmark_controller::runner::bench_datafusion === Cloning PR branch ===
2026-10-28T15:26:03.626188Z  INFO benchmark_controller::runner::shell running command cmd=git, args=["clone", "--depth=200", "https://github.com/apache/datafusion.git", "/workspace/datafusion-branch"], cwd="/"
2026-10-28T15:26:13.631906Z  INFO benchmark_controller::runner::shell running command cmd=git, args=["fetch", "origin", "refs/pull/26849/head:rich-T-kid/optimize-dictionary-key-scatter", "main"], cwd="/workspace/datafusion-branch"
2026-10-28T15:26:18.633271Z  INFO benchmark_controller::runner::shell running command cmd=git, args=["checkout", "rich-T-kid/optimize-dictionary-key-scatter"], cwd="/workspace/datafusion-branch"
2026-10-28T15:26:23.635526Z  INFO benchmark_controller::runner::shell running command cmd=git, args=["merge-base", "HEAD", "origin/main"], cwd="/workspace/datafusion-branch"
2026-10-28T15:26:28.638002Z  INFO benchmark_controller::runner::bench_datafusion === Checking out custom changed ref === changed_ref=d8ff9f4c4
2026-10-28T15:26:28.638018Z  INFO benchmark_controller::runner::shell running command cmd=git, args=["fetch", "origin", "refs/pull/26849/head"], cwd="/workspace/datafusion-branch"
2026-10-28T15:26:33.640239Z  INFO benchmark_controller::runner::shell running command cmd=git, args=["fetch", "origin"], cwd="/workspace/datafusion-branch"
2026-10-28T15:26:38.642931Z  INFO benchmark_controller::runner::shell running command cmd=git, args=["-c", "advice.detachedHead=false", "checkout", "d8ff9f4c4"], cwd="/workspace/datafusion-branch"
2026-10-28T15:26:43.644820Z ERROR runner benchmark failed error=git checkout: git -c advice.detachedHead=false checkout d8ff9f4c4 exited with code 1
stdout:

stderr:
error: pathspec 'd8ff9f4c4' did not match any file(s) known to git
Kubernetes message
Job has reached the specified backoff limit

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes

env:
  BENCH_FILTER: dictionary_scatter
baseline:
  ref: "f37a77d8a70d1c692c273c20bf7f75be325ac6a9"
changed:
  ref: "1194bbc7cac88dd49f8e74baa323bc042c282b82"

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6473192091-4049-rpm42 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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 1194bbc (1194bbc) to f37a77d diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dictionary_scatter"
baseline:
  ref: "f37a77d8a70d1c692c273c20bf7f75be325ac6a9"
changed:
  ref: "1194bbc7cac88dd49f8e74baa323bc042c282b82"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing 1194bbc (1194bbc) to f37a77d diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dictionary_scatter"
baseline:
  ref: "f37a77d8a70d1c692c273c20bf7f75be325ac6a9"
changed:
  ref: "1194bbc7cac88dd49f8e74baa323bc042c282b82"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

group                                               HEAD                                    rich-T-kid_optimize-dictionary-key-scatter
-----                                               ----                                    ------------------------------------------
dictionary_scatter: cardinality 1048576, combine    1.00  1938.9±129.73µs        ? ?/sec    1.04      2.0±0.06ms        ? ?/sec
dictionary_scatter: cardinality 1048576, fresh      1.02  1427.9±34.19µs        ? ?/sec     1.00  1396.4±35.47µs        ? ?/sec
dictionary_scatter: cardinality 128, combine        1.01   494.4±18.07µs        ? ?/sec     1.00   487.6±22.20µs        ? ?/sec
dictionary_scatter: cardinality 128, fresh          1.01    337.8±4.84µs        ? ?/sec     1.00    334.4±5.08µs        ? ?/sec
dictionary_scatter: cardinality 65536, combine      1.00   751.9±33.88µs        ? ?/sec     1.05   789.7±46.29µs        ? ?/sec
dictionary_scatter: cardinality 65536, fresh        1.02   687.4±18.12µs        ? ?/sec     1.00    676.9±7.23µs        ? ?/sec
dictionary_scatter: cardinality 8388608, combine    1.17      5.3±0.33ms        ? ?/sec     1.00      4.5±0.43ms        ? ?/sec
dictionary_scatter: cardinality 8388608, fresh      1.01      4.2±0.21ms        ? ?/sec     1.00      4.2±0.18ms        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 160.0s
Peak memory 117.1 MiB
Avg memory 35.1 MiB
CPU user 111.2s
CPU sys 0.1s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 150.0s
Peak memory 113.6 MiB
Avg memory 33.4 MiB
CPU user 107.1s
CPU sys 0.1s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid
Rich-T-kid force-pushed the rich-T-kid/optimize-dictionary-key-scatter branch from 1194bbc to db1b138 Compare October 28, 2026 15:43
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes
env:
BENCH_FILTER: dict

