Skip to content

feat(parquet): cap dictionary reads using NDV estimates - #27500

Open
Rich-T-kid wants to merge 5 commits into
apache:mainfrom
Rich-T-kid:rich-T-kid/parquet-dictionary-ndv-threshold
Open

Rich-T-kid wants to merge 5 commits into
apache:mainfrom
Rich-T-kid:rich-T-kid/parquet-dictionary-ndv-threshold

Conversation

@Rich-T-kid

@Rich-T-kid Rich-T-kid commented Nov 17, 2026 •

Copy link
Copy Markdown
Contributor

Which issue does this PR close?

Rationale for this change

Low-cardinality Parquet string and binary columns can be cheaper to decode and process as dictionaries. Large row-group dictionaries can instead be slower because each output batch can retain the whole dictionary.

For a multi-file table, keeping a dictionary table type when any file remains plain would require dictionary-encoding that file's values during reading or coercion. That adds conversion and memory overhead, so the conservative choice is to keep the field plain when files disagree.

What changes are included in this PR?

Adds datafusion.execution.parquet.dictionary_read_max_distinct_values, a cap for promoting inferred Parquet string and binary columns to dictionaries.

Promotion behavior

This option applies when DataFusion infers the table schema. A user-supplied plain schema is not promoted.

  • NDV available: For each file, DataFusion sums the row-group NDVs. A top-level plain string or binary field is promoted when every nonempty row group has a dictionary page and the sum is at most cap.
  • NDV missing: The row group's non-null row count is used when null_count is available; otherwise its total row count is used. Missing NDV therefore does not automatically prevent promotion.
  • Dictionary page missing: If any nonempty row group lacks a dictionary page, that file's field remains plain regardless of its cardinality estimate.
  • Multiple files: Each file is evaluated independently. If a field is dictionary in some participating files and plain in others, the merged table schema uses the plain type. Promoting the plain files solely to make the representation uniform would add dictionary-encoding overhead during reads, so the conservative choice is to keep the field plain when files disagree.
  • Declared dictionaries: The cap does not demote a dictionary declared by a file's embedded Arrow schema. It can still become plain when reconciling it with another file that declares the same field as plain.
  • Legacy option: Setting dictionary_read_max_distinct_values enables dictionary reads by itself. When both options are set, the cap controls promotion. Without a cap, enable_rle_to_dictionary = true retains the existing unbounded behavior.

Row-group estimates are deliberately conservative: values repeated across row groups may be counted more than once.

What is the testing strategy for this PR?

SQL logic tests cover cap-only activation, exact and above-cap estimates, the dictionary-page requirement, multi-file reconciliation, declared dictionaries, and precedence over the legacy flag. Focused metadata coverage checks the non-null-row fallback for missing NDV. Serialization coverage checks that the new option round trips. Formatting, workspace Clippy, and typo checks pass.

Benchmark evidence

Hosted benchmarks on final head f323f0265 compare the existing unbounded behavior (enable_rle_to_dictionary = true with no cap) against the same flag with dictionary_read_max_distinct_values set to 20,000 or 40,000. Both sides use the same commit, so the cap is the only changed setting. Each comparison ran TPC-H, TPC-DS, and partitioned ClickBench on arm64 and amd64 with 12 CPUs and 65 GiB.

Negative values mean the capped run was faster.

Cap Architecture TPC-H TPC-DS ClickBench
20,000 arm64 -6.65% -4.20% -50.71%
20,000 amd64 -5.04% -1.06% -53.69%
40,000 arm64 -5.91% +1.60% -50.64%
40,000 amd64 -4.64% -2.07% -53.27%

All 12 suite jobs completed successfully. The 20,000 cap improved all six suite totals. The 40,000 cap improved five of six; arm64 TPC-DS was 1.60% slower. Both caps reduced partitioned ClickBench total time by about 51% on arm64 and 53% on amd64 compared with unbounded promotion.

Three per-query median regressions reproduced with both caps on both architectures, and the distribution comparison also classified each as slower in all four runs: TPC-H Q19 (9–13%), TPC-DS Q15 (6–9%), and ClickBench Q28 (7–18%). The crossover remains workload-dependent, so the option documentation recommends a conservative cap and representative benchmarking.

Are there any user-facing changes?

Yes. ParquetOptions gains dictionary_read_max_distinct_values: Option<usize>. Exhaustive struct literals must add the field or use ..Default::default(). The default is None, so existing behavior is unchanged.

@github-actions github-actions Bot added documentation Improvements or additions to documentation sqllogictest SQL Logic Tests (.slt) common Related to common crate proto Related to proto crate datasource Changes to the datasource crate labels Nov 17, 2026
@Rich-T-kid
Rich-T-kid marked this pull request as draft November 17, 2026 04:53
@Rich-T-kid
Rich-T-kid force-pushed the rich-T-kid/parquet-dictionary-ndv-threshold branch 2 times, most recently from 85dbfb4 to 432586c Compare November 17, 2026 05:20
@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark tpch tpcds clickbench_partitioned

env:
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: arm64

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark tpch tpcds clickbench_partitioned

env:
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: amd64

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark tpch tpcds clickbench_partitioned

env:
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "false"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: arm64

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark tpch tpcds clickbench_partitioned

env:
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "false"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939383888-5261-hg9lb 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939383888-5262-rd98d 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939383888-5263-5p2x8 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939383514-5264-pc4fw 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 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939383514-5265-zcq4c 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 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

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 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃     HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 39.77 ms │                                    40.97 ms │     no change │
│ QQuery 2  │ 19.55 ms │                                    18.67 ms │     no change │
│ QQuery 3  │ 29.25 ms │                                    28.93 ms │     no change │
│ QQuery 4  │ 17.81 ms │                                    17.86 ms │     no change │
│ QQuery 5  │ 36.27 ms │                                    36.22 ms │     no change │
│ QQuery 6  │ 13.99 ms │                                    14.00 ms │     no change │
│ QQuery 7  │ 42.24 ms │                                    42.94 ms │     no change │
│ QQuery 8  │ 40.57 ms │                                    40.60 ms │     no change │
│ QQuery 9  │ 50.95 ms │                                    48.83 ms │     no change │
│ QQuery 10 │ 58.54 ms │                                    41.51 ms │ +1.41x faster │
│ QQuery 11 │ 14.31 ms │                                    13.97 ms │     no change │
│ QQuery 12 │ 20.68 ms │                                    17.99 ms │ +1.15x faster │
│ QQuery 13 │ 48.68 ms │                                    41.95 ms │ +1.16x faster │
│ QQuery 14 │ 24.38 ms │                                    24.69 ms │     no change │
│ QQuery 15 │ 31.69 ms │                                    31.62 ms │     no change │
│ QQuery 16 │ 14.08 ms │                                    13.76 ms │     no change │
│ QQuery 17 │ 71.27 ms │                                    71.89 ms │     no change │
│ QQuery 18 │ 64.84 ms │                                    40.38 ms │ +1.61x faster │
│ QQuery 19 │ 29.67 ms │                                    32.30 ms │  1.09x slower │
│ QQuery 20 │ 34.92 ms │                                    32.53 ms │ +1.07x faster │
│ QQuery 21 │ 54.31 ms │                                    57.62 ms │  1.06x slower │
│ QQuery 22 │ 16.34 ms │                                    14.27 ms │ +1.14x faster │
└───────────┴──────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 774.09ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 723.52ms │
│ Average Time (HEAD)                                        │  35.19ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │  32.89ms │
│ Queries Faster                                             │        6 │
│ Queries Slower                                             │        2 │
│ Queries with No Change                                     │       14 │
│ Queries with Failure                                       │        0 │
└────────────────────────────────────────────────────────────┴──────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 39.77 / 40.79 ±0.98 / 42.24 ms │              40.97 / 42.36 ±1.20 / 43.93 ms │     no change │
│ QQuery 2  │ 19.55 / 20.21 ±0.73 / 21.62 ms │              18.67 / 19.01 ±0.20 / 19.25 ms │ +1.06x faster │
│ QQuery 3  │ 29.25 / 29.54 ±0.36 / 30.21 ms │              28.93 / 29.11 ±0.20 / 29.47 ms │     no change │
│ QQuery 4  │ 17.81 / 18.31 ±0.71 / 19.69 ms │              17.86 / 17.95 ±0.07 / 18.06 ms │     no change │
│ QQuery 5  │ 36.27 / 36.66 ±0.30 / 37.17 ms │              36.22 / 36.46 ±0.24 / 36.87 ms │     no change │
│ QQuery 6  │ 13.99 / 14.05 ±0.06 / 14.16 ms │              14.00 / 14.12 ±0.07 / 14.21 ms │     no change │
│ QQuery 7  │ 42.24 / 43.88 ±1.04 / 45.01 ms │              42.94 / 43.88 ±0.83 / 45.36 ms │     no change │
│ QQuery 8  │ 40.57 / 41.50 ±1.04 / 43.54 ms │              40.60 / 41.31 ±0.52 / 42.20 ms │     no change │
│ QQuery 9  │ 50.95 / 52.55 ±1.11 / 53.85 ms │              48.83 / 51.06 ±1.45 / 53.00 ms │     no change │
│ QQuery 10 │ 58.54 / 59.49 ±0.98 / 61.38 ms │              41.51 / 41.98 ±0.62 / 43.20 ms │ +1.42x faster │
│ QQuery 11 │ 14.31 / 14.95 ±0.75 / 16.12 ms │              13.97 / 14.13 ±0.19 / 14.45 ms │ +1.06x faster │
│ QQuery 12 │ 20.68 / 20.92 ±0.17 / 21.11 ms │              17.99 / 18.39 ±0.38 / 19.09 ms │ +1.14x faster │
│ QQuery 13 │ 48.68 / 50.13 ±1.94 / 53.98 ms │              41.95 / 43.76 ±3.12 / 49.98 ms │ +1.15x faster │
│ QQuery 14 │ 24.38 / 24.54 ±0.13 / 24.71 ms │              24.69 / 24.99 ±0.17 / 25.16 ms │     no change │
│ QQuery 15 │ 31.69 / 32.15 ±0.40 / 32.87 ms │              31.62 / 31.88 ±0.25 / 32.32 ms │     no change │
│ QQuery 16 │ 14.08 / 14.16 ±0.11 / 14.38 ms │              13.76 / 14.12 ±0.43 / 14.97 ms │     no change │
│ QQuery 17 │ 71.27 / 72.57 ±1.96 / 76.40 ms │              71.89 / 72.53 ±0.53 / 73.46 ms │     no change │
│ QQuery 18 │ 64.84 / 67.71 ±1.85 / 69.77 ms │              40.38 / 42.55 ±1.79 / 45.79 ms │ +1.59x faster │
│ QQuery 19 │ 29.67 / 30.10 ±0.45 / 30.93 ms │              32.30 / 32.82 ±0.58 / 33.94 ms │  1.09x slower │
│ QQuery 20 │ 34.92 / 35.12 ±0.17 / 35.40 ms │              32.53 / 32.71 ±0.18 / 33.05 ms │ +1.07x faster │
│ QQuery 21 │ 54.31 / 56.51 ±1.70 / 59.46 ms │              57.62 / 58.58 ±0.72 / 59.77 ms │     no change │
│ QQuery 22 │ 16.34 / 16.57 ±0.19 / 16.85 ms │              14.27 / 14.34 ±0.05 / 14.42 ms │ +1.16x faster │
└───────────┴────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 792.42ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 738.03ms │
│ Average Time (HEAD)                                        │  36.02ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │  33.55ms │
│ Queries Faster                                             │        8 │
│ Queries Slower                                             │        1 │
│ Queries with No Change                                     │       13 │
│ Queries with Failure                                       │        0 │
└────────────────────────────────────────────────────────────┴──────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca | Changed: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca

tpch — tpch_sf1

Query Base Changed Change
Query 1 20.1 MiB 30.0 MiB +49.1%
Query 2 23.1 MiB 15.9 MiB -31.3%
Query 3 10.6 MiB 10.6 MiB +0.0%
Query 4 32.2 MiB 43.1 MiB +34.0%
Query 5 42.7 MiB 42.9 MiB +0.5%
Query 6 832 B 832 B +0.0%
Query 7 104.4 MiB 106.7 MiB +2.2%
Query 8 39.8 MiB 39.8 MiB -0.0%
Query 9 141.6 MiB 141.6 MiB -0.0%
Query 10 53.3 MiB 38.6 MiB -27.6%
Query 11 51.7 MiB 51.9 MiB +0.4%
Query 12 22.5 MiB 22.7 MiB +0.9%
Query 13 62.5 MiB 62.3 MiB -0.2%
Query 14 6.7 MiB 6.7 MiB +0.0%
Query 15 12.8 MiB 12.2 MiB -4.8%
Query 16 33.9 MiB 35.2 MiB +3.9%
Query 17 150.3 MiB 150.3 MiB +0.0%
Query 18 106.0 MiB 167.1 MiB +57.7%
Query 19 209.7 KiB 725.7 KiB +246.1%
Query 20 61.6 MiB 61.3 MiB -0.5%
Query 21 35.9 MiB 42.2 MiB +17.7%
Query 22 41.5 MiB 32.4 MiB -22.0%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpch base (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 150.3 MiB 1.1 GiB 1010.8 MiB 7.7×
tpch changed (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 167.1 MiB 1.1 GiB 970.8 MiB 6.8×
Resource Usage

tpch — base (merge-base)

Metric Value
Wall time 5.0s
Peak memory 1.1 GiB
Avg memory 488.4 MiB
CPU user 20.6s
CPU sys 1.7s
Peak spill 0 B

tpch — branch

Metric Value
Wall time 5.0s
Peak memory 1.1 GiB
Avg memory 473.2 MiB
CPU user 19.9s
CPU sys 1.9s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939383514-5266-gn8f5 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 432586c (432586c) to 432586c diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