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6473557189-4060-6hlkc 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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-dictionary-key-scatter (db1b138) to af752e5 (merge-base) diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dict"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-dictionary-key-scatter (db1b138) to af752e5 (merge-base) diff

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

group                                        HEAD                                   rich-T-kid_optimize-dictionary-key-scatter
-----                                        ----                                   ------------------------------------------
dictionary_utf8_int32: multiple, no nulls    1.00     10.8±0.02µs        ? ?/sec    1.00     10.8±0.03µs        ? ?/sec
dictionary_utf8_int32: multiple, nulls       1.93     26.3±0.08µs        ? ?/sec    1.00     13.6±0.02µs        ? ?/sec
dictionary_utf8_int32: single, no nulls      1.00      3.2±0.01µs        ? ?/sec    1.01      3.3±0.02µs        ? ?/sec
dictionary_utf8_int32: single, nulls         2.08      8.2±0.00µs        ? ?/sec    1.00      3.9±0.00µs        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 80.0s
Peak memory 32.8 MiB
Avg memory 12.3 MiB
CPU user 48.4s
CPU sys 0.0s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 95.0s
Peak memory 32.6 MiB
Avg memory 10.3 MiB
CPU user 44.7s
CPU sys 0.0s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid
Rich-T-kid force-pushed the rich-T-kid/optimize-dictionary-key-scatter branch 2 times, most recently from 91256d1 to d5444aa Compare October 28, 2026 16:58
@Rich-T-kid Rich-T-kid changed the title Speed up nullable dictionary hash scattering Pipeline non-null dictionary hash scatter on AArch64 Oct 28, 2026
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes
env:
BENCH_FILTER: dict

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6475315520-4070-6v4mq 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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-dictionary-key-scatter (d5444aa) to af752e5 (merge-base) diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dict"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-dictionary-key-scatter (d5444aa) to af752e5 (merge-base) diff

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

group                                        HEAD                                   rich-T-kid_optimize-dictionary-key-scatter
-----                                        ----                                   ------------------------------------------
dictionary_utf8_int32: multiple, no nulls    1.03     10.8±0.01µs        ? ?/sec    1.00     10.5±0.04µs        ? ?/sec
dictionary_utf8_int32: multiple, nulls       1.01     26.5±0.08µs        ? ?/sec    1.00     26.3±0.14µs        ? ?/sec
dictionary_utf8_int32: single, no nulls      1.00      3.3±0.05µs        ? ?/sec    1.00      3.3±0.01µs        ? ?/sec
dictionary_utf8_int32: single, nulls         1.00      8.1±0.01µs        ? ?/sec    1.00      8.1±0.01µs        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 85.0s
Peak memory 36.0 MiB
Avg memory 12.7 MiB
CPU user 46.8s
CPU sys 0.0s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 95.0s
Peak memory 31.3 MiB
Avg memory 10.1 MiB
CPU user 44.5s
CPU sys 0.0s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid
Rich-T-kid force-pushed the rich-T-kid/optimize-dictionary-key-scatter branch from d5444aa to 427927c Compare October 28, 2026 17:18
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes
env:
BENCH_FILTER: dict

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6475745406-4071-xm8hp 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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-dictionary-key-scatter (427927c) to af752e5 (merge-base) diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dict"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-dictionary-key-scatter (427927c) to af752e5 (merge-base) diff

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

group                                        HEAD                                   rich-T-kid_optimize-dictionary-key-scatter
-----                                        ----                                   ------------------------------------------
dictionary_utf8_int32: multiple, no nulls    1.03     10.8±0.01µs        ? ?/sec    1.00     10.4±0.01µs        ? ?/sec
dictionary_utf8_int32: multiple, nulls       1.01     26.3±0.04µs        ? ?/sec    1.00     26.2±0.02µs        ? ?/sec
dictionary_utf8_int32: single, no nulls      1.00      3.3±0.00µs        ? ?/sec    1.00      3.3±0.00µs        ? ?/sec
dictionary_utf8_int32: single, nulls         1.01      8.2±0.00µs        ? ?/sec    1.00      8.1±0.00µs        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 85.0s
Peak memory 32.4 MiB
Avg memory 11.9 MiB
CPU user 45.4s
CPU sys 0.0s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 95.0s
Peak memory 32.5 MiB
Avg memory 10.5 MiB
CPU user 44.8s
CPU sys 0.0s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid
Rich-T-kid force-pushed the rich-T-kid/optimize-dictionary-key-scatter branch from 427927c to 78d4a14 Compare October 28, 2026 17:27
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes
env:
BENCH_FILTER: dict

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6475939500-4073-22lwx 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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-dictionary-key-scatter (78d4a14) to af752e5 (merge-base) diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dict"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-dictionary-key-scatter (78d4a14) to af752e5 (merge-base) diff

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

group                                        HEAD                                   rich-T-kid_optimize-dictionary-key-scatter
-----                                        ----                                   ------------------------------------------
dictionary_utf8_int32: multiple, no nulls    1.01     10.8±0.01µs        ? ?/sec    1.00     10.7±0.04µs        ? ?/sec
dictionary_utf8_int32: multiple, nulls       1.00     26.4±0.03µs        ? ?/sec    1.00     26.3±0.02µs        ? ?/sec
dictionary_utf8_int32: single, no nulls      1.00      3.2±0.01µs        ? ?/sec    1.01      3.3±0.00µs        ? ?/sec
dictionary_utf8_int32: single, nulls         1.00      8.2±0.01µs        ? ?/sec    1.02      8.3±0.00µs        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 80.0s
Peak memory 31.1 MiB
Avg memory 12.3 MiB
CPU user 44.9s
CPU sys 0.1s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 100.1s
Peak memory 34.9 MiB
Avg memory 10.3 MiB
CPU user 44.9s
CPU sys 0.0s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes
env:
BENCH_FILTER: dict

@Rich-T-kid
Rich-T-kid force-pushed the rich-T-kid/optimize-dictionary-key-scatter branch from 78d4a14 to 93838ec Compare October 28, 2026 17:34
@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6476090797-4074-49qhf 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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-dictionary-key-scatter (93838ec) to af752e5 (merge-base) diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dict"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-dictionary-key-scatter (93838ec) to af752e5 (merge-base) diff

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

group                                        HEAD                                   rich-T-kid_optimize-dictionary-key-scatter
-----                                        ----                                   ------------------------------------------
dictionary_utf8_int32: multiple, no nulls    1.04     10.8±0.01µs        ? ?/sec    1.00     10.4±0.01µs        ? ?/sec
dictionary_utf8_int32: multiple, nulls       1.03     26.5±0.02µs        ? ?/sec    1.00     25.6±0.01µs        ? ?/sec
dictionary_utf8_int32: single, no nulls      1.00      3.3±0.00µs        ? ?/sec    1.00      3.3±0.01µs        ? ?/sec
dictionary_utf8_int32: single, nulls         1.01      8.2±0.01µs        ? ?/sec    1.00      8.1±0.00µs        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 80.0s
Peak memory 32.8 MiB
Avg memory 12.6 MiB
CPU user 45.2s
CPU sys 0.0s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 95.0s
Peak memory 35.2 MiB
Avg memory 10.7 MiB
CPU user 44.7s
CPU sys 0.1s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes
env:
BENCH_FILTER: dict

@Rich-T-kid
Rich-T-kid force-pushed the rich-T-kid/optimize-dictionary-key-scatter branch from 93838ec to 785b7df Compare October 28, 2026 17:39
@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6476205147-4075-vgqzt 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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-dictionary-key-scatter (785b7df) to af752e5 (merge-base) diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dict"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-dictionary-key-scatter (785b7df) to af752e5 (merge-base) diff