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 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │    5.74 ms │                                     5.77 ms │     no change │
│ QQuery 2  │   77.11 ms │                                    79.47 ms │     no change │
│ QQuery 3  │   29.89 ms │                                    29.52 ms │     no change │
│ QQuery 4  │  555.67 ms │                                   486.45 ms │ +1.14x faster │
│ QQuery 5  │   51.44 ms │                                    51.09 ms │     no change │
│ QQuery 6  │   34.12 ms │                                    36.12 ms │  1.06x slower │
│ QQuery 7  │   72.62 ms │                                    72.80 ms │     no change │
│ QQuery 8  │   39.64 ms │                                    37.20 ms │ +1.07x faster │
│ QQuery 9  │   53.87 ms │                                    51.65 ms │     no change │
│ QQuery 10 │   59.93 ms │                                    60.85 ms │     no change │
│ QQuery 11 │  397.05 ms │                                   301.25 ms │ +1.32x faster │
│ QQuery 12 │   27.47 ms │                                    29.43 ms │  1.07x slower │
│ QQuery 13 │  115.32 ms │                                   114.77 ms │     no change │
│ QQuery 14 │  384.88 ms │                                   381.06 ms │     no change │
│ QQuery 15 │   53.20 ms │                                    56.81 ms │  1.07x slower │
│ QQuery 16 │    6.39 ms │                                     6.27 ms │     no change │
│ QQuery 17 │   80.78 ms │                                    76.51 ms │ +1.06x faster │
│ QQuery 18 │  108.17 ms │                                   107.74 ms │     no change │
│ QQuery 19 │   42.74 ms │                                    41.55 ms │     no change │
│ QQuery 20 │   32.63 ms │                                    36.36 ms │  1.11x slower │
│ QQuery 21 │   17.71 ms │                                    17.22 ms │     no change │
│ QQuery 22 │   63.85 ms │                                    65.70 ms │     no change │
│ QQuery 23 │  314.27 ms │                                   301.91 ms │     no change │
│ QQuery 24 │  208.07 ms │                                   201.14 ms │     no change │
│ QQuery 25 │  106.31 ms │                                   105.27 ms │     no change │
│ QQuery 26 │   45.94 ms │                                    46.06 ms │     no change │
│ QQuery 27 │    5.71 ms │                                     5.15 ms │ +1.11x faster │
│ QQuery 28 │   61.24 ms │                                    61.52 ms │     no change │
│ QQuery 29 │   95.15 ms │                                    91.91 ms │     no change │
│ QQuery 30 │   39.71 ms │                                    33.07 ms │ +1.20x faster │
│ QQuery 31 │  113.18 ms │                                   109.07 ms │     no change │
│ QQuery 32 │   20.78 ms │                                    20.49 ms │     no change │
│ QQuery 33 │   38.62 ms │                                    36.91 ms │     no change │
│ QQuery 34 │    6.08 ms │                                     5.55 ms │ +1.10x faster │
│ QQuery 35 │   74.96 ms │                                    71.97 ms │     no change │
│ QQuery 36 │    6.65 ms │                                     5.40 ms │ +1.23x faster │
│ QQuery 37 │    8.31 ms │                                     6.80 ms │ +1.22x faster │
│ QQuery 38 │   68.35 ms │                                    60.98 ms │ +1.12x faster │
│ QQuery 39 │   76.12 ms │                                    74.03 ms │     no change │
│ QQuery 40 │   24.44 ms │                                    22.57 ms │ +1.08x faster │
│ QQuery 41 │   14.86 ms │                                    13.25 ms │ +1.12x faster │
│ QQuery 42 │   25.13 ms │                                    23.95 ms │     no change │
│ QQuery 43 │    5.72 ms │                                     4.79 ms │ +1.19x faster │
│ QQuery 44 │    6.48 ms │                                     5.72 ms │ +1.13x faster │
│ QQuery 45 │   41.24 ms │                                    38.02 ms │ +1.08x faster │
│ QQuery 46 │   12.28 ms │                                    12.25 ms │     no change │
│ QQuery 47 │  216.04 ms │                                   235.74 ms │  1.09x slower │
│ QQuery 48 │   92.47 ms │                                    91.88 ms │     no change │
│ QQuery 49 │   70.58 ms │                                    70.18 ms │     no change │
│ QQuery 50 │   59.41 ms │                                    56.81 ms │     no change │
│ QQuery 51 │   95.43 ms │                                    92.91 ms │     no change │
│ QQuery 52 │   25.13 ms │                                    24.61 ms │     no change │
│ QQuery 53 │   30.66 ms │                                    29.84 ms │     no change │
│ QQuery 54 │   55.00 ms │                                    54.03 ms │     no change │
│ QQuery 55 │   24.75 ms │                                    23.63 ms │     no change │
│ QQuery 56 │   39.61 ms │                                    38.52 ms │     no change │
│ QQuery 57 │  156.25 ms │                                   170.46 ms │  1.09x slower │
│ QQuery 58 │  105.08 ms │                                   106.89 ms │     no change │
│ QQuery 59 │  107.23 ms │                                   114.29 ms │  1.07x slower │
│ QQuery 60 │   40.16 ms │                                    38.75 ms │     no change │
│ QQuery 61 │    9.80 ms │                                     8.90 ms │ +1.10x faster │
│ QQuery 62 │   38.92 ms │                                    43.74 ms │  1.12x slower │
│ QQuery 63 │   30.32 ms │                                    29.53 ms │     no change │
│ QQuery 64 │  370.49 ms │                                   355.80 ms │     no change │
│ QQuery 65 │   75.41 ms │                                    68.53 ms │ +1.10x faster │
│ QQuery 66 │   75.66 ms │                                    74.10 ms │     no change │
│ QQuery 67 │  201.74 ms │                                   238.12 ms │  1.18x slower │
│ QQuery 68 │    7.00 ms │                                     6.66 ms │     no change │
│ QQuery 69 │   54.67 ms │                                    56.47 ms │     no change │
│ QQuery 70 │   98.73 ms │                                   102.02 ms │     no change │
│ QQuery 71 │   35.62 ms │                                    35.54 ms │     no change │
│ QQuery 72 │ 1602.47 ms │                                  1708.83 ms │  1.07x slower │
│ QQuery 73 │    5.91 ms │                                     5.43 ms │ +1.09x faster │
│ QQuery 74 │  252.61 ms │                                   206.36 ms │ +1.22x faster │
│ QQuery 75 │  142.47 ms │                                   138.74 ms │     no change │
│ QQuery 76 │   35.44 ms │                                    35.00 ms │     no change │
│ QQuery 77 │   61.25 ms │                                    60.15 ms │     no change │
│ QQuery 78 │  274.10 ms │                                   274.36 ms │     no change │
│ QQuery 79 │   69.18 ms │                                    66.32 ms │     no change │
│ QQuery 80 │   96.36 ms │                                    96.58 ms │     no change │
│ QQuery 81 │   29.50 ms │                                    26.39 ms │ +1.12x faster │
│ QQuery 82 │   18.12 ms │                                    17.02 ms │ +1.06x faster │
│ QQuery 83 │   33.47 ms │                                    33.20 ms │     no change │
│ QQuery 84 │   31.07 ms │                                    29.55 ms │     no change │
│ QQuery 85 │   96.27 ms │                                    97.57 ms │     no change │
│ QQuery 86 │   24.72 ms │                                    25.75 ms │     no change │
│ QQuery 87 │   66.04 ms │                                    63.80 ms │     no change │
│ QQuery 88 │   60.32 ms │                                    60.57 ms │     no change │
│ QQuery 89 │   35.77 ms │                                    36.04 ms │     no change │
│ QQuery 90 │   16.47 ms │                                    16.69 ms │     no change │
│ QQuery 91 │   39.94 ms │                                    41.14 ms │     no change │
│ QQuery 92 │   29.64 ms │                                    30.59 ms │     no change │
│ QQuery 93 │   49.77 ms │                                    49.85 ms │     no change │
│ QQuery 94 │   38.76 ms │                                    38.80 ms │     no change │
│ QQuery 95 │   80.34 ms │                                    79.64 ms │     no change │
│ QQuery 96 │   24.27 ms │                                    23.99 ms │     no change │
│ QQuery 97 │   52.85 ms │                                    52.89 ms │     no change │
│ QQuery 98 │   38.61 ms │                                    44.57 ms │  1.15x slower │
│ QQuery 99 │   57.34 ms │                                    64.94 ms │  1.13x slower │
└───────────┴────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 9209.00ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 9096.08ms │
│ Average Time (HEAD)                                        │   93.02ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   91.88ms │
│ Queries Faster                                             │        21 │
│ Queries Slower                                             │        12 │
│ Queries with No Change                                     │        66 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │           5.74 / 6.32 ±1.00 / 8.32 ms │                 5.77 / 6.27 ±0.85 / 7.96 ms │     no change │
│ QQuery 2  │        77.11 / 77.45 ±0.24 / 77.73 ms │              79.47 / 79.98 ±0.38 / 80.54 ms │     no change │
│ QQuery 3  │        29.89 / 30.15 ±0.20 / 30.41 ms │              29.52 / 29.74 ±0.23 / 30.17 ms │     no change │
│ QQuery 4  │    555.67 / 571.54 ±14.64 / 597.85 ms │           486.45 / 490.97 ±5.04 / 500.40 ms │ +1.16x faster │
│ QQuery 5  │        51.44 / 51.75 ±0.27 / 52.06 ms │              51.09 / 51.77 ±0.49 / 52.38 ms │     no change │
│ QQuery 6  │        34.12 / 35.21 ±0.69 / 36.10 ms │              36.12 / 36.79 ±0.49 / 37.46 ms │     no change │
│ QQuery 7  │        72.62 / 75.85 ±4.29 / 84.34 ms │              72.80 / 73.73 ±0.59 / 74.32 ms │     no change │
│ QQuery 8  │        39.64 / 39.91 ±0.22 / 40.20 ms │              37.20 / 39.16 ±3.27 / 45.68 ms │     no change │
│ QQuery 9  │        53.87 / 55.07 ±0.71 / 55.90 ms │              51.65 / 54.41 ±2.03 / 57.63 ms │     no change │
│ QQuery 10 │        59.93 / 61.20 ±2.17 / 65.53 ms │              60.85 / 61.13 ±0.22 / 61.44 ms │     no change │
│ QQuery 11 │     397.05 / 403.77 ±5.77 / 412.22 ms │           301.25 / 307.48 ±3.66 / 312.37 ms │ +1.31x faster │
│ QQuery 12 │        27.47 / 27.80 ±0.23 / 28.05 ms │              29.43 / 29.94 ±0.32 / 30.30 ms │  1.08x slower │
│ QQuery 13 │     115.32 / 117.31 ±1.14 / 118.69 ms │           114.77 / 115.54 ±0.62 / 116.56 ms │     no change │
│ QQuery 14 │     384.88 / 390.43 ±3.23 / 393.74 ms │           381.06 / 386.78 ±3.76 / 392.68 ms │     no change │
│ QQuery 15 │        53.20 / 53.84 ±0.39 / 54.38 ms │              56.81 / 59.41 ±1.82 / 61.62 ms │  1.10x slower │
│ QQuery 16 │           6.39 / 6.52 ±0.09 / 6.68 ms │                 6.27 / 6.38 ±0.14 / 6.65 ms │     no change │
│ QQuery 17 │        80.78 / 85.04 ±5.33 / 95.36 ms │              76.51 / 78.17 ±0.93 / 79.38 ms │ +1.09x faster │
│ QQuery 18 │     108.17 / 110.21 ±1.74 / 112.92 ms │           107.74 / 108.59 ±0.67 / 109.40 ms │     no change │
│ QQuery 19 │        42.74 / 43.73 ±0.86 / 45.27 ms │              41.55 / 41.76 ±0.29 / 42.32 ms │     no change │
│ QQuery 20 │        32.63 / 33.36 ±0.59 / 34.40 ms │              36.36 / 37.06 ±0.45 / 37.57 ms │  1.11x slower │
│ QQuery 21 │        17.71 / 18.07 ±0.23 / 18.31 ms │              17.22 / 17.49 ±0.19 / 17.80 ms │     no change │
│ QQuery 22 │        63.85 / 64.21 ±0.22 / 64.44 ms │              65.70 / 68.02 ±2.89 / 73.58 ms │  1.06x slower │
│ QQuery 23 │     314.27 / 323.33 ±5.04 / 329.00 ms │           301.91 / 307.61 ±2.95 / 309.78 ms │     no change │
│ QQuery 24 │     208.07 / 213.21 ±2.65 / 215.49 ms │           201.14 / 203.05 ±1.22 / 204.33 ms │     no change │
│ QQuery 25 │     106.31 / 108.85 ±2.82 / 112.99 ms │           105.27 / 108.28 ±2.81 / 113.54 ms │     no change │
│ QQuery 26 │        45.94 / 47.66 ±1.86 / 51.29 ms │              46.06 / 46.48 ±0.30 / 46.77 ms │     no change │
│ QQuery 27 │           5.71 / 5.81 ±0.10 / 5.98 ms │                 5.15 / 5.30 ±0.11 / 5.50 ms │ +1.10x faster │
│ QQuery 28 │        61.24 / 61.68 ±0.26 / 61.93 ms │              61.52 / 62.05 ±0.38 / 62.63 ms │     no change │
│ QQuery 29 │       95.15 / 97.45 ±2.60 / 102.17 ms │              91.91 / 93.79 ±1.55 / 96.18 ms │     no change │
│ QQuery 30 │        39.71 / 41.12 ±1.78 / 44.54 ms │              33.07 / 34.05 ±0.90 / 35.73 ms │ +1.21x faster │
│ QQuery 31 │     113.18 / 114.04 ±0.57 / 114.69 ms │           109.07 / 110.59 ±2.52 / 115.63 ms │     no change │
│ QQuery 32 │        20.78 / 21.35 ±0.31 / 21.62 ms │              20.49 / 20.63 ±0.15 / 20.91 ms │     no change │
│ QQuery 33 │        38.62 / 39.00 ±0.41 / 39.77 ms │              36.91 / 37.42 ±0.53 / 38.30 ms │     no change │
│ QQuery 34 │          6.08 / 7.20 ±2.07 / 11.35 ms │                 5.55 / 6.13 ±0.86 / 7.82 ms │ +1.17x faster │
│ QQuery 35 │        74.96 / 76.45 ±0.88 / 77.52 ms │              71.97 / 73.12 ±0.78 / 74.27 ms │     no change │
│ QQuery 36 │           6.65 / 6.77 ±0.08 / 6.91 ms │                 5.40 / 5.55 ±0.23 / 6.01 ms │ +1.22x faster │
│ QQuery 37 │           8.31 / 8.52 ±0.13 / 8.65 ms │                 6.80 / 6.97 ±0.09 / 7.08 ms │ +1.22x faster │
│ QQuery 38 │        68.35 / 69.19 ±0.60 / 70.16 ms │              60.98 / 62.41 ±1.08 / 63.78 ms │ +1.11x faster │
│ QQuery 39 │        76.12 / 79.12 ±2.05 / 81.60 ms │              74.03 / 74.97 ±0.60 / 75.86 ms │ +1.06x faster │
│ QQuery 40 │        24.44 / 24.81 ±0.35 / 25.40 ms │              22.57 / 23.29 ±0.48 / 24.00 ms │ +1.07x faster │
│ QQuery 41 │        14.86 / 15.15 ±0.17 / 15.37 ms │              13.25 / 13.33 ±0.06 / 13.42 ms │ +1.14x faster │
│ QQuery 42 │        25.13 / 25.59 ±0.39 / 26.28 ms │              23.95 / 24.39 ±0.25 / 24.67 ms │     no change │
│ QQuery 43 │           5.72 / 5.87 ±0.09 / 5.98 ms │                 4.79 / 4.91 ±0.07 / 4.98 ms │ +1.19x faster │
│ QQuery 44 │           6.48 / 6.58 ±0.13 / 6.83 ms │                 5.72 / 5.82 ±0.10 / 5.99 ms │ +1.13x faster │
│ QQuery 45 │        41.24 / 42.51 ±0.74 / 43.55 ms │              38.02 / 39.91 ±2.65 / 45.10 ms │ +1.07x faster │
│ QQuery 46 │        12.28 / 12.57 ±0.31 / 13.14 ms │              12.25 / 12.77 ±0.39 / 13.46 ms │     no change │
│ QQuery 47 │     216.04 / 229.50 ±7.97 / 237.31 ms │           235.74 / 240.35 ±3.59 / 244.85 ms │     no change │
│ QQuery 48 │        92.47 / 93.60 ±0.86 / 94.95 ms │              91.88 / 93.06 ±1.36 / 95.56 ms │     no change │
│ QQuery 49 │        70.58 / 71.28 ±0.49 / 72.02 ms │              70.18 / 71.16 ±0.57 / 71.85 ms │     no change │
│ QQuery 50 │        59.41 / 59.85 ±0.43 / 60.61 ms │              56.81 / 58.15 ±1.09 / 60.15 ms │     no change │
│ QQuery 51 │       95.43 / 97.09 ±1.57 / 100.07 ms │              92.91 / 95.93 ±2.08 / 99.30 ms │     no change │
│ QQuery 52 │        25.13 / 25.48 ±0.37 / 26.16 ms │              24.61 / 24.71 ±0.10 / 24.84 ms │     no change │
│ QQuery 53 │        30.66 / 30.84 ±0.12 / 31.02 ms │              29.84 / 31.31 ±2.19 / 35.67 ms │     no change │
│ QQuery 54 │        55.00 / 56.23 ±0.95 / 57.74 ms │              54.03 / 54.99 ±1.08 / 56.89 ms │     no change │
│ QQuery 55 │        24.75 / 25.07 ±0.35 / 25.71 ms │              23.63 / 23.89 ±0.24 / 24.32 ms │     no change │
│ QQuery 56 │        39.61 / 39.97 ±0.26 / 40.27 ms │              38.52 / 39.38 ±0.81 / 40.67 ms │     no change │
│ QQuery 57 │     156.25 / 158.99 ±2.84 / 163.50 ms │           170.46 / 173.69 ±1.98 / 176.43 ms │  1.09x slower │
│ QQuery 58 │     105.08 / 106.34 ±0.85 / 107.33 ms │           106.89 / 109.28 ±2.20 / 113.38 ms │     no change │
│ QQuery 59 │     107.23 / 109.28 ±1.27 / 110.74 ms │           114.29 / 115.17 ±0.78 / 116.59 ms │  1.05x slower │
│ QQuery 60 │        40.16 / 40.74 ±0.47 / 41.24 ms │              38.75 / 38.97 ±0.20 / 39.28 ms │     no change │
│ QQuery 61 │          9.80 / 9.95 ±0.12 / 10.10 ms │                 8.90 / 9.16 ±0.23 / 9.53 ms │ +1.09x faster │
│ QQuery 62 │        38.92 / 39.17 ±0.20 / 39.51 ms │              43.74 / 45.37 ±2.64 / 50.62 ms │  1.16x slower │
│ QQuery 63 │        30.32 / 31.06 ±0.63 / 31.88 ms │              29.53 / 30.35 ±0.53 / 31.12 ms │     no change │
│ QQuery 64 │     370.49 / 376.33 ±6.27 / 388.52 ms │           355.80 / 361.47 ±3.81 / 364.98 ms │     no change │
│ QQuery 65 │        75.41 / 77.90 ±4.30 / 86.49 ms │              68.53 / 69.85 ±0.88 / 71.23 ms │ +1.12x faster │
│ QQuery 66 │        75.66 / 76.00 ±0.29 / 76.43 ms │              74.10 / 75.26 ±0.68 / 75.97 ms │     no change │
│ QQuery 67 │     201.74 / 206.10 ±3.71 / 212.14 ms │           238.12 / 241.65 ±3.61 / 248.00 ms │  1.17x slower │
│ QQuery 68 │           7.00 / 7.14 ±0.09 / 7.25 ms │                 6.66 / 6.82 ±0.13 / 6.99 ms │     no change │
│ QQuery 69 │        54.67 / 55.11 ±0.23 / 55.27 ms │              56.47 / 57.08 ±0.79 / 58.58 ms │     no change │
│ QQuery 70 │      98.73 / 101.44 ±3.77 / 108.89 ms │           102.02 / 102.56 ±0.45 / 103.33 ms │     no change │
│ QQuery 71 │        35.62 / 36.40 ±0.59 / 37.12 ms │              35.54 / 39.94 ±7.95 / 55.83 ms │  1.10x slower │
│ QQuery 72 │ 1602.47 / 1673.32 ±51.02 / 1759.64 ms │       1708.83 / 1752.19 ±45.71 / 1838.35 ms │     no change │
│ QQuery 73 │           5.91 / 5.98 ±0.05 / 6.05 ms │                 5.43 / 5.57 ±0.14 / 5.83 ms │ +1.07x faster │
│ QQuery 74 │     252.61 / 256.24 ±2.55 / 259.69 ms │           206.36 / 212.12 ±4.92 / 220.65 ms │ +1.21x faster │
│ QQuery 75 │     142.47 / 143.37 ±0.58 / 143.97 ms │           138.74 / 142.73 ±5.12 / 152.73 ms │     no change │
│ QQuery 76 │        35.44 / 35.69 ±0.24 / 36.10 ms │              35.00 / 35.20 ±0.16 / 35.44 ms │     no change │
│ QQuery 77 │        61.25 / 61.70 ±0.41 / 62.44 ms │              60.15 / 60.22 ±0.07 / 60.32 ms │     no change │
│ QQuery 78 │    274.10 / 296.77 ±22.02 / 326.69 ms │          274.36 / 289.49 ±14.93 / 315.01 ms │     no change │
│ QQuery 79 │        69.18 / 70.50 ±1.39 / 73.17 ms │              66.32 / 68.19 ±1.02 / 69.26 ms │     no change │
│ QQuery 80 │       96.36 / 99.55 ±3.89 / 107.17 ms │            96.58 / 100.82 ±4.48 / 109.28 ms │     no change │
│ QQuery 81 │        29.50 / 30.33 ±0.83 / 31.87 ms │              26.39 / 26.62 ±0.26 / 27.09 ms │ +1.14x faster │
│ QQuery 82 │        18.12 / 18.60 ±0.49 / 19.53 ms │              17.02 / 18.28 ±1.46 / 21.03 ms │     no change │
│ QQuery 83 │        33.47 / 34.01 ±0.31 / 34.30 ms │              33.20 / 33.79 ±0.33 / 34.20 ms │     no change │
│ QQuery 84 │        31.07 / 31.52 ±0.29 / 31.90 ms │              29.55 / 29.82 ±0.16 / 30.01 ms │ +1.06x faster │
│ QQuery 85 │       96.27 / 99.34 ±3.81 / 106.86 ms │            97.57 / 101.62 ±3.25 / 107.40 ms │     no change │
│ QQuery 86 │        24.72 / 24.97 ±0.14 / 25.13 ms │              25.75 / 26.42 ±0.56 / 27.39 ms │  1.06x slower │
│ QQuery 87 │        66.04 / 67.13 ±1.05 / 68.98 ms │              63.80 / 65.41 ±0.81 / 66.01 ms │     no change │
│ QQuery 88 │        60.32 / 61.04 ±0.64 / 62.00 ms │              60.57 / 63.70 ±5.30 / 74.29 ms │     no change │
│ QQuery 89 │        35.77 / 36.78 ±1.14 / 38.97 ms │              36.04 / 36.79 ±0.55 / 37.52 ms │     no change │
│ QQuery 90 │        16.47 / 17.01 ±0.78 / 18.55 ms │              16.69 / 17.02 ±0.35 / 17.57 ms │     no change │
│ QQuery 91 │        39.94 / 40.62 ±0.42 / 41.09 ms │              41.14 / 41.77 ±0.35 / 42.16 ms │     no change │
│ QQuery 92 │        29.64 / 29.96 ±0.25 / 30.32 ms │              30.59 / 31.23 ±0.81 / 32.78 ms │     no change │
│ QQuery 93 │        49.77 / 50.58 ±0.82 / 51.94 ms │              49.85 / 50.86 ±0.60 / 51.62 ms │     no change │
│ QQuery 94 │        38.76 / 39.31 ±0.28 / 39.57 ms │              38.80 / 41.19 ±2.72 / 46.51 ms │     no change │
│ QQuery 95 │        80.34 / 81.38 ±0.86 / 82.37 ms │              79.64 / 80.13 ±0.42 / 80.82 ms │     no change │
│ QQuery 96 │        24.27 / 24.36 ±0.07 / 24.47 ms │              23.99 / 24.26 ±0.17 / 24.43 ms │     no change │
│ QQuery 97 │        52.85 / 53.40 ±0.40 / 54.09 ms │              52.89 / 53.61 ±0.73 / 54.73 ms │     no change │
│ QQuery 98 │        38.61 / 38.96 ±0.27 / 39.37 ms │              44.57 / 46.13 ±1.44 / 48.83 ms │  1.18x slower │
│ QQuery 99 │        57.34 / 57.81 ±0.36 / 58.20 ms │              64.94 / 65.54 ±0.54 / 66.55 ms │  1.13x slower │
└───────────┴───────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 9453.68ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 9293.71ms │
│ Average Time (HEAD)                                        │   95.49ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   93.88ms │
│ Queries Faster                                             │        21 │
│ Queries Slower                                             │        12 │
│ Queries with No Change                                     │        66 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca | Changed: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca

tpcds — tpcds_sf1

Query Base Changed Change
Query 1 135.7 KiB 199.6 KiB +47.1%
Query 2 91.0 MiB 73.1 MiB -19.7%
Query 3 1.9 MiB 3.5 MiB +82.9%
Query 4 85.5 MiB 120.5 MiB +40.9%
Query 5 13.0 MiB 14.7 MiB +13.2%
Query 6 4.4 MiB 7.8 MiB +77.0%
Query 7 3.6 MiB 10.4 MiB +189.1%
Query 8 1.8 MiB 3.6 MiB +104.1%
Query 9 32.1 KiB 32.1 KiB +0.0%
Query 10 2.0 MiB 3.1 MiB +53.3%
Query 11 58.1 MiB 82.9 MiB +42.9%
Query 12 114.4 MiB 131.1 MiB +14.6%
Query 13 13.0 MiB 16.9 MiB +29.9%
Query 14 170.9 MiB 170.9 MiB +0.0%
Query 15 4.1 MiB 6.8 MiB +66.3%
Query 16 118.6 KiB 161.5 KiB +36.2%
Query 17 39.5 MiB 39.5 MiB +0.0%
Query 18 21.2 MiB 21.2 MiB -0.3%
Query 19 2.1 MiB 4.1 MiB +92.2%
Query 20 114.6 MiB 130.4 MiB +13.8%
Query 21 3.2 MiB 4.5 MiB +38.7%
Query 22 61.8 MiB 93.4 MiB +51.2%
Query 23 212.9 MiB 215.3 MiB +1.1%
Query 24 36.4 MiB 45.9 MiB +25.9%
Query 25 85.2 MiB 85.2 MiB -0.0%
Query 26 3.4 MiB 10.2 MiB +201.9%
Query 27 929.6 KiB 987.7 KiB +6.2%
Query 28 1.3 MiB 1.3 MiB +0.0%
Query 29 97.2 MiB 97.0 MiB -0.2%
Query 30 3.7 MiB 6.9 MiB +88.3%
Query 31 13.4 MiB 18.0 MiB +34.3%
Query 32 2.7 MiB 2.5 MiB -8.1%
Query 33 4.2 MiB 4.0 MiB -3.7%
Query 34 96.3 KiB 96.3 KiB +0.0%
Query 35 16.6 MiB 26.6 MiB +60.3%
Query 36 10.3 MiB 10.6 MiB +2.4%
Query 37 3.3 MiB 3.8 MiB +14.8%
Query 38 10.4 MiB 19.7 MiB +88.1%
Query 39 41.5 MiB 42.6 MiB +2.5%
Query 40 11.5 MiB 13.4 MiB +16.7%
Query 41 2.1 MiB 2.1 MiB +3.9%
Query 42 1.2 MiB 1.7 MiB +36.0%
Query 43 275.4 KiB 192.4 KiB -30.1%
Query 44 81.1 KiB 192.4 KiB +137.3%
Query 45 4.6 MiB 8.0 MiB +71.8%
Query 46 5.4 MiB 9.4 MiB +73.5%
Query 47 138.2 MiB 157.3 MiB +13.8%
Query 48 11.4 MiB 15.2 MiB +33.9%
Query 49 62.4 MiB 61.6 MiB -1.3%
Query 50 16.5 MiB 14.6 MiB -11.5%
Query 51 178.0 MiB 177.7 MiB -0.2%
Query 52 2.0 MiB 3.3 MiB +68.0%
Query 53 30.6 MiB 30.6 MiB +0.0%
Query 54 1.9 MiB 4.2 MiB +119.8%
Query 55 1.7 MiB 3.0 MiB +73.0%
Query 56 2.8 MiB 5.0 MiB +79.3%
Query 57 120.8 MiB 141.9 MiB +17.4%
Query 58 4.4 MiB 6.3 MiB +43.3%
Query 59 11.6 MiB 14.9 MiB +28.1%
Query 60 5.0 MiB 7.0 MiB +39.7%
Query 61 2.4 KiB 2.4 KiB +0.0%
Query 62 3.7 MiB 7.7 MiB +111.2%
Query 63 30.4 MiB 20.6 MiB -32.4%
Query 64 106.2 MiB 128.0 MiB +20.5%
Query 65 29.5 MiB 28.8 MiB -2.3%
Query 66 15.0 MiB 16.0 MiB +6.8%
Query 67 117.0 MiB 265.2 MiB +126.7%
Query 68 218.6 KiB 96.3 KiB -56.0%
Query 69 3.8 MiB 47.3 MiB +1139.1%
Query 70 32.5 MiB 50.0 MiB +54.0%
Query 71 50.4 MiB 70.8 MiB +40.4%
Query 72 91.2 MiB 91.1 MiB -0.1%
Query 73 96.3 KiB 96.3 KiB +0.0%
Query 74 35.2 MiB 49.4 MiB +40.4%
Query 75 23.7 MiB 22.6 MiB -4.9%
Query 76 4.2 MiB 5.1 MiB +20.5%
Query 77 4.1 MiB 4.0 MiB -2.0%
Query 78 316.9 MiB 316.9 MiB +0.0%
Query 79 8.7 MiB 12.4 MiB +42.6%
Query 80 41.6 MiB 44.0 MiB +5.8%
Query 81 4.1 MiB 11.4 MiB +176.3%
Query 82 4.9 MiB 3.8 MiB -23.2%
Query 83 8.2 MiB 11.6 MiB +41.7%
Query 84 11.6 MiB 12.0 MiB +3.1%
Query 85 17.0 MiB 21.6 MiB +27.1%
Query 86 82.1 MiB 105.2 MiB +28.2%
Query 87 12.9 MiB 20.6 MiB +59.8%
Query 88 1.6 MiB 1.6 MiB +6.1%
Query 89 112.2 MiB 130.0 MiB +15.9%
Query 90 348.6 KiB 348.6 KiB +0.0%
Query 91 31.3 MiB 22.9 MiB -26.9%
Query 92 2.5 MiB 2.5 MiB +0.0%
Query 93 15.1 MiB 15.1 MiB +0.0%
Query 94 3.7 MiB 3.7 MiB -0.0%
Query 95 19.4 MiB 19.4 MiB +0.0%
Query 96 224.7 KiB 224.7 KiB +0.0%
Query 97 33.1 MiB 33.1 MiB +0.2%
Query 98 114.7 MiB 131.5 MiB +14.6%
Query 99 1.9 MiB 7.8 MiB +307.8%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpcds base (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 316.9 MiB 2.3 GiB 2.0 GiB 7.5×
tpcds changed (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 316.9 MiB 2.1 GiB 1.8 GiB 6.7×
Resource Usage

tpcds — base (merge-base)

Metric Value
Wall time 50.0s
Peak memory 2.3 GiB
Avg memory 1.5 GiB
CPU user 191.2s
CPU sys 5.9s
Peak spill 0 B

tpcds — branch

Metric Value
Wall time 50.0s
Peak memory 2.1 GiB
Avg memory 1.4 GiB
CPU user 194.9s
CPU sys 6.0s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃     HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 69.34 ms │                                    75.51 ms │  1.09x slower │
│ QQuery 2  │ 20.87 ms │                                    20.67 ms │     no change │
│ QQuery 3  │ 41.80 ms │                                    42.03 ms │     no change │
│ QQuery 4  │ 25.01 ms │                                    25.14 ms │     no change │
│ QQuery 5  │ 51.00 ms │                                    51.90 ms │     no change │
│ QQuery 6  │ 20.53 ms │                                    21.78 ms │  1.06x slower │
│ QQuery 7  │ 63.53 ms │                                    59.23 ms │ +1.07x faster │
│ QQuery 8  │ 55.87 ms │                                    57.62 ms │     no change │
│ QQuery 9  │ 78.60 ms │                                    77.95 ms │     no change │
│ QQuery 10 │ 75.71 ms │                                    57.68 ms │ +1.31x faster │
│ QQuery 11 │ 15.49 ms │                                    15.45 ms │     no change │
│ QQuery 12 │ 29.49 ms │                                    28.17 ms │     no change │
│ QQuery 13 │ 58.35 ms │                                    47.85 ms │ +1.22x faster │
│ QQuery 14 │ 35.42 ms │                                    33.44 ms │ +1.06x faster │
│ QQuery 15 │ 45.72 ms │                                    48.74 ms │  1.07x slower │
│ QQuery 16 │ 15.52 ms │                                    15.98 ms │     no change │
│ QQuery 17 │ 89.38 ms │                                    92.14 ms │     no change │
│ QQuery 18 │ 84.61 ms │                                    61.03 ms │ +1.39x faster │
│ QQuery 19 │ 45.14 ms │                                    51.07 ms │  1.13x slower │
│ QQuery 20 │ 47.83 ms │                                    44.09 ms │ +1.08x faster │
│ QQuery 21 │ 83.56 ms │                                    87.60 ms │     no change │
│ QQuery 22 │ 17.97 ms │                                    17.15 ms │     no change │
└───────────┴──────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 1070.74ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 1032.22ms │
│ Average Time (HEAD)                                        │   48.67ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   46.92ms │
│ Queries Faster                                             │         6 │
│ Queries Slower                                             │         4 │
│ Queries with No Change                                     │        12 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 69.34 / 76.89 ±5.77 / 84.34 ms │              75.51 / 79.15 ±2.32 / 82.31 ms │     no change │
│ QQuery 2  │ 20.87 / 22.58 ±1.95 / 26.32 ms │              20.67 / 22.57 ±1.88 / 26.15 ms │     no change │
│ QQuery 3  │ 41.80 / 43.40 ±0.84 / 44.27 ms │              42.03 / 43.94 ±1.47 / 46.41 ms │     no change │
│ QQuery 4  │ 25.01 / 25.69 ±0.60 / 26.65 ms │              25.14 / 25.75 ±0.58 / 26.68 ms │     no change │
│ QQuery 5  │ 51.00 / 52.91 ±1.55 / 54.98 ms │              51.90 / 52.91 ±0.79 / 54.22 ms │     no change │
│ QQuery 6  │ 20.53 / 22.42 ±1.16 / 23.81 ms │              21.78 / 23.02 ±1.03 / 24.85 ms │     no change │
│ QQuery 7  │ 63.53 / 64.50 ±0.84 / 65.97 ms │              59.23 / 62.74 ±2.13 / 65.50 ms │     no change │
│ QQuery 8  │ 55.87 / 59.60 ±2.57 / 63.55 ms │              57.62 / 59.94 ±1.48 / 61.54 ms │     no change │
│ QQuery 9  │ 78.60 / 80.63 ±1.13 / 81.93 ms │              77.95 / 79.46 ±1.71 / 82.25 ms │     no change │
│ QQuery 10 │ 75.71 / 81.31 ±3.40 / 85.82 ms │              57.68 / 59.98 ±1.86 / 63.14 ms │ +1.36x faster │
│ QQuery 11 │ 15.49 / 16.64 ±0.91 / 17.92 ms │              15.45 / 16.50 ±0.96 / 17.97 ms │     no change │
│ QQuery 12 │ 29.49 / 30.27 ±0.71 / 31.59 ms │              28.17 / 28.85 ±0.69 / 30.05 ms │     no change │
│ QQuery 13 │ 58.35 / 62.95 ±6.21 / 75.21 ms │              47.85 / 49.45 ±2.22 / 53.80 ms │ +1.27x faster │
│ QQuery 14 │ 35.42 / 36.17 ±0.93 / 37.81 ms │              33.44 / 36.17 ±2.02 / 39.65 ms │     no change │
│ QQuery 15 │ 45.72 / 47.78 ±1.51 / 50.23 ms │              48.74 / 49.88 ±0.85 / 51.29 ms │     no change │
│ QQuery 16 │ 15.52 / 15.87 ±0.22 / 16.15 ms │              15.98 / 16.73 ±0.65 / 17.70 ms │  1.05x slower │
│ QQuery 17 │ 89.38 / 93.12 ±2.12 / 95.95 ms │              92.14 / 93.62 ±0.92 / 94.40 ms │     no change │
│ QQuery 18 │ 84.61 / 90.42 ±4.99 / 96.41 ms │              61.03 / 64.60 ±2.04 / 67.28 ms │ +1.40x faster │
│ QQuery 19 │ 45.14 / 46.91 ±1.04 / 47.94 ms │              51.07 / 51.86 ±0.64 / 52.75 ms │  1.11x slower │
│ QQuery 20 │ 47.83 / 50.50 ±3.45 / 57.24 ms │              44.09 / 46.26 ±1.66 / 48.99 ms │ +1.09x faster │
│ QQuery 21 │ 83.56 / 85.67 ±2.51 / 90.31 ms │              87.60 / 89.98 ±3.55 / 97.00 ms │  1.05x slower │
│ QQuery 22 │ 17.97 / 19.72 ±1.02 / 20.85 ms │              17.15 / 17.65 ±0.39 / 18.26 ms │ +1.12x faster │
└───────────┴────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 1125.95ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 1071.03ms │
│ Average Time (HEAD)                                        │   51.18ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   48.68ms │
│ Queries Faster                                             │         5 │
│ Queries Slower                                             │         3 │
│ Queries with No Change                                     │        14 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca | Changed: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca

tpch — tpch_sf1

Query Base Changed Change
Query 1 21.2 MiB 31.4 MiB +47.7%
Query 2 23.1 MiB 15.9 MiB -31.3%
Query 3 11.1 MiB 11.2 MiB +0.2%
Query 4 22.2 MiB 33.4 MiB +50.3%
Query 5 42.7 MiB 43.2 MiB +1.0%
Query 6 832 B 832 B +0.0%
Query 7 103.6 MiB 107.1 MiB +3.4%
Query 8 33.1 MiB 39.8 MiB +20.2%
Query 9 140.9 MiB 140.6 MiB -0.2%
Query 10 50.0 MiB 43.9 MiB -12.2%
Query 11 31.7 MiB 51.2 MiB +61.4%
Query 12 22.5 MiB 22.6 MiB +0.5%
Query 13 92.3 MiB 72.3 MiB -21.6%
Query 14 6.7 MiB 6.7 MiB +0.0%
Query 15 13.0 MiB 12.4 MiB -4.5%
Query 16 42.5 MiB 45.0 MiB +6.0%
Query 17 153.3 MiB 154.5 MiB +0.8%
Query 18 116.3 MiB 167.1 MiB +43.6%
Query 19 209.7 KiB 725.7 KiB +246.1%
Query 20 59.8 MiB 57.3 MiB -4.2%
Query 21 36.1 MiB 42.4 MiB +17.6%
Query 22 31.5 MiB 32.2 MiB +2.1%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpch base (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 153.3 MiB 974.0 MiB 820.7 MiB 6.4×
tpch changed (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 167.1 MiB 1.1 GiB 991.0 MiB 6.9×
Resource Usage

tpch — base (merge-base)

Metric Value
Wall time 10.0s
Peak memory 974.0 MiB
Avg memory 391.7 MiB
CPU user 35.2s
CPU sys 1.9s
Peak spill 0 B

tpch — branch

Metric Value
Wall time 10.0s
Peak memory 1.1 GiB
Avg memory 424.6 MiB
CPU user 34.6s
CPU sys 2.0s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939384597-5269-vxhtl 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "false"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939384237-5270-c4qqj 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 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "false"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939384597-5268-ph258 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "false"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6939384597-5267-zgt75 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "false"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │    5.77 ms │                                     5.35 ms │ +1.08x faster │
│ QQuery 2  │   85.84 ms │                                    93.20 ms │  1.09x slower │
│ QQuery 3  │   35.26 ms │                                    35.11 ms │     no change │
│ QQuery 4  │  582.11 ms │                                   480.79 ms │ +1.21x faster │
│ QQuery 5  │   59.08 ms │                                    56.22 ms │     no change │
│ QQuery 6  │   39.85 ms │                                    39.56 ms │     no change │
│ QQuery 7  │   88.61 ms │                                    87.97 ms │     no change │
│ QQuery 8  │   44.98 ms │                                    42.13 ms │ +1.07x faster │
│ QQuery 9  │   83.18 ms │                                    80.87 ms │     no change │
│ QQuery 10 │   66.72 ms │                                    68.71 ms │     no change │
│ QQuery 11 │  368.58 ms │                                   280.89 ms │ +1.31x faster │
│ QQuery 12 │   33.23 ms │                                    34.27 ms │     no change │
│ QQuery 13 │  138.06 ms │                                   136.86 ms │     no change │
│ QQuery 14 │  516.27 ms │                                   521.82 ms │     no change │
│ QQuery 15 │   53.69 ms │                                    53.93 ms │     no change │
│ QQuery 16 │    6.27 ms │                                     5.93 ms │ +1.06x faster │
│ QQuery 17 │  104.01 ms │                                    97.51 ms │ +1.07x faster │
│ QQuery 18 │  120.88 ms │                                   121.64 ms │     no change │
│ QQuery 19 │   49.32 ms │                                    49.69 ms │     no change │
│ QQuery 20 │   37.46 ms │                                    40.78 ms │  1.09x slower │
│ QQuery 21 │   22.13 ms │                                    21.61 ms │     no change │
│ QQuery 22 │   94.01 ms │                                   106.30 ms │  1.13x slower │
│ QQuery 23 │  404.04 ms │                                   394.00 ms │     no change │
│ QQuery 24 │  258.59 ms │                                   243.23 ms │ +1.06x faster │
│ QQuery 25 │  139.98 ms │                                   136.26 ms │     no change │
│ QQuery 26 │   52.41 ms │                                    52.86 ms │     no change │
│ QQuery 27 │    5.60 ms │                                     4.91 ms │ +1.14x faster │
│ QQuery 28 │   89.80 ms │                                    82.68 ms │ +1.09x faster │
│ QQuery 29 │  127.25 ms │                                   123.03 ms │     no change │
│ QQuery 30 │   42.24 ms │                                    36.06 ms │ +1.17x faster │
│ QQuery 31 │  132.41 ms │                                   127.21 ms │     no change │
│ QQuery 32 │   23.98 ms │                                    22.58 ms │ +1.06x faster │
│ QQuery 33 │   43.92 ms │                                    42.38 ms │     no change │
│ QQuery 34 │    5.88 ms │                                     5.29 ms │ +1.11x faster │
│ QQuery 35 │   77.94 ms │                                    78.66 ms │     no change │
│ QQuery 36 │    6.18 ms │                                     5.46 ms │ +1.13x faster │
│ QQuery 37 │    8.35 ms │                                     7.01 ms │ +1.19x faster │
│ QQuery 38 │   74.66 ms │                                    71.43 ms │     no change │
│ QQuery 39 │   97.64 ms │                                    89.79 ms │ +1.09x faster │
│ QQuery 40 │   29.10 ms │                                    28.71 ms │     no change │
│ QQuery 41 │   15.19 ms │                                    13.83 ms │ +1.10x faster │
│ QQuery 42 │   30.09 ms │                                    29.87 ms │     no change │
│ QQuery 43 │    4.83 ms │                                     4.60 ms │     no change │
│ QQuery 44 │    5.80 ms │                                     5.89 ms │     no change │
│ QQuery 45 │   39.06 ms │                                    38.90 ms │     no change │
│ QQuery 46 │   12.54 ms │                                    13.58 ms │  1.08x slower │
│ QQuery 47 │  262.01 ms │                                   267.04 ms │     no change │
│ QQuery 48 │  112.32 ms │                                   112.81 ms │     no change │
│ QQuery 49 │   86.04 ms │                                    90.56 ms │  1.05x slower │
│ QQuery 50 │   73.47 ms │                                    74.20 ms │     no change │
│ QQuery 51 │  120.82 ms │                                   114.78 ms │     no change │
│ QQuery 52 │   30.22 ms │                                    29.80 ms │     no change │
│ QQuery 53 │   38.07 ms │                                    37.68 ms │     no change │
│ QQuery 54 │   62.89 ms │                                    60.58 ms │     no change │
│ QQuery 55 │   30.48 ms │                                    29.01 ms │     no change │
│ QQuery 56 │   45.79 ms │                                    45.88 ms │     no change │
│ QQuery 57 │  187.71 ms │                                   201.67 ms │  1.07x slower │
│ QQuery 58 │  120.73 ms │                                   117.74 ms │     no change │
│ QQuery 59 │  129.00 ms │                                   140.70 ms │  1.09x slower │
│ QQuery 60 │   47.58 ms │                                    44.17 ms │ +1.08x faster │
│ QQuery 61 │    8.98 ms │                                     8.17 ms │ +1.10x faster │
│ QQuery 62 │   46.74 ms │                                    49.81 ms │  1.07x slower │
│ QQuery 63 │   38.10 ms │                                    34.98 ms │ +1.09x faster │
│ QQuery 64 │  442.59 ms │                                   426.70 ms │     no change │
│ QQuery 65 │   88.70 ms │                                    80.97 ms │ +1.10x faster │
│ QQuery 66 │   82.26 ms │                                    78.98 ms │     no change │
│ QQuery 67 │  215.95 ms │                                   258.73 ms │  1.20x slower │
│ QQuery 68 │    6.56 ms │                                     6.40 ms │     no change │
│ QQuery 69 │   62.99 ms │                                    62.33 ms │     no change │
│ QQuery 70 │  115.40 ms │                                   124.57 ms │  1.08x slower │
│ QQuery 71 │   40.51 ms │                                    39.73 ms │     no change │
│ QQuery 72 │ 2253.28 ms │                                  2261.55 ms │     no change │
│ QQuery 73 │    5.16 ms │                                     5.15 ms │     no change │
│ QQuery 74 │  258.38 ms │                                   203.18 ms │ +1.27x faster │
│ QQuery 75 │  169.08 ms │                                   184.06 ms │  1.09x slower │
│ QQuery 76 │   41.96 ms │                                    45.50 ms │  1.08x slower │
│ QQuery 77 │   68.53 ms │                                    71.25 ms │     no change │
│ QQuery 78 │  367.37 ms │                                   374.41 ms │     no change │
│ QQuery 79 │   76.15 ms │                                    79.89 ms │     no change │
│ QQuery 80 │  123.99 ms │                                   128.07 ms │     no change │
│ QQuery 81 │   31.65 ms │                                    29.06 ms │ +1.09x faster │
│ QQuery 82 │   20.60 ms │                                    20.20 ms │     no change │
│ QQuery 83 │   33.31 ms │                                    35.89 ms │  1.08x slower │
│ QQuery 84 │   33.34 ms │                                    33.01 ms │     no change │
│ QQuery 85 │  110.69 ms │                                   120.36 ms │  1.09x slower │
│ QQuery 86 │   28.45 ms │                                    31.14 ms │  1.09x slower │
│ QQuery 87 │   71.30 ms │                                    74.10 ms │     no change │
│ QQuery 88 │   84.92 ms │                                    83.19 ms │     no change │
│ QQuery 89 │   42.48 ms │                                    43.93 ms │     no change │
│ QQuery 90 │   18.50 ms │                                    18.35 ms │     no change │
│ QQuery 91 │   45.60 ms │                                    46.31 ms │     no change │
│ QQuery 92 │   37.87 ms │                                    36.86 ms │     no change │
│ QQuery 93 │   69.20 ms │                                    70.35 ms │     no change │
│ QQuery 94 │   45.93 ms │                                    47.04 ms │     no change │
│ QQuery 95 │  101.65 ms │                                   102.88 ms │     no change │
│ QQuery 96 │   27.39 ms │                                    28.04 ms │     no change │
│ QQuery 97 │   68.91 ms │                                    68.75 ms │     no change │
│ QQuery 98 │   45.42 ms │                                    51.95 ms │  1.14x slower │
│ QQuery 99 │   68.12 ms │                                    78.52 ms │  1.15x slower │
└───────────┴────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 11207.95ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 11030.36ms │
│ Average Time (HEAD)                                        │   113.21ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   111.42ms │
│ Queries Faster                                             │         22 │
│ Queries Slower                                             │         17 │
│ Queries with No Change                                     │         60 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │          5.77 / 7.27 ±2.39 / 12.01 ms │                5.35 / 6.73 ±2.09 / 10.88 ms │ +1.08x faster │
│ QQuery 2  │       85.84 / 92.86 ±7.98 / 107.54 ms │            93.20 / 103.24 ±7.19 / 115.68 ms │  1.11x slower │
│ QQuery 3  │        35.26 / 39.23 ±4.50 / 45.37 ms │              35.11 / 39.44 ±4.26 / 45.81 ms │     no change │
│ QQuery 4  │    582.11 / 609.39 ±22.16 / 639.52 ms │          480.79 / 499.33 ±10.40 / 508.20 ms │ +1.22x faster │
│ QQuery 5  │        59.08 / 69.70 ±8.34 / 82.63 ms │              56.22 / 60.79 ±4.10 / 66.76 ms │ +1.15x faster │
│ QQuery 6  │        39.85 / 44.68 ±5.41 / 51.54 ms │              39.56 / 42.01 ±4.30 / 50.57 ms │ +1.06x faster │
│ QQuery 7  │       88.61 / 92.60 ±6.36 / 105.21 ms │             87.97 / 91.78 ±5.46 / 102.58 ms │     no change │
│ QQuery 8  │        44.98 / 49.40 ±4.88 / 58.51 ms │              42.13 / 45.10 ±4.48 / 53.93 ms │ +1.10x faster │
│ QQuery 9  │        83.18 / 89.02 ±4.90 / 94.87 ms │              80.87 / 85.03 ±4.53 / 91.15 ms │     no change │
│ QQuery 10 │        66.72 / 71.83 ±4.20 / 77.31 ms │              68.71 / 73.21 ±3.43 / 78.35 ms │     no change │
│ QQuery 11 │     368.58 / 378.28 ±8.80 / 392.35 ms │          280.89 / 301.28 ±12.08 / 317.82 ms │ +1.26x faster │
│ QQuery 12 │        33.23 / 33.59 ±0.31 / 33.95 ms │              34.27 / 35.09 ±0.79 / 36.52 ms │     no change │
│ QQuery 13 │     138.06 / 145.61 ±8.35 / 158.14 ms │           136.86 / 149.30 ±8.92 / 161.56 ms │     no change │
│ QQuery 14 │    516.27 / 539.30 ±15.49 / 560.19 ms │          521.82 / 535.59 ±12.16 / 551.75 ms │     no change │
│ QQuery 15 │        53.69 / 57.63 ±3.78 / 63.96 ms │              53.93 / 54.99 ±0.99 / 56.84 ms │     no change │
│ QQuery 16 │           6.27 / 6.68 ±0.34 / 7.19 ms │                 5.93 / 6.28 ±0.24 / 6.63 ms │ +1.06x faster │
│ QQuery 17 │     104.01 / 110.25 ±5.70 / 120.73 ms │             97.51 / 99.27 ±1.48 / 101.36 ms │ +1.11x faster │
│ QQuery 18 │     120.88 / 125.23 ±5.83 / 136.69 ms │           121.64 / 126.21 ±4.13 / 131.36 ms │     no change │
│ QQuery 19 │        49.32 / 50.69 ±0.71 / 51.27 ms │              49.69 / 54.37 ±3.52 / 59.29 ms │  1.07x slower │
│ QQuery 20 │        37.46 / 40.64 ±3.20 / 46.72 ms │              40.78 / 41.46 ±0.67 / 42.70 ms │     no change │
│ QQuery 21 │        22.13 / 23.28 ±0.69 / 24.22 ms │              21.61 / 23.20 ±1.04 / 24.44 ms │     no change │
│ QQuery 22 │       94.01 / 97.40 ±2.22 / 100.73 ms │           106.30 / 110.21 ±2.89 / 114.28 ms │  1.13x slower │
│ QQuery 23 │     404.04 / 416.33 ±6.79 / 424.77 ms │           394.00 / 400.61 ±8.02 / 415.94 ms │     no change │
│ QQuery 24 │     258.59 / 264.81 ±6.28 / 276.28 ms │          243.23 / 263.95 ±15.46 / 286.13 ms │     no change │
│ QQuery 25 │     139.98 / 147.51 ±7.44 / 158.26 ms │           136.26 / 139.13 ±2.15 / 142.57 ms │ +1.06x faster │
│ QQuery 26 │        52.41 / 54.63 ±1.49 / 57.07 ms │              52.86 / 53.55 ±0.64 / 54.61 ms │     no change │
│ QQuery 27 │           5.60 / 5.70 ±0.12 / 5.94 ms │                 4.91 / 5.08 ±0.16 / 5.32 ms │ +1.12x faster │
│ QQuery 28 │        89.80 / 91.37 ±1.10 / 92.62 ms │              82.68 / 90.95 ±5.41 / 99.47 ms │     no change │
│ QQuery 29 │     127.25 / 134.39 ±6.06 / 144.41 ms │           123.03 / 133.96 ±8.13 / 142.58 ms │     no change │
│ QQuery 30 │        42.24 / 43.39 ±0.93 / 44.40 ms │              36.06 / 36.87 ±0.54 / 37.71 ms │ +1.18x faster │
│ QQuery 31 │    132.41 / 145.86 ±10.72 / 162.26 ms │           127.21 / 137.91 ±7.83 / 148.62 ms │ +1.06x faster │
│ QQuery 32 │        23.98 / 24.91 ±0.65 / 25.84 ms │              22.58 / 23.59 ±0.61 / 24.25 ms │ +1.06x faster │
│ QQuery 33 │        43.92 / 48.78 ±5.35 / 56.28 ms │              42.38 / 45.50 ±5.57 / 56.63 ms │ +1.07x faster │
│ QQuery 34 │           5.88 / 6.15 ±0.47 / 7.09 ms │                 5.29 / 5.57 ±0.26 / 6.06 ms │ +1.10x faster │
│ QQuery 35 │        77.94 / 80.15 ±2.24 / 84.38 ms │             78.66 / 85.02 ±8.45 / 101.36 ms │  1.06x slower │
│ QQuery 36 │           6.18 / 6.51 ±0.41 / 7.31 ms │                 5.46 / 5.81 ±0.52 / 6.84 ms │ +1.12x faster │
│ QQuery 37 │           8.35 / 8.57 ±0.12 / 8.69 ms │                 7.01 / 7.94 ±0.77 / 9.08 ms │ +1.08x faster │
│ QQuery 38 │        74.66 / 77.66 ±4.11 / 85.78 ms │              71.43 / 77.49 ±4.94 / 84.39 ms │     no change │
│ QQuery 39 │      97.64 / 101.11 ±3.18 / 105.14 ms │             89.79 / 96.35 ±4.42 / 103.11 ms │     no change │
│ QQuery 40 │        29.10 / 30.28 ±0.62 / 30.91 ms │              28.71 / 29.98 ±0.89 / 31.42 ms │     no change │
│ QQuery 41 │        15.19 / 15.48 ±0.26 / 15.97 ms │              13.83 / 13.99 ±0.16 / 14.25 ms │ +1.11x faster │
│ QQuery 42 │        30.09 / 31.93 ±3.04 / 37.93 ms │              29.87 / 32.60 ±4.20 / 40.85 ms │     no change │
│ QQuery 43 │           4.83 / 5.00 ±0.22 / 5.44 ms │                 4.60 / 4.78 ±0.16 / 5.05 ms │     no change │
│ QQuery 44 │           5.80 / 6.00 ±0.16 / 6.20 ms │                 5.89 / 6.03 ±0.22 / 6.46 ms │     no change │
│ QQuery 45 │        39.06 / 39.72 ±0.63 / 40.86 ms │              38.90 / 40.45 ±0.87 / 41.52 ms │     no change │
│ QQuery 46 │        12.54 / 13.00 ±0.28 / 13.37 ms │              13.58 / 14.44 ±0.85 / 15.74 ms │  1.11x slower │
│ QQuery 47 │    262.01 / 278.94 ±18.39 / 313.81 ms │          267.04 / 288.92 ±12.93 / 305.80 ms │     no change │
│ QQuery 48 │     112.32 / 116.59 ±6.85 / 130.26 ms │           112.81 / 118.90 ±6.71 / 127.65 ms │     no change │
│ QQuery 49 │       86.04 / 96.59 ±6.62 / 104.78 ms │             90.56 / 96.45 ±3.84 / 102.28 ms │     no change │
│ QQuery 50 │        73.47 / 83.19 ±7.12 / 90.89 ms │              74.20 / 74.87 ±0.78 / 76.29 ms │ +1.11x faster │
│ QQuery 51 │     120.82 / 132.16 ±7.66 / 143.11 ms │           114.78 / 122.39 ±5.36 / 130.70 ms │ +1.08x faster │
│ QQuery 52 │        30.22 / 32.94 ±3.80 / 40.48 ms │              29.80 / 33.90 ±4.44 / 40.75 ms │     no change │
│ QQuery 53 │        38.07 / 40.94 ±5.37 / 51.68 ms │              37.68 / 40.32 ±4.10 / 48.50 ms │     no change │
│ QQuery 54 │        62.89 / 69.00 ±7.15 / 79.02 ms │              60.58 / 63.37 ±2.77 / 68.58 ms │ +1.09x faster │
│ QQuery 55 │        30.48 / 34.65 ±4.43 / 41.59 ms │              29.01 / 31.31 ±3.90 / 39.09 ms │ +1.11x faster │
│ QQuery 56 │        45.79 / 49.87 ±4.02 / 56.07 ms │              45.88 / 48.15 ±2.04 / 51.52 ms │     no change │
│ QQuery 57 │     187.71 / 191.53 ±5.29 / 201.70 ms │           201.67 / 207.44 ±6.83 / 220.05 ms │  1.08x slower │
│ QQuery 58 │     120.73 / 128.70 ±6.70 / 137.59 ms │          117.74 / 144.61 ±26.09 / 192.14 ms │  1.12x slower │
│ QQuery 59 │    129.00 / 137.49 ±10.21 / 150.18 ms │           140.70 / 147.63 ±6.67 / 156.21 ms │  1.07x slower │
│ QQuery 60 │        47.58 / 49.39 ±2.79 / 54.93 ms │              44.17 / 49.53 ±4.42 / 55.05 ms │     no change │
│ QQuery 61 │          8.98 / 9.82 ±1.00 / 11.68 ms │                 8.17 / 8.65 ±0.44 / 9.38 ms │ +1.13x faster │
│ QQuery 62 │        46.74 / 47.42 ±0.49 / 48.27 ms │              49.81 / 51.26 ±0.93 / 52.66 ms │  1.08x slower │
│ QQuery 63 │        38.10 / 38.68 ±0.58 / 39.66 ms │              34.98 / 42.27 ±5.91 / 48.69 ms │  1.09x slower │
│ QQuery 64 │    442.59 / 459.10 ±15.51 / 485.94 ms │           426.70 / 439.75 ±7.92 / 450.29 ms │     no change │
│ QQuery 65 │       88.70 / 93.87 ±5.42 / 104.25 ms │            80.97 / 90.10 ±11.11 / 111.89 ms │     no change │
│ QQuery 66 │        82.26 / 83.11 ±0.57 / 83.97 ms │              78.98 / 82.29 ±4.08 / 90.31 ms │     no change │
│ QQuery 67 │     215.95 / 227.91 ±9.89 / 244.21 ms │          258.73 / 273.10 ±15.28 / 296.99 ms │  1.20x slower │
│ QQuery 68 │           6.56 / 7.02 ±0.53 / 8.04 ms │                 6.40 / 6.87 ±0.65 / 8.15 ms │     no change │
│ QQuery 69 │        62.99 / 68.77 ±4.91 / 75.70 ms │              62.33 / 63.60 ±0.91 / 64.85 ms │ +1.08x faster │
│ QQuery 70 │     115.40 / 119.16 ±3.93 / 126.54 ms │           124.57 / 137.34 ±8.36 / 147.37 ms │  1.15x slower │
│ QQuery 71 │        40.51 / 43.36 ±2.76 / 47.11 ms │              39.73 / 44.46 ±4.36 / 51.08 ms │     no change │
│ QQuery 72 │ 2253.28 / 2274.42 ±19.89 / 2308.03 ms │       2261.55 / 2306.34 ±38.76 / 2369.89 ms │     no change │
│ QQuery 73 │           5.16 / 5.56 ±0.46 / 6.44 ms │                 5.15 / 5.42 ±0.41 / 6.24 ms │     no change │
│ QQuery 74 │     258.38 / 266.56 ±6.03 / 273.41 ms │          203.18 / 217.81 ±11.35 / 236.07 ms │ +1.22x faster │
│ QQuery 75 │     169.08 / 177.79 ±6.16 / 185.93 ms │           184.06 / 191.15 ±6.68 / 203.59 ms │  1.08x slower │
│ QQuery 76 │        41.96 / 46.89 ±5.24 / 54.68 ms │              45.50 / 47.86 ±1.54 / 50.27 ms │     no change │
│ QQuery 77 │        68.53 / 71.58 ±1.95 / 74.66 ms │              71.25 / 74.63 ±4.42 / 83.25 ms │     no change │
│ QQuery 78 │    367.37 / 389.97 ±25.57 / 434.71 ms │          374.41 / 396.68 ±27.91 / 449.57 ms │     no change │
│ QQuery 79 │      76.15 / 85.80 ±10.59 / 100.58 ms │              79.89 / 85.60 ±7.10 / 98.32 ms │     no change │
│ QQuery 80 │     123.99 / 127.92 ±4.14 / 134.01 ms │           128.07 / 136.32 ±4.41 / 140.70 ms │  1.07x slower │
│ QQuery 81 │        31.65 / 32.56 ±0.73 / 33.55 ms │              29.06 / 30.23 ±0.93 / 31.90 ms │ +1.08x faster │
│ QQuery 82 │        20.60 / 21.99 ±1.12 / 23.79 ms │              20.20 / 23.39 ±4.14 / 31.57 ms │  1.06x slower │
│ QQuery 83 │        33.31 / 34.00 ±0.43 / 34.44 ms │              35.89 / 36.54 ±0.70 / 37.80 ms │  1.07x slower │
│ QQuery 84 │        33.34 / 35.32 ±3.11 / 41.52 ms │              33.01 / 34.25 ±1.25 / 36.50 ms │     no change │
│ QQuery 85 │     110.69 / 122.07 ±9.35 / 134.51 ms │           120.36 / 127.92 ±7.60 / 138.55 ms │     no change │
│ QQuery 86 │        28.45 / 29.64 ±0.64 / 30.25 ms │              31.14 / 31.76 ±0.46 / 32.21 ms │  1.07x slower │
│ QQuery 87 │        71.30 / 76.34 ±5.27 / 82.98 ms │              74.10 / 81.53 ±5.89 / 88.78 ms │  1.07x slower │
│ QQuery 88 │        84.92 / 87.56 ±2.96 / 93.22 ms │              83.19 / 87.82 ±3.52 / 93.20 ms │     no change │
│ QQuery 89 │        42.48 / 46.38 ±4.93 / 56.08 ms │              43.93 / 46.89 ±3.91 / 54.48 ms │     no change │
│ QQuery 90 │        18.50 / 19.09 ±0.45 / 19.70 ms │              18.35 / 18.74 ±0.22 / 19.02 ms │     no change │
│ QQuery 91 │        45.60 / 47.02 ±0.76 / 47.77 ms │              46.31 / 47.52 ±0.67 / 48.29 ms │     no change │
│ QQuery 92 │        37.87 / 39.01 ±1.03 / 40.48 ms │              36.86 / 37.81 ±0.65 / 38.67 ms │     no change │
│ QQuery 93 │        69.20 / 72.70 ±4.34 / 80.74 ms │              70.35 / 73.70 ±4.50 / 82.57 ms │     no change │
│ QQuery 94 │        45.93 / 46.94 ±0.65 / 47.61 ms │              47.04 / 48.14 ±1.22 / 50.49 ms │     no change │
│ QQuery 95 │     101.65 / 106.46 ±3.00 / 109.50 ms │           102.88 / 105.73 ±1.68 / 107.50 ms │     no change │
│ QQuery 96 │        27.39 / 30.23 ±3.86 / 37.87 ms │              28.04 / 30.91 ±5.01 / 40.92 ms │     no change │
│ QQuery 97 │        68.91 / 71.78 ±2.50 / 76.10 ms │              68.75 / 72.60 ±4.93 / 81.52 ms │     no change │
│ QQuery 98 │        45.42 / 51.98 ±6.72 / 61.13 ms │              51.95 / 57.60 ±6.19 / 66.18 ms │  1.11x slower │
│ QQuery 99 │        68.12 / 69.01 ±1.31 / 71.56 ms │              78.52 / 82.68 ±5.57 / 93.57 ms │  1.20x slower │
└───────────┴───────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 11700.56ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 11580.51ms │
│ Average Time (HEAD)                                        │   118.19ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   116.97ms │
│ Queries Faster                                             │         26 │
│ Queries Slower                                             │         20 │
│ Queries with No Change                                     │         53 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca | Changed: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca

tpcds — tpcds_sf1

Query Base Changed Change
Query 1 139.6 KiB 142.0 KiB +1.7%
Query 2 56.6 MiB 52.2 MiB -7.8%
Query 3 2.4 MiB 4.1 MiB +68.4%
Query 4 85.6 MiB 121.6 MiB +42.1%
Query 5 14.0 MiB 16.0 MiB +13.8%
Query 6 4.7 MiB 7.5 MiB +60.0%
Query 7 7.9 MiB 12.1 MiB +53.5%
Query 8 2.0 MiB 3.8 MiB +88.1%
Query 9 32.1 KiB 32.1 KiB +0.0%
Query 10 2.0 MiB 3.1 MiB +53.3%
Query 11 58.1 MiB 85.1 MiB +46.3%
Query 12 114.4 MiB 130.3 MiB +13.9%
Query 13 15.7 MiB 18.9 MiB +20.6%
Query 14 170.9 MiB 170.9 MiB +0.0%
Query 15 4.5 MiB 7.1 MiB +57.9%
Query 16 118.8 KiB 161.5 KiB +36.0%
Query 17 39.8 MiB 39.6 MiB -0.6%
Query 18 21.3 MiB 34.2 MiB +60.8%
Query 19 2.4 MiB 5.9 MiB +151.2%
Query 20 114.6 MiB 90.3 MiB -21.2%
Query 21 3.3 MiB 5.6 MiB +72.5%
Query 22 61.8 MiB 94.7 MiB +53.4%
Query 23 211.6 MiB 215.6 MiB +1.9%
Query 24 36.4 MiB 45.9 MiB +25.9%
Query 25 85.3 MiB 85.3 MiB -0.1%
Query 26 9.9 MiB 11.8 MiB +19.9%
Query 27 929.6 KiB 987.7 KiB +6.2%
Query 28 1.5 MiB 1.5 MiB -0.1%
Query 29 97.2 MiB 96.9 MiB -0.3%
Query 30 3.6 MiB 7.8 MiB +117.2%
Query 31 13.4 MiB 18.0 MiB +34.3%
Query 32 2.9 MiB 2.9 MiB +0.0%
Query 33 4.2 MiB 4.2 MiB +0.0%
Query 34 12.7 KiB 96.3 KiB +656.9%
Query 35 19.3 MiB 30.2 MiB +56.8%
Query 36 10.3 MiB 10.6 MiB +2.4%
Query 37 3.3 MiB 3.8 MiB +14.8%
Query 38 10.4 MiB 19.9 MiB +90.8%
Query 39 35.1 MiB 42.7 MiB +21.6%
Query 40 12.0 MiB 14.4 MiB +19.6%
Query 41 2.1 MiB 2.5 MiB +21.0%
Query 42 1.6 MiB 2.0 MiB +23.0%
Query 43 275.4 KiB 192.4 KiB -30.1%
Query 44 97.6 KiB 192.4 KiB +97.2%
Query 45 4.6 MiB 8.3 MiB +82.0%
Query 46 5.4 MiB 9.4 MiB +73.5%
Query 47 128.5 MiB 157.3 MiB +22.4%
Query 48 12.5 MiB 16.4 MiB +30.9%
Query 49 42.1 MiB 34.4 MiB -18.3%
Query 50 16.3 MiB 14.8 MiB -9.6%
Query 51 178.1 MiB 177.7 MiB -0.3%
Query 52 2.1 MiB 4.0 MiB +84.8%
Query 53 30.6 MiB 20.6 MiB -32.7%
Query 54 2.0 MiB 4.1 MiB +103.2%
Query 55 2.1 MiB 3.5 MiB +68.2%
Query 56 3.2 MiB 5.2 MiB +63.3%
Query 57 130.6 MiB 140.9 MiB +7.9%
Query 58 4.6 MiB 6.8 MiB +47.9%
Query 59 12.6 MiB 14.9 MiB +18.6%
Query 60 5.7 MiB 9.5 MiB +67.6%
Query 61 2.5 KiB 2.5 KiB +0.0%
Query 62 4.3 MiB 8.4 MiB +95.1%
Query 63 30.6 MiB 20.6 MiB -32.7%
Query 64 125.3 MiB 108.0 MiB -13.8%
Query 65 29.5 MiB 29.4 MiB -0.4%
Query 66 18.0 MiB 16.0 MiB -11.5%
Query 67 116.3 MiB 265.7 MiB +128.6%
Query 68 218.6 KiB 96.3 KiB -56.0%
Query 69 3.8 MiB 47.9 MiB +1177.8%
Query 70 32.5 MiB 50.0 MiB +54.1%
Query 71 50.6 MiB 51.5 MiB +1.6%
Query 72 101.6 MiB 106.0 MiB +4.4%
Query 73 96.3 KiB 96.3 KiB +0.0%
Query 74 35.2 MiB 50.0 MiB +41.9%
Query 75 21.7 MiB 21.7 MiB +0.0%
Query 76 4.7 MiB 7.1 MiB +50.5%
Query 77 4.4 MiB 4.4 MiB +0.0%
Query 78 316.9 MiB 316.9 MiB +0.0%
Query 79 5.8 MiB 10.7 MiB +83.3%
Query 80 41.2 MiB 44.8 MiB +8.8%
Query 81 4.7 MiB 11.4 MiB +141.4%
Query 82 4.9 MiB 3.8 MiB -23.2%
Query 83 8.4 MiB 11.8 MiB +40.3%
Query 84 11.6 MiB 12.0 MiB +3.1%
Query 85 17.2 MiB 21.9 MiB +27.5%
Query 86 82.1 MiB 75.2 MiB -8.4%
Query 87 12.9 MiB 21.0 MiB +62.7%
Query 88 1.7 MiB 1.9 MiB +16.0%
Query 89 112.2 MiB 109.1 MiB -2.8%
Query 90 348.6 KiB 356.5 KiB +2.3%
Query 91 31.6 MiB 22.1 MiB -30.0%
Query 92 2.9 MiB 2.9 MiB +0.0%
Query 93 15.4 MiB 15.1 MiB -1.6%
Query 94 3.7 MiB 4.0 MiB +7.6%
Query 95 19.4 MiB 19.4 MiB +0.0%
Query 96 321.0 KiB 224.7 KiB -30.0%
Query 97 33.1 MiB 33.2 MiB +0.2%
Query 98 114.7 MiB 100.4 MiB -12.4%
Query 99 4.1 MiB 8.5 MiB +106.6%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpcds base (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 316.9 MiB 1.8 GiB 1.5 GiB 5.9×
tpcds changed (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 316.9 MiB 1.9 GiB 1.6 GiB 6.2×
Resource Usage

tpcds — base (merge-base)

Metric Value
Wall time 60.0s
Peak memory 1.8 GiB
Avg memory 1.3 GiB
CPU user 253.1s
CPU sys 6.1s
Peak spill 0 B

tpcds — branch

Metric Value
Wall time 60.0s
Peak memory 1.9 GiB
Avg memory 1.4 GiB
CPU user 254.8s
CPU sys 6.5s
Peak spill 0 B

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 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "false"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query     ┃     HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃    Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 1  │ 40.26 ms │                                    40.01 ms │ no change │
│ QQuery 2  │ 17.97 ms │                                    18.10 ms │ no change │
│ QQuery 3  │ 27.97 ms │                                    27.93 ms │ no change │
│ QQuery 4  │ 17.30 ms │                                    17.00 ms │ no change │
│ QQuery 5  │ 34.64 ms │                                    34.37 ms │ no change │
│ QQuery 6  │ 13.57 ms │                                    13.47 ms │ no change │
│ QQuery 7  │ 40.42 ms │                                    40.60 ms │ no change │
│ QQuery 8  │ 39.83 ms │                                    40.01 ms │ no change │
│ QQuery 9  │ 48.15 ms │                                    49.54 ms │ no change │
│ QQuery 10 │ 39.72 ms │                                    40.05 ms │ no change │
│ QQuery 11 │ 13.28 ms │                                    13.09 ms │ no change │
│ QQuery 12 │ 17.55 ms │                                    17.66 ms │ no change │
│ QQuery 13 │ 38.91 ms │                                    39.01 ms │ no change │
│ QQuery 14 │ 24.04 ms │                                    23.94 ms │ no change │
│ QQuery 15 │ 30.42 ms │                                    30.46 ms │ no change │
│ QQuery 16 │ 13.30 ms │                                    13.39 ms │ no change │
│ QQuery 17 │ 66.32 ms │                                    66.64 ms │ no change │
│ QQuery 18 │ 39.67 ms │                                    40.22 ms │ no change │
│ QQuery 19 │ 31.46 ms │                                    31.58 ms │ no change │
│ QQuery 20 │ 31.06 ms │                                    31.65 ms │ no change │
│ QQuery 21 │ 54.53 ms │                                    54.34 ms │ no change │
│ QQuery 22 │ 13.51 ms │                                    13.24 ms │ no change │
└───────────┴──────────┴─────────────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 693.90ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 696.29ms │
│ Average Time (HEAD)                                        │  31.54ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │  31.65ms │
│ Queries Faster                                             │        0 │
│ Queries Slower                                             │        0 │
│ Queries with No Change                                     │       22 │
│ Queries with Failure                                       │        0 │
└────────────────────────────────────────────────────────────┴──────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃    Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ QQuery 1  │ 40.26 / 41.63 ±2.34 / 46.29 ms │              40.01 / 41.63 ±1.31 / 43.24 ms │ no change │
│ QQuery 2  │ 17.97 / 18.26 ±0.25 / 18.68 ms │              18.10 / 18.37 ±0.24 / 18.78 ms │ no change │
│ QQuery 3  │ 27.97 / 28.25 ±0.30 / 28.82 ms │              27.93 / 28.21 ±0.31 / 28.77 ms │ no change │
│ QQuery 4  │ 17.30 / 17.95 ±0.62 / 18.82 ms │              17.00 / 17.33 ±0.22 / 17.66 ms │ no change │
│ QQuery 5  │ 34.64 / 34.92 ±0.17 / 35.09 ms │              34.37 / 34.69 ±0.22 / 34.98 ms │ no change │
│ QQuery 6  │ 13.57 / 14.12 ±0.69 / 15.48 ms │              13.47 / 13.91 ±0.39 / 14.42 ms │ no change │
│ QQuery 7  │ 40.42 / 41.45 ±1.19 / 43.63 ms │              40.60 / 42.38 ±1.23 / 43.97 ms │ no change │
│ QQuery 8  │ 39.83 / 40.70 ±1.32 / 43.32 ms │              40.01 / 41.02 ±1.47 / 43.93 ms │ no change │
│ QQuery 9  │ 48.15 / 49.51 ±1.24 / 51.58 ms │              49.54 / 50.47 ±0.86 / 51.64 ms │ no change │
│ QQuery 10 │ 39.72 / 39.93 ±0.21 / 40.31 ms │              40.05 / 40.40 ±0.55 / 41.49 ms │ no change │
│ QQuery 11 │ 13.28 / 13.64 ±0.30 / 14.17 ms │              13.09 / 13.52 ±0.44 / 14.33 ms │ no change │
│ QQuery 12 │ 17.55 / 18.23 ±0.67 / 19.45 ms │              17.66 / 18.08 ±0.51 / 19.03 ms │ no change │
│ QQuery 13 │ 38.91 / 39.23 ±0.21 / 39.50 ms │              39.01 / 39.27 ±0.17 / 39.45 ms │ no change │
│ QQuery 14 │ 24.04 / 25.23 ±1.29 / 26.96 ms │              23.94 / 25.14 ±2.03 / 29.17 ms │ no change │
│ QQuery 15 │ 30.42 / 30.98 ±0.63 / 32.17 ms │              30.46 / 31.30 ±0.74 / 32.36 ms │ no change │
│ QQuery 16 │ 13.30 / 13.75 ±0.40 / 14.50 ms │              13.39 / 13.52 ±0.10 / 13.69 ms │ no change │
│ QQuery 17 │ 66.32 / 67.32 ±1.61 / 70.52 ms │              66.64 / 67.81 ±1.01 / 69.70 ms │ no change │
│ QQuery 18 │ 39.67 / 40.73 ±0.83 / 41.90 ms │              40.22 / 40.47 ±0.24 / 40.84 ms │ no change │
│ QQuery 19 │ 31.46 / 32.08 ±0.81 / 33.67 ms │              31.58 / 32.41 ±1.46 / 35.31 ms │ no change │
│ QQuery 20 │ 31.06 / 31.62 ±0.41 / 32.32 ms │              31.65 / 31.83 ±0.18 / 32.07 ms │ no change │
│ QQuery 21 │ 54.53 / 55.75 ±0.96 / 57.13 ms │              54.34 / 55.35 ±0.92 / 57.02 ms │ no change │
│ QQuery 22 │ 13.51 / 13.64 ±0.14 / 13.87 ms │              13.24 / 13.41 ±0.17 / 13.71 ms │ no change │
└───────────┴────────────────────────────────┴─────────────────────────────────────────────┴───────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 708.94ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 710.52ms │
│ Average Time (HEAD)                                        │  32.22ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │  32.30ms │
│ Queries Faster                                             │        0 │
│ Queries Slower                                             │        0 │
│ Queries with No Change                                     │       22 │
│ Queries with Failure                                       │        0 │
└────────────────────────────────────────────────────────────┴──────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca | Changed: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca

tpch — tpch_sf1

Query Base Changed Change
Query 1 31.4 MiB 31.4 MiB -0.0%
Query 2 15.9 MiB 15.9 MiB +0.0%
Query 3 10.6 MiB 10.6 MiB +0.0%
Query 4 32.7 MiB 23.4 MiB -28.6%
Query 5 53.1 MiB 43.2 MiB -18.8%
Query 6 832 B 832 B +0.0%
Query 7 105.9 MiB 103.1 MiB -2.7%
Query 8 39.8 MiB 39.9 MiB +0.1%
Query 9 167.5 MiB 140.6 MiB -16.0%
Query 10 40.4 MiB 38.7 MiB -4.2%
Query 11 51.4 MiB 41.9 MiB -18.4%
Query 12 21.7 MiB 22.7 MiB +4.7%
Query 13 52.6 MiB 72.3 MiB +37.5%
Query 14 6.7 MiB 6.7 MiB +0.0%
Query 15 11.9 MiB 12.2 MiB +2.6%
Query 16 43.8 MiB 44.8 MiB +2.3%
Query 17 150.3 MiB 150.3 MiB +0.0%
Query 18 167.1 MiB 167.1 MiB +0.0%
Query 19 725.7 KiB 725.7 KiB +0.0%
Query 20 59.8 MiB 57.6 MiB -3.7%
Query 21 42.2 MiB 42.2 MiB +0.0%
Query 22 32.3 MiB 31.8 MiB -1.7%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpch base (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 167.5 MiB 1.3 GiB 1.1 GiB 7.7×
tpch changed (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 167.1 MiB 1.2 GiB 1.0 GiB 7.4×
Resource Usage

tpch — base (merge-base)

Metric Value
Wall time 5.0s
Peak memory 1.3 GiB
Avg memory 499.2 MiB
CPU user 18.9s
CPU sys 1.6s
Peak spill 0 B

tpch — branch

Metric Value
Wall time 5.0s
Peak memory 1.2 GiB
Avg memory 489.1 MiB
CPU user 18.9s
CPU sys 1.6s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "false"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃     HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 78.18 ms │                                    73.82 ms │ +1.06x faster │
│ QQuery 2  │ 20.60 ms │                                    21.38 ms │     no change │
│ QQuery 3  │ 42.63 ms │                                    43.28 ms │     no change │
│ QQuery 4  │ 25.39 ms │                                    25.29 ms │     no change │
│ QQuery 5  │ 51.05 ms │                                    50.58 ms │     no change │
│ QQuery 6  │ 22.45 ms │                                    21.54 ms │     no change │
│ QQuery 7  │ 60.88 ms │                                    61.71 ms │     no change │
│ QQuery 8  │ 58.29 ms │                                    57.94 ms │     no change │
│ QQuery 9  │ 78.28 ms │                                    77.00 ms │     no change │
│ QQuery 10 │ 56.83 ms │                                    56.18 ms │     no change │
│ QQuery 11 │ 15.93 ms │                                    15.53 ms │     no change │
│ QQuery 12 │ 28.99 ms │                                    28.48 ms │     no change │
│ QQuery 13 │ 48.80 ms │                                    48.30 ms │     no change │
│ QQuery 14 │ 35.68 ms │                                    36.31 ms │     no change │
│ QQuery 15 │ 47.67 ms │                                    48.78 ms │     no change │
│ QQuery 16 │ 16.12 ms │                                    16.76 ms │     no change │
│ QQuery 17 │ 92.99 ms │                                    93.85 ms │     no change │
│ QQuery 18 │ 64.32 ms │                                    64.10 ms │     no change │
│ QQuery 19 │ 48.55 ms │                                    49.57 ms │     no change │
│ QQuery 20 │ 44.13 ms │                                    44.55 ms │     no change │
│ QQuery 21 │ 90.36 ms │                                    83.58 ms │ +1.08x faster │
│ QQuery 22 │ 17.06 ms │                                    15.71 ms │ +1.09x faster │
└───────────┴──────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 1045.20ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 1034.24ms │
│ Average Time (HEAD)                                        │   47.51ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   47.01ms │
│ Queries Faster                                             │         3 │
│ Queries Slower                                             │         0 │
│ Queries with No Change                                     │        19 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 78.18 / 81.21 ±3.33 / 87.45 ms │              73.82 / 76.78 ±3.55 / 83.65 ms │ +1.06x faster │
│ QQuery 2  │ 20.60 / 22.19 ±1.57 / 24.30 ms │              21.38 / 23.05 ±1.44 / 25.66 ms │     no change │
│ QQuery 3  │ 42.63 / 44.61 ±1.05 / 45.54 ms │              43.28 / 44.74 ±1.16 / 46.71 ms │     no change │
│ QQuery 4  │ 25.39 / 26.40 ±0.70 / 27.57 ms │              25.29 / 26.00 ±0.87 / 27.66 ms │     no change │
│ QQuery 5  │ 51.05 / 53.12 ±3.57 / 60.24 ms │              50.58 / 52.62 ±1.30 / 54.41 ms │     no change │
│ QQuery 6  │ 22.45 / 23.05 ±0.67 / 24.34 ms │              21.54 / 23.22 ±1.04 / 24.61 ms │     no change │
│ QQuery 7  │ 60.88 / 63.71 ±3.44 / 70.25 ms │              61.71 / 63.87 ±1.38 / 66.07 ms │     no change │
│ QQuery 8  │ 58.29 / 59.91 ±1.17 / 61.20 ms │              57.94 / 60.30 ±1.69 / 62.26 ms │     no change │
│ QQuery 9  │ 78.28 / 79.79 ±1.12 / 81.69 ms │              77.00 / 79.14 ±1.87 / 82.22 ms │     no change │
│ QQuery 10 │ 56.83 / 60.33 ±2.70 / 63.40 ms │              56.18 / 58.39 ±2.38 / 63.01 ms │     no change │
│ QQuery 11 │ 15.93 / 17.25 ±1.14 / 19.28 ms │              15.53 / 16.47 ±0.86 / 17.61 ms │     no change │
│ QQuery 12 │ 28.99 / 30.83 ±2.64 / 36.03 ms │              28.48 / 29.39 ±0.79 / 30.74 ms │     no change │
│ QQuery 13 │ 48.80 / 51.26 ±2.15 / 55.08 ms │              48.30 / 51.32 ±3.16 / 55.91 ms │     no change │
│ QQuery 14 │ 35.68 / 36.69 ±0.89 / 38.22 ms │              36.31 / 37.03 ±0.46 / 37.60 ms │     no change │
│ QQuery 15 │ 47.67 / 48.42 ±0.85 / 49.94 ms │              48.78 / 49.82 ±0.78 / 51.14 ms │     no change │
│ QQuery 16 │ 16.12 / 16.87 ±0.77 / 18.33 ms │              16.76 / 17.25 ±0.45 / 17.98 ms │     no change │
│ QQuery 17 │ 92.99 / 95.22 ±1.64 / 97.37 ms │              93.85 / 95.76 ±1.42 / 97.53 ms │     no change │
│ QQuery 18 │ 64.32 / 65.63 ±1.17 / 67.10 ms │              64.10 / 65.76 ±1.11 / 66.89 ms │     no change │
│ QQuery 19 │ 48.55 / 50.77 ±1.41 / 52.44 ms │              49.57 / 53.21 ±2.07 / 54.82 ms │     no change │
│ QQuery 20 │ 44.13 / 46.42 ±2.24 / 49.65 ms │              44.55 / 45.97 ±1.84 / 49.57 ms │     no change │
│ QQuery 21 │ 90.36 / 91.87 ±1.27 / 93.79 ms │              83.58 / 87.25 ±1.93 / 88.87 ms │ +1.05x faster │
│ QQuery 22 │ 17.06 / 17.85 ±0.64 / 18.48 ms │              15.71 / 17.33 ±0.97 / 18.58 ms │     no change │
└───────────┴────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 1083.41ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 1074.66ms │
│ Average Time (HEAD)                                        │   49.25ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   48.85ms │
│ Queries Faster                                             │         2 │
│ Queries Slower                                             │         0 │
│ Queries with No Change                                     │        20 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca | Changed: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca

tpch — tpch_sf1

Query Base Changed Change
Query 1 30.0 MiB 21.4 MiB -28.7%
Query 2 15.9 MiB 15.9 MiB +0.0%
Query 3 11.1 MiB 11.1 MiB +0.0%
Query 4 23.5 MiB 23.6 MiB +0.3%
Query 5 42.7 MiB 33.6 MiB -21.2%
Query 6 832 B 832 B +0.0%
Query 7 110.0 MiB 104.7 MiB -4.8%
Query 8 39.8 MiB 39.8 MiB -0.0%
Query 9 142.1 MiB 142.2 MiB +0.1%
Query 10 42.1 MiB 43.9 MiB +4.2%
Query 11 41.5 MiB 41.5 MiB +0.0%
Query 12 22.7 MiB 22.7 MiB -0.0%
Query 13 43.2 MiB 82.6 MiB +91.1%
Query 14 6.7 MiB 6.7 MiB +0.0%
Query 15 12.1 MiB 13.0 MiB +7.9%
Query 16 45.3 MiB 44.8 MiB -1.1%
Query 17 153.5 MiB 153.6 MiB +0.1%
Query 18 167.1 MiB 167.1 MiB +0.0%
Query 19 725.7 KiB 725.7 KiB +0.0%
Query 20 58.9 MiB 58.8 MiB -0.1%
Query 21 42.4 MiB 42.4 MiB +0.0%
Query 22 41.8 MiB 31.9 MiB -23.7%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpch base (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 167.1 MiB 1.1 GiB 948.4 MiB 6.7×
tpch changed (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 167.1 MiB 1.1 GiB 956.1 MiB 6.7×
Resource Usage

tpch — base (merge-base)

Metric Value
Wall time 10.0s
Peak memory 1.1 GiB
Avg memory 419.1 MiB
CPU user 34.9s
CPU sys 2.0s
Peak spill 0 B

tpch — branch

Metric Value
Wall time 10.0s
Peak memory 1.1 GiB
Avg memory 423.4 MiB
CPU user 34.9s
CPU sys 1.9s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing 432586c (432586c) to 432586c diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "false"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
changed:
  ref: "432586c6f82b2a7f96933dc0501cf02f2a2a1eca"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "5"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │    4.86 ms │                                     4.69 ms │     no change │
│ QQuery 2  │   76.35 ms │                                    76.38 ms │     no change │
│ QQuery 3  │   28.64 ms │                                    29.30 ms │     no change │
│ QQuery 4  │  422.35 ms │                                   409.31 ms │     no change │
│ QQuery 5  │   48.61 ms │                                    47.45 ms │     no change │
│ QQuery 6  │   33.71 ms │                                    33.73 ms │     no change │
│ QQuery 7  │   72.71 ms │                                    72.70 ms │     no change │
│ QQuery 8  │   35.12 ms │                                    34.09 ms │     no change │
│ QQuery 9  │   70.29 ms │                                    67.10 ms │     no change │
│ QQuery 10 │   56.80 ms │                                    55.30 ms │     no change │
│ QQuery 11 │  239.99 ms │                                   253.41 ms │  1.06x slower │
│ QQuery 12 │   29.56 ms │                                    29.68 ms │     no change │
│ QQuery 13 │  111.35 ms │                                   114.77 ms │     no change │
│ QQuery 14 │  437.89 ms │                                   418.64 ms │     no change │
│ QQuery 15 │   46.31 ms │                                    45.59 ms │     no change │
│ QQuery 16 │    4.90 ms │                                     5.19 ms │  1.06x slower │
│ QQuery 17 │   84.22 ms │                                    81.07 ms │     no change │
│ QQuery 18 │   99.98 ms │                                    98.82 ms │     no change │
│ QQuery 19 │   40.27 ms │                                    41.61 ms │     no change │
│ QQuery 20 │   35.60 ms │                                    33.26 ms │ +1.07x faster │
│ QQuery 21 │   19.87 ms │                                    19.20 ms │     no change │
│ QQuery 22 │   92.78 ms │                                    93.82 ms │     no change │
│ QQuery 23 │  332.43 ms │                                   339.94 ms │     no change │
│ QQuery 24 │  221.06 ms │                                   212.23 ms │     no change │
│ QQuery 25 │  119.33 ms │                                   115.62 ms │     no change │
│ QQuery 26 │   45.01 ms │                                    45.05 ms │     no change │
│ QQuery 27 │    4.00 ms │                                     4.25 ms │  1.06x slower │
│ QQuery 28 │   70.96 ms │                                    74.23 ms │     no change │
│ QQuery 29 │  107.13 ms │                                   114.66 ms │  1.07x slower │
│ QQuery 30 │   30.59 ms │                                    30.94 ms │     no change │
│ QQuery 31 │  107.92 ms │                                   105.96 ms │     no change │
│ QQuery 32 │   20.37 ms │                                    19.27 ms │ +1.06x faster │
│ QQuery 33 │   35.52 ms │                                    36.42 ms │     no change │
│ QQuery 34 │    4.60 ms │                                     4.60 ms │     no change │
│ QQuery 35 │   65.72 ms │                                    66.29 ms │     no change │
│ QQuery 36 │    5.13 ms │                                     4.83 ms │ +1.06x faster │
│ QQuery 37 │    5.97 ms │                                     6.53 ms │  1.09x slower │
│ QQuery 38 │   60.51 ms │                                    61.52 ms │     no change │
│ QQuery 39 │   79.07 ms │                                    75.73 ms │     no change │
│ QQuery 40 │   24.46 ms │                                    23.36 ms │     no change │
│ QQuery 41 │   11.67 ms │                                    11.60 ms │     no change │
│ QQuery 42 │   24.27 ms │                                    23.59 ms │     no change │
│ QQuery 43 │    3.86 ms │                                     3.80 ms │     no change │
│ QQuery 44 │    4.82 ms │                                     4.82 ms │     no change │
│ QQuery 45 │   33.47 ms │                                    32.66 ms │     no change │
│ QQuery 46 │   10.94 ms │                                    12.10 ms │  1.11x slower │
│ QQuery 47 │  224.63 ms │                                   234.94 ms │     no change │
│ QQuery 48 │   92.88 ms │                                    94.63 ms │     no change │
│ QQuery 49 │   76.68 ms │                                    71.44 ms │ +1.07x faster │
│ QQuery 50 │   61.39 ms │                                    62.65 ms │     no change │
│ QQuery 51 │   99.65 ms │                                   101.25 ms │     no change │
│ QQuery 52 │   24.44 ms │                                    25.22 ms │     no change │
│ QQuery 53 │   29.59 ms │                                    29.60 ms │     no change │
│ QQuery 54 │   49.45 ms │                                    49.86 ms │     no change │
│ QQuery 55 │   24.35 ms │                                    22.81 ms │ +1.07x faster │
│ QQuery 56 │   38.30 ms │                                    36.10 ms │ +1.06x faster │
│ QQuery 57 │  164.71 ms │                                   167.55 ms │     no change │
│ QQuery 58 │  100.21 ms │                                   102.78 ms │     no change │
│ QQuery 59 │  116.97 ms │                                   115.11 ms │     no change │
│ QQuery 60 │   38.12 ms │                                    38.34 ms │     no change │
│ QQuery 61 │    7.26 ms │                                     7.08 ms │     no change │
│ QQuery 62 │   43.29 ms │                                    42.64 ms │     no change │
│ QQuery 63 │   31.19 ms │                                    30.92 ms │     no change │
│ QQuery 64 │  377.28 ms │                                   368.55 ms │     no change │
│ QQuery 65 │   70.26 ms │                                    73.90 ms │  1.05x slower │
│ QQuery 66 │   68.75 ms │                                    69.39 ms │     no change │
│ QQuery 67 │  230.30 ms │                                   225.58 ms │     no change │
│ QQuery 68 │    5.72 ms │                                     5.52 ms │     no change │
│ QQuery 69 │   52.39 ms │                                    52.88 ms │     no change │
│ QQuery 70 │  103.54 ms │                                   100.66 ms │     no change │
│ QQuery 71 │   34.98 ms │                                    34.12 ms │     no change │
│ QQuery 72 │ 1960.96 ms │                                  1981.23 ms │     no change │
│ QQuery 73 │    4.43 ms │                                     4.40 ms │     no change │
│ QQuery 74 │  174.73 ms │                                   174.93 ms │     no change │
│ QQuery 75 │  149.71 ms │                                   142.89 ms │     no change │
│ QQuery 76 │   35.26 ms │                                    37.07 ms │  1.05x slower │
│ QQuery 77 │   57.17 ms │                                    57.60 ms │     no change │
│ QQuery 78 │  336.50 ms │                                   323.23 ms │     no change │
│ QQuery 79 │   65.20 ms │                                    67.24 ms │     no change │
│ QQuery 80 │  107.23 ms │                                   103.98 ms │     no change │
│ QQuery 81 │   24.22 ms │                                    24.33 ms │     no change │
│ QQuery 82 │   16.62 ms │                                    16.50 ms │     no change │
│ QQuery 83 │   29.92 ms │                                    28.82 ms │     no change │
│ QQuery 84 │   26.69 ms │                                    27.48 ms │     no change │
│ QQuery 85 │   95.96 ms │                                    99.25 ms │     no change │
│ QQuery 86 │   25.54 ms │                                    25.19 ms │     no change │
│ QQuery 87 │   60.80 ms │                                    60.95 ms │     no change │
│ QQuery 88 │   71.74 ms │                                    71.93 ms │     no change │
│ QQuery 89 │   35.07 ms │                                    35.43 ms │     no change │
│ QQuery 90 │   15.59 ms │                                    15.47 ms │     no change │
│ QQuery 91 │   39.01 ms │                                    38.90 ms │     no change │
│ QQuery 92 │   31.80 ms │                                    31.56 ms │     no change │
│ QQuery 93 │   57.49 ms │                                    54.08 ms │ +1.06x faster │
│ QQuery 94 │   37.35 ms │                                    37.21 ms │     no change │
│ QQuery 95 │   90.07 ms │                                    83.57 ms │ +1.08x faster │
│ QQuery 96 │   23.09 ms │                                    22.97 ms │     no change │
│ QQuery 97 │   51.62 ms │                                    54.51 ms │  1.06x slower │
│ QQuery 98 │   42.70 ms │                                    43.47 ms │     no change │
│ QQuery 99 │   64.31 ms │                                    63.45 ms │     no change │
└───────────┴────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 9418.09ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 9386.29ms │
│ Average Time (HEAD)                                        │   95.13ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   94.81ms │
│ Queries Faster                                             │         8 │
│ Queries Slower                                             │         9 │
│ Queries with No Change                                     │        82 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │           4.86 / 6.06 ±1.83 / 9.61 ms │                 4.69 / 6.03 ±1.75 / 9.48 ms │     no change │
│ QQuery 2  │        76.35 / 83.05 ±8.34 / 99.12 ms │              76.38 / 80.28 ±3.06 / 85.49 ms │     no change │
│ QQuery 3  │        28.64 / 31.13 ±2.86 / 36.59 ms │              29.30 / 35.70 ±4.91 / 41.54 ms │  1.15x slower │
│ QQuery 4  │     422.35 / 432.47 ±7.39 / 440.19 ms │           409.31 / 426.69 ±9.25 / 435.22 ms │     no change │
│ QQuery 5  │        48.61 / 53.46 ±6.22 / 65.51 ms │              47.45 / 54.22 ±3.47 / 57.40 ms │     no change │
│ QQuery 6  │        33.71 / 34.48 ±0.57 / 35.32 ms │              33.73 / 36.74 ±3.54 / 43.60 ms │  1.07x slower │
│ QQuery 7  │        72.71 / 77.48 ±6.51 / 90.28 ms │              72.70 / 77.98 ±7.01 / 91.04 ms │     no change │
│ QQuery 8  │        35.12 / 39.18 ±3.23 / 43.13 ms │              34.09 / 35.42 ±0.91 / 36.53 ms │ +1.11x faster │
│ QQuery 9  │        70.29 / 72.65 ±2.13 / 76.31 ms │              67.10 / 72.35 ±3.70 / 78.63 ms │     no change │
│ QQuery 10 │        56.80 / 60.07 ±4.02 / 65.91 ms │              55.30 / 56.91 ±1.01 / 58.10 ms │ +1.06x faster │
│ QQuery 11 │    239.99 / 251.79 ±13.41 / 277.09 ms │           253.41 / 263.72 ±5.41 / 268.49 ms │     no change │
│ QQuery 12 │        29.56 / 30.35 ±0.77 / 31.78 ms │              29.68 / 29.98 ±0.31 / 30.44 ms │     no change │
│ QQuery 13 │     111.35 / 115.05 ±3.34 / 120.33 ms │           114.77 / 126.17 ±7.50 / 134.76 ms │  1.10x slower │
│ QQuery 14 │    437.89 / 462.41 ±28.82 / 518.36 ms │          418.64 / 459.26 ±25.99 / 492.82 ms │     no change │
│ QQuery 15 │        46.31 / 50.71 ±4.65 / 57.94 ms │              45.59 / 49.18 ±4.23 / 56.66 ms │     no change │
│ QQuery 16 │           4.90 / 5.31 ±0.55 / 6.39 ms │                 5.19 / 5.36 ±0.20 / 5.75 ms │     no change │
│ QQuery 17 │       84.22 / 92.44 ±7.12 / 102.04 ms │             81.07 / 92.37 ±8.98 / 103.11 ms │     no change │
│ QQuery 18 │      99.98 / 103.50 ±3.73 / 110.67 ms │            98.82 / 100.21 ±1.16 / 101.71 ms │     no change │
│ QQuery 19 │        40.27 / 46.43 ±6.99 / 57.51 ms │              41.61 / 45.29 ±5.44 / 56.05 ms │     no change │
│ QQuery 20 │        35.60 / 41.22 ±3.39 / 44.45 ms │              33.26 / 34.82 ±1.04 / 36.12 ms │ +1.18x faster │
│ QQuery 21 │        19.87 / 21.11 ±0.77 / 22.02 ms │              19.20 / 20.64 ±1.12 / 22.34 ms │     no change │
│ QQuery 22 │       92.78 / 96.74 ±3.17 / 101.91 ms │              93.82 / 95.77 ±1.15 / 97.17 ms │     no change │
│ QQuery 23 │    332.43 / 353.93 ±19.88 / 385.70 ms │          339.94 / 354.13 ±13.93 / 380.03 ms │     no change │
│ QQuery 24 │     221.06 / 232.61 ±6.90 / 241.77 ms │          212.23 / 224.06 ±11.73 / 244.69 ms │     no change │
│ QQuery 25 │     119.33 / 123.32 ±2.15 / 125.66 ms │           115.62 / 119.34 ±2.36 / 123.05 ms │     no change │
│ QQuery 26 │        45.01 / 46.83 ±1.21 / 48.60 ms │              45.05 / 46.52 ±1.13 / 48.19 ms │     no change │
│ QQuery 27 │           4.00 / 4.31 ±0.25 / 4.73 ms │                 4.25 / 4.31 ±0.04 / 4.37 ms │     no change │
│ QQuery 28 │        70.96 / 81.30 ±7.36 / 91.49 ms │              74.23 / 76.87 ±2.63 / 81.51 ms │ +1.06x faster │
│ QQuery 29 │     107.13 / 116.72 ±6.23 / 124.44 ms │           114.66 / 117.23 ±3.44 / 123.81 ms │     no change │
│ QQuery 30 │        30.59 / 31.56 ±0.82 / 32.87 ms │              30.94 / 32.95 ±1.54 / 35.55 ms │     no change │
│ QQuery 31 │    107.92 / 121.68 ±10.31 / 135.57 ms │           105.96 / 112.73 ±5.45 / 121.76 ms │ +1.08x faster │
│ QQuery 32 │        20.37 / 21.50 ±0.80 / 22.84 ms │              19.27 / 21.48 ±1.63 / 23.97 ms │     no change │
│ QQuery 33 │        35.52 / 37.86 ±2.56 / 42.80 ms │              36.42 / 39.25 ±3.54 / 46.20 ms │     no change │
│ QQuery 34 │           4.60 / 4.80 ±0.18 / 5.11 ms │                 4.60 / 4.82 ±0.21 / 5.17 ms │     no change │
│ QQuery 35 │        65.72 / 71.13 ±4.29 / 76.52 ms │              66.29 / 72.80 ±6.08 / 83.04 ms │     no change │
│ QQuery 36 │           5.13 / 5.26 ±0.11 / 5.43 ms │                 4.83 / 5.11 ±0.27 / 5.53 ms │     no change │
│ QQuery 37 │           5.97 / 6.25 ±0.24 / 6.62 ms │                 6.53 / 6.66 ±0.17 / 6.98 ms │  1.07x slower │
│ QQuery 38 │        60.51 / 65.28 ±4.05 / 70.50 ms │              61.52 / 64.03 ±2.54 / 68.75 ms │     no change │
│ QQuery 39 │        79.07 / 81.83 ±1.60 / 83.54 ms │              75.73 / 80.25 ±2.84 / 84.34 ms │     no change │
│ QQuery 40 │        24.46 / 26.66 ±1.79 / 29.33 ms │              23.36 / 24.29 ±0.90 / 25.36 ms │ +1.10x faster │
│ QQuery 41 │        11.67 / 12.07 ±0.42 / 12.74 ms │              11.60 / 12.39 ±1.12 / 14.60 ms │     no change │
│ QQuery 42 │        24.27 / 26.50 ±2.34 / 30.85 ms │              23.59 / 25.89 ±2.47 / 30.62 ms │     no change │
│ QQuery 43 │           3.86 / 4.20 ±0.22 / 4.52 ms │                 3.80 / 3.91 ±0.14 / 4.19 ms │ +1.07x faster │
│ QQuery 44 │           4.82 / 4.93 ±0.14 / 5.16 ms │                 4.82 / 5.09 ±0.24 / 5.46 ms │     no change │
│ QQuery 45 │        33.47 / 35.88 ±2.64 / 41.00 ms │              32.66 / 34.06 ±0.79 / 35.08 ms │ +1.05x faster │
│ QQuery 46 │        10.94 / 11.89 ±0.78 / 12.98 ms │              12.10 / 12.84 ±1.10 / 15.02 ms │  1.08x slower │
│ QQuery 47 │    224.63 / 244.85 ±12.56 / 255.68 ms │           234.94 / 246.29 ±8.02 / 259.48 ms │     no change │
│ QQuery 48 │       92.88 / 97.92 ±6.42 / 110.54 ms │             94.63 / 98.37 ±6.00 / 110.18 ms │     no change │
│ QQuery 49 │        76.68 / 83.75 ±4.64 / 88.75 ms │              71.44 / 82.66 ±7.05 / 93.59 ms │     no change │
│ QQuery 50 │        61.39 / 65.24 ±3.93 / 72.13 ms │              62.65 / 68.40 ±5.48 / 78.05 ms │     no change │
│ QQuery 51 │      99.65 / 105.65 ±4.87 / 114.49 ms │           101.25 / 106.58 ±5.84 / 117.98 ms │     no change │
│ QQuery 52 │        24.44 / 27.15 ±3.07 / 32.98 ms │              25.22 / 25.95 ±0.88 / 27.65 ms │     no change │
│ QQuery 53 │        29.59 / 32.38 ±3.16 / 38.50 ms │              29.60 / 30.47 ±0.60 / 31.46 ms │ +1.06x faster │
│ QQuery 54 │        49.45 / 50.91 ±1.10 / 51.94 ms │              49.86 / 52.90 ±3.28 / 58.89 ms │     no change │
│ QQuery 55 │        24.35 / 26.83 ±3.76 / 34.32 ms │              22.81 / 24.85 ±1.10 / 26.02 ms │ +1.08x faster │
│ QQuery 56 │        38.30 / 46.41 ±4.41 / 50.13 ms │              36.10 / 37.54 ±0.94 / 39.04 ms │ +1.24x faster │
│ QQuery 57 │     164.71 / 173.27 ±5.63 / 182.09 ms │           167.55 / 176.73 ±6.04 / 185.98 ms │     no change │
│ QQuery 58 │     100.21 / 112.05 ±7.79 / 120.35 ms │           102.78 / 109.06 ±5.10 / 117.08 ms │     no change │
│ QQuery 59 │    116.97 / 130.70 ±18.76 / 165.17 ms │          115.11 / 125.40 ±12.25 / 143.79 ms │     no change │
│ QQuery 60 │        38.12 / 43.20 ±3.43 / 47.95 ms │              38.34 / 43.62 ±3.88 / 48.83 ms │     no change │
│ QQuery 61 │           7.26 / 7.61 ±0.30 / 8.16 ms │                 7.08 / 7.27 ±0.14 / 7.52 ms │     no change │
│ QQuery 62 │        43.29 / 43.57 ±0.20 / 43.87 ms │              42.64 / 44.04 ±1.04 / 45.41 ms │     no change │
│ QQuery 63 │        31.19 / 34.84 ±3.32 / 40.31 ms │              30.92 / 33.23 ±4.22 / 41.67 ms │     no change │
│ QQuery 64 │    377.28 / 400.36 ±17.58 / 423.18 ms │          368.55 / 385.32 ±11.27 / 404.01 ms │     no change │
│ QQuery 65 │        70.26 / 75.24 ±5.27 / 84.79 ms │              73.90 / 77.50 ±3.94 / 84.83 ms │     no change │
│ QQuery 66 │        68.75 / 73.36 ±5.49 / 83.10 ms │              69.39 / 71.32 ±1.43 / 72.96 ms │     no change │
│ QQuery 67 │    230.30 / 240.67 ±10.05 / 257.21 ms │           225.58 / 232.34 ±5.33 / 239.03 ms │     no change │
│ QQuery 68 │           5.72 / 6.30 ±0.56 / 7.26 ms │                 5.52 / 6.09 ±0.54 / 7.10 ms │     no change │
│ QQuery 69 │        52.39 / 58.63 ±5.55 / 68.75 ms │              52.88 / 55.37 ±1.55 / 57.63 ms │ +1.06x faster │
│ QQuery 70 │    103.54 / 111.88 ±11.20 / 132.38 ms │           100.66 / 112.07 ±7.50 / 121.57 ms │     no change │
│ QQuery 71 │        34.98 / 38.56 ±3.50 / 43.41 ms │              34.12 / 40.95 ±3.70 / 44.54 ms │  1.06x slower │
│ QQuery 72 │ 1960.96 / 2014.96 ±44.41 / 2074.75 ms │       1981.23 / 2007.40 ±26.84 / 2047.05 ms │     no change │
│ QQuery 73 │           4.43 / 4.97 ±0.47 / 5.79 ms │                 4.40 / 4.79 ±0.48 / 5.74 ms │     no change │
│ QQuery 74 │     174.73 / 184.65 ±8.42 / 195.39 ms │          174.93 / 185.00 ±11.88 / 205.14 ms │     no change │
│ QQuery 75 │     149.71 / 157.04 ±6.30 / 165.19 ms │           142.89 / 153.60 ±7.25 / 163.08 ms │     no change │
│ QQuery 76 │        35.26 / 42.05 ±5.28 / 48.85 ms │              37.07 / 39.44 ±2.02 / 42.12 ms │ +1.07x faster │
│ QQuery 77 │        57.17 / 62.05 ±3.56 / 65.64 ms │              57.60 / 61.73 ±5.64 / 72.88 ms │     no change │
│ QQuery 78 │    336.50 / 347.48 ±10.35 / 364.34 ms │          323.23 / 348.27 ±22.05 / 382.50 ms │     no change │
│ QQuery 79 │        65.20 / 67.85 ±2.35 / 71.74 ms │              67.24 / 69.47 ±1.83 / 72.27 ms │     no change │
│ QQuery 80 │     107.23 / 110.25 ±2.23 / 113.41 ms │           103.98 / 110.16 ±5.15 / 117.51 ms │     no change │
│ QQuery 81 │        24.22 / 26.05 ±1.32 / 27.78 ms │              24.33 / 26.06 ±1.14 / 27.35 ms │     no change │
│ QQuery 82 │        16.62 / 17.50 ±0.64 / 18.56 ms │              16.50 / 18.12 ±1.19 / 20.05 ms │     no change │
│ QQuery 83 │        29.92 / 31.92 ±2.64 / 37.07 ms │              28.82 / 30.66 ±0.96 / 31.65 ms │     no change │
│ QQuery 84 │        26.69 / 29.71 ±2.95 / 34.67 ms │              27.48 / 29.18 ±1.33 / 30.95 ms │     no change │
│ QQuery 85 │      95.96 / 101.78 ±4.64 / 109.92 ms │            99.25 / 104.64 ±5.01 / 114.11 ms │     no change │
│ QQuery 86 │        25.54 / 26.75 ±0.98 / 28.20 ms │              25.19 / 25.92 ±0.57 / 26.87 ms │     no change │
│ QQuery 87 │        60.80 / 64.84 ±4.85 / 72.40 ms │              60.95 / 63.90 ±2.14 / 67.16 ms │     no change │
│ QQuery 88 │        71.74 / 73.89 ±1.86 / 77.03 ms │              71.93 / 75.31 ±2.17 / 78.70 ms │     no change │
│ QQuery 89 │        35.07 / 38.68 ±4.49 / 47.52 ms │              35.43 / 38.44 ±3.81 / 45.75 ms │     no change │
│ QQuery 90 │        15.59 / 16.87 ±2.07 / 20.99 ms │              15.47 / 16.89 ±2.00 / 20.77 ms │     no change │
│ QQuery 91 │        39.01 / 41.74 ±3.80 / 48.85 ms │              38.90 / 42.48 ±3.52 / 49.02 ms │     no change │
│ QQuery 92 │        31.80 / 32.09 ±0.31 / 32.60 ms │              31.56 / 34.36 ±1.49 / 35.61 ms │  1.07x slower │
│ QQuery 93 │        57.49 / 59.41 ±1.80 / 62.51 ms │              54.08 / 60.54 ±6.70 / 72.27 ms │     no change │
│ QQuery 94 │        37.35 / 39.72 ±1.54 / 41.18 ms │              37.21 / 37.88 ±0.40 / 38.29 ms │     no change │
│ QQuery 95 │        90.07 / 91.28 ±0.94 / 92.44 ms │              83.57 / 88.16 ±2.75 / 92.11 ms │     no change │
│ QQuery 96 │        23.09 / 25.56 ±3.81 / 33.11 ms │              22.97 / 23.96 ±0.96 / 25.67 ms │ +1.07x faster │
│ QQuery 97 │        51.62 / 59.36 ±7.30 / 71.83 ms │              54.51 / 61.77 ±5.40 / 71.00 ms │     no change │
│ QQuery 98 │        42.70 / 43.63 ±1.06 / 45.67 ms │              43.47 / 49.82 ±5.45 / 56.48 ms │  1.14x slower │
│ QQuery 99 │        64.31 / 65.73 ±0.88 / 67.08 ms │              63.45 / 68.73 ±6.26 / 81.02 ms │     no change │
└───────────┴───────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 9944.48ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 9877.15ms │
│ Average Time (HEAD)                                        │  100.45ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   99.77ms │
│ Queries Faster                                             │        14 │
│ Queries Slower                                             │         8 │
│ Queries with No Change                                     │        77 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca | Changed: 432586c6f82b2a7f96933dc0501cf02f2a2a1eca

tpcds — tpcds_sf1

Query Base Changed Change
Query 1 142.0 KiB 137.5 KiB -3.2%
Query 2 59.5 MiB 67.9 MiB +14.0%
Query 3 4.1 MiB 4.1 MiB +0.0%
Query 4 121.6 MiB 121.6 MiB +0.0%
Query 5 16.0 MiB 16.0 MiB +0.0%
Query 6 7.6 MiB 7.6 MiB -0.4%
Query 7 12.1 MiB 12.1 MiB +0.0%
Query 8 3.8 MiB 3.8 MiB +0.0%
Query 9 32.1 KiB 32.1 KiB +0.0%
Query 10 3.1 MiB 3.1 MiB +0.0%
Query 11 85.1 MiB 85.1 MiB +0.0%
Query 12 50.0 MiB 131.1 MiB +162.4%
Query 13 19.6 MiB 19.6 MiB +0.0%
Query 14 170.9 MiB 170.9 MiB +0.0%
Query 15 7.1 MiB 7.1 MiB +0.0%
Query 16 161.5 KiB 161.5 KiB +0.0%
Query 17 40.5 MiB 40.1 MiB -1.1%
Query 18 24.9 MiB 34.2 MiB +37.4%
Query 19 6.4 MiB 5.5 MiB -14.3%
Query 20 111.2 MiB 110.4 MiB -0.7%
Query 21 5.8 MiB 5.8 MiB +0.0%
Query 22 93.4 MiB 95.5 MiB +2.3%
Query 23 215.6 MiB 216.0 MiB +0.2%
Query 24 45.9 MiB 45.9 MiB +0.0%
Query 25 84.9 MiB 84.6 MiB -0.4%
Query 26 11.8 MiB 11.8 MiB +0.0%
Query 27 987.7 KiB 987.7 KiB +0.0%
Query 28 1.3 MiB 1.6 MiB +19.9%
Query 29 97.4 MiB 97.0 MiB -0.4%
Query 30 7.5 MiB 7.8 MiB +3.8%
Query 31 18.0 MiB 18.0 MiB -0.0%
Query 32 2.9 MiB 2.9 MiB +0.0%
Query 33 4.2 MiB 4.2 MiB +0.0%
Query 34 96.3 KiB 96.3 KiB +0.0%
Query 35 29.0 MiB 26.6 MiB -8.2%
Query 36 10.6 MiB 10.6 MiB +0.0%
Query 37 3.8 MiB 3.8 MiB +0.0%
Query 38 20.1 MiB 20.0 MiB -0.6%
Query 39 33.8 MiB 32.9 MiB -2.6%
Query 40 14.4 MiB 14.4 MiB +0.0%
Query 41 2.1 MiB 2.1 MiB +0.0%
Query 42 2.0 MiB 2.0 MiB +0.0%
Query 43 192.4 KiB 192.4 KiB +0.0%
Query 44 192.4 KiB 192.4 KiB +0.0%
Query 45 8.3 MiB 8.3 MiB +0.0%
Query 46 9.4 MiB 9.4 MiB +0.0%
Query 47 157.3 MiB 157.3 MiB +0.0%
Query 48 16.4 MiB 16.4 MiB -0.0%
Query 49 41.1 MiB 41.1 MiB -0.0%
Query 50 14.8 MiB 14.8 MiB +0.0%
Query 51 177.7 MiB 177.7 MiB +0.0%
Query 52 4.0 MiB 4.0 MiB +0.0%
Query 53 30.4 MiB 20.6 MiB -32.4%
Query 54 4.0 MiB 4.0 MiB +0.0%
Query 55 3.5 MiB 3.5 MiB +0.0%
Query 56 5.0 MiB 5.2 MiB +4.3%
Query 57 140.9 MiB 141.9 MiB +0.6%
Query 58 6.8 MiB 6.8 MiB +0.0%
Query 59 14.9 MiB 14.9 MiB +0.0%
Query 60 7.3 MiB 7.3 MiB -0.1%
Query 61 2.4 KiB 2.3 KiB -3.3%
Query 62 8.4 MiB 8.4 MiB +0.0%
Query 63 30.6 MiB 20.6 MiB -32.7%
Query 64 113.3 MiB 111.1 MiB -1.9%
Query 65 29.4 MiB 29.4 MiB +0.0%
Query 66 16.0 MiB 23.4 MiB +46.5%
Query 67 261.5 MiB 260.4 MiB -0.4%
Query 68 96.3 KiB 96.3 KiB +0.0%
Query 69 47.9 MiB 47.9 MiB +0.0%
Query 70 50.0 MiB 50.0 MiB -0.0%
Query 71 60.5 MiB 51.1 MiB -15.4%
Query 72 106.3 MiB 106.2 MiB -0.1%
Query 73 96.3 KiB 96.3 KiB +0.0%
Query 74 50.0 MiB 50.0 MiB +0.0%
Query 75 21.5 MiB 21.5 MiB +0.0%
Query 76 7.1 MiB 7.1 MiB -0.0%
Query 77 4.4 MiB 4.4 MiB +0.0%
Query 78 316.9 MiB 316.9 MiB -0.0%
Query 79 10.2 MiB 10.2 MiB +0.0%
Query 80 41.9 MiB 42.4 MiB +1.0%
Query 81 11.4 MiB 11.9 MiB +4.2%
Query 82 3.8 MiB 3.8 MiB +0.0%
Query 83 12.3 MiB 11.9 MiB -2.9%
Query 84 12.0 MiB 12.0 MiB +0.0%
Query 85 21.9 MiB 22.2 MiB +1.3%
Query 86 95.2 MiB 105.2 MiB +10.5%
Query 87 20.6 MiB 21.0 MiB +1.8%
Query 88 1.7 MiB 1.5 MiB -10.9%
Query 89 59.0 MiB 69.0 MiB +17.0%
Query 90 348.6 KiB 338.7 KiB -2.8%
Query 91 22.1 MiB 22.9 MiB +3.5%
Query 92 2.9 MiB 2.9 MiB +0.0%
Query 93 15.4 MiB 15.1 MiB -1.6%
Query 94 4.0 MiB 4.0 MiB +0.0%
Query 95 19.4 MiB 19.4 MiB +0.0%
Query 96 224.7 KiB 224.7 KiB +0.0%
Query 97 33.2 MiB 33.1 MiB -0.2%
Query 98 111.4 MiB 110.6 MiB -0.8%
Query 99 8.5 MiB 8.5 MiB +0.0%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpcds base (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 316.9 MiB 2.1 GiB 1.8 GiB 6.9×
tpcds changed (432586c6f82b2a7f96933dc0501cf02f2a2a1eca) 316.9 MiB 2.0 GiB 1.7 GiB 6.5×
Resource Usage

tpcds — base (merge-base)

Metric Value
Wall time 55.0s
Peak memory 2.1 GiB
Avg memory 1.4 GiB
CPU user 217.8s
CPU sys 5.3s
Peak spill 0 B

tpcds — branch

Metric Value
Wall time 50.0s
Peak memory 2.0 GiB
Avg memory 1.5 GiB
CPU user 218.5s
CPU sys 5.4s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark tpch tpcds clickbench_partitioned

env:
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: arm64

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

run benchmark tpch tpcds clickbench_partitioned

env:
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: amd64

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991648359-5452-fbtch 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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991649318-5457-zgtcg 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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991648359-5453-jxcjl 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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991649318-5456-n7zf5 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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991649318-5455-t4xzx 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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991648869-5458-b2ffp 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991648869-5459-b86dx 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991648359-5454-g5v28 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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"

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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃     HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 39.39 ms │                                    40.41 ms │     no change │
│ QQuery 2  │ 19.05 ms │                                    18.97 ms │     no change │
│ QQuery 3  │ 28.35 ms │                                    28.25 ms │     no change │
│ QQuery 4  │ 17.50 ms │                                    17.57 ms │     no change │
│ QQuery 5  │ 35.57 ms │                                    35.59 ms │     no change │
│ QQuery 6  │ 13.77 ms │                                    13.70 ms │     no change │
│ QQuery 7  │ 41.36 ms │                                    41.75 ms │     no change │
│ QQuery 8  │ 40.67 ms │                                    40.61 ms │     no change │
│ QQuery 9  │ 49.34 ms │                                    48.83 ms │     no change │
│ QQuery 10 │ 57.70 ms │                                    40.70 ms │ +1.42x faster │
│ QQuery 11 │ 13.66 ms │                                    13.78 ms │     no change │
│ QQuery 12 │ 20.38 ms │                                    17.69 ms │ +1.15x faster │
│ QQuery 13 │ 47.05 ms │                                    40.41 ms │ +1.16x faster │
│ QQuery 14 │ 23.73 ms │                                    24.10 ms │     no change │
│ QQuery 15 │ 31.39 ms │                                    31.11 ms │     no change │
│ QQuery 16 │ 13.82 ms │                                    13.57 ms │     no change │
│ QQuery 17 │ 69.91 ms │                                    68.41 ms │     no change │
│ QQuery 18 │ 63.64 ms │                                    40.57 ms │ +1.57x faster │
│ QQuery 19 │ 29.07 ms │                                    31.75 ms │  1.09x slower │
│ QQuery 20 │ 33.93 ms │                                    32.32 ms │     no change │
│ QQuery 21 │ 53.47 ms │                                    54.19 ms │     no change │
│ QQuery 22 │ 15.75 ms │                                    13.78 ms │ +1.14x faster │
└───────────┴──────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 758.49ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 708.07ms │
│ Average Time (HEAD)                                        │  34.48ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │  32.18ms │
│ Queries Faster                                             │        5 │
│ Queries Slower                                             │        1 │
│ Queries with No Change                                     │       16 │
│ Queries with Failure                                       │        0 │
└────────────────────────────────────────────────────────────┴──────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 39.39 / 40.76 ±1.54 / 42.71 ms │              40.41 / 41.22 ±1.06 / 43.30 ms │     no change │
│ QQuery 2  │ 19.05 / 19.56 ±0.49 / 20.38 ms │              18.97 / 19.43 ±0.70 / 20.82 ms │     no change │
│ QQuery 3  │ 28.35 / 29.42 ±0.94 / 31.11 ms │              28.25 / 28.43 ±0.15 / 28.65 ms │     no change │
│ QQuery 4  │ 17.50 / 17.58 ±0.08 / 17.68 ms │              17.57 / 18.08 ±0.61 / 19.27 ms │     no change │
│ QQuery 5  │ 35.57 / 35.66 ±0.10 / 35.84 ms │              35.59 / 35.76 ±0.11 / 35.86 ms │     no change │
│ QQuery 6  │ 13.77 / 13.84 ±0.08 / 13.97 ms │              13.70 / 14.16 ±0.67 / 15.49 ms │     no change │
│ QQuery 7  │ 41.36 / 42.71 ±1.53 / 45.62 ms │              41.75 / 43.11 ±1.51 / 45.92 ms │     no change │
│ QQuery 8  │ 40.67 / 42.84 ±2.26 / 46.92 ms │              40.61 / 41.20 ±0.86 / 42.89 ms │     no change │
│ QQuery 9  │ 49.34 / 50.46 ±0.58 / 50.97 ms │              48.83 / 50.48 ±1.19 / 52.01 ms │     no change │
│ QQuery 10 │ 57.70 / 59.04 ±0.89 / 60.17 ms │              40.70 / 41.05 ±0.23 / 41.40 ms │ +1.44x faster │
│ QQuery 11 │ 13.66 / 13.95 ±0.20 / 14.28 ms │              13.78 / 14.18 ±0.52 / 15.20 ms │     no change │
│ QQuery 12 │ 20.38 / 20.64 ±0.33 / 21.29 ms │              17.69 / 18.15 ±0.27 / 18.46 ms │ +1.14x faster │
│ QQuery 13 │ 47.05 / 48.25 ±1.65 / 51.48 ms │              40.41 / 41.35 ±0.71 / 42.26 ms │ +1.17x faster │
│ QQuery 14 │ 23.73 / 24.00 ±0.14 / 24.14 ms │              24.10 / 24.24 ±0.08 / 24.32 ms │     no change │
│ QQuery 15 │ 31.39 / 31.55 ±0.10 / 31.70 ms │              31.11 / 32.05 ±0.53 / 32.71 ms │     no change │
│ QQuery 16 │ 13.82 / 13.88 ±0.06 / 13.98 ms │              13.57 / 13.69 ±0.10 / 13.88 ms │     no change │
│ QQuery 17 │ 69.91 / 70.91 ±0.78 / 72.08 ms │              68.41 / 69.30 ±0.74 / 70.50 ms │     no change │
│ QQuery 18 │ 63.64 / 65.23 ±1.26 / 67.25 ms │              40.57 / 40.80 ±0.21 / 41.20 ms │ +1.60x faster │
│ QQuery 19 │ 29.07 / 29.36 ±0.29 / 29.86 ms │              31.75 / 32.76 ±1.63 / 35.99 ms │  1.12x slower │
│ QQuery 20 │ 33.93 / 34.41 ±0.67 / 35.71 ms │              32.32 / 32.78 ±0.33 / 33.25 ms │     no change │
│ QQuery 21 │ 53.47 / 54.21 ±0.56 / 55.04 ms │              54.19 / 55.85 ±1.43 / 57.64 ms │     no change │
│ QQuery 22 │ 15.75 / 15.85 ±0.11 / 16.06 ms │              13.78 / 14.39 ±0.86 / 16.10 ms │ +1.10x faster │
└───────────┴────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 774.10ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 722.46ms │
│ Average Time (HEAD)                                        │  35.19ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │  32.84ms │
│ Queries Faster                                             │        5 │
│ Queries Slower                                             │        1 │
│ Queries with No Change                                     │       16 │
│ Queries with Failure                                       │        0 │
└────────────────────────────────────────────────────────────┴──────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

tpch — tpch_sf1

Query Base Changed Change
Query 1 30.2 MiB 40.0 MiB +32.7%
Query 2 23.1 MiB 24.7 MiB +7.0%
Query 3 10.9 MiB 10.9 MiB +0.0%
Query 4 32.1 MiB 33.3 MiB +3.7%
Query 5 42.8 MiB 42.7 MiB -0.0%
Query 6 832 B 832 B +0.0%
Query 7 105.6 MiB 105.2 MiB -0.4%
Query 8 39.8 MiB 32.4 MiB -18.7%
Query 9 142.2 MiB 142.2 MiB +0.0%
Query 10 53.3 MiB 36.9 MiB -30.9%