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

group                                        HEAD                                   rich-T-kid_optimize-dictionary-key-scatter
-----                                        ----                                   ------------------------------------------
dictionary_utf8_int32: multiple, no nulls    1.04     10.8±0.01µs        ? ?/sec    1.00     10.4±0.01µs        ? ?/sec
dictionary_utf8_int32: multiple, nulls       1.03     26.4±0.07µs        ? ?/sec    1.00     25.8±0.02µs        ? ?/sec
dictionary_utf8_int32: single, no nulls      1.01      3.3±0.07µs        ? ?/sec    1.00      3.3±0.01µs        ? ?/sec
dictionary_utf8_int32: single, nulls         1.00      8.2±0.01µs        ? ?/sec    1.01      8.2±0.00µs        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 80.0s
Peak memory 36.9 MiB
Avg memory 13.9 MiB
CPU user 45.4s
CPU sys 0.1s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 95.0s
Peak memory 33.7 MiB
Avg memory 10.7 MiB
CPU user 44.8s
CPU sys 0.1s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid
Rich-T-kid force-pushed the rich-T-kid/optimize-dictionary-key-scatter branch 2 times, most recently from 00d3076 to a142b66 Compare October 28, 2026 17:51
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes
env:
BENCH_FILTER: dict

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6476481733-4076-gg489 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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-dictionary-key-scatter (a142b66) to af752e5 (merge-base) diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dict"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-dictionary-key-scatter (a142b66) to af752e5 (merge-base) diff

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

group                                        HEAD                                   rich-T-kid_optimize-dictionary-key-scatter
-----                                        ----                                   ------------------------------------------
dictionary_utf8_int32: multiple, no nulls    1.00     10.9±0.01µs        ? ?/sec    1.05     11.4±0.02µs        ? ?/sec
dictionary_utf8_int32: multiple, nulls       1.00     26.4±0.03µs        ? ?/sec    1.01     26.6±0.09µs        ? ?/sec
dictionary_utf8_int32: single, no nulls      1.00      3.3±0.00µs        ? ?/sec    1.00      3.2±0.01µs        ? ?/sec
dictionary_utf8_int32: single, nulls         1.00      8.1±0.01µs        ? ?/sec    1.02      8.3±0.02µs        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 80.0s
Peak memory 35.0 MiB
Avg memory 13.0 MiB
CPU user 45.1s
CPU sys 0.0s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 95.0s
Peak memory 34.2 MiB
Avg memory 10.7 MiB
CPU user 48.4s
CPU sys 0.0s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid
Rich-T-kid force-pushed the rich-T-kid/optimize-dictionary-key-scatter branch from a142b66 to fc0397a Compare October 28, 2026 18:20
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark with_hashes
env:
BENCH_FILTER: dict

@Rich-T-kid Rich-T-kid closed this Oct 28, 2026
@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6477086127-4081-22l6d 6.12.105+ #1 SMP Sat Sep 19 07:55:25 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-dictionary-key-scatter (fc0397a) to af752e5 (merge-base) diff

Run configuration
run benchmark with_hashes
env:
  BENCH_FILTER: "dict"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing rich-T-kid/optimize-dictionary-key-scatter (fc0397a) to af752e5 (merge-base) diff

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

group                                        HEAD                                   rich-T-kid_optimize-dictionary-key-scatter
-----                                        ----                                   ------------------------------------------
dictionary_utf8_int32: multiple, no nulls    1.00     10.8±0.01µs        ? ?/sec    1.03     11.1±0.02µs        ? ?/sec
dictionary_utf8_int32: multiple, nulls       1.04     26.5±0.03µs        ? ?/sec    1.00     25.6±0.03µs        ? ?/sec
dictionary_utf8_int32: single, no nulls      1.00      3.2±0.00µs        ? ?/sec    1.00      3.3±0.00µs        ? ?/sec
dictionary_utf8_int32: single, nulls         1.00      8.1±0.00µs        ? ?/sec    1.00      8.1±0.01µs        ? ?/sec

Resource Usage

with_hashes — base (merge-base)

Metric Value
Wall time 80.0s
Peak memory 36.7 MiB
Avg memory 13.3 MiB
CPU user 44.9s
CPU sys 0.0s
Peak spill 0 B

with_hashes — branch

Metric Value
Wall time 95.0s
Peak memory 32.4 MiB
Avg memory 10.5 MiB
CPU user 45.5s
CPU sys 0.1s
Peak spill 0 B

File an issue against this benchmark runner

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

common Related to common crate

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants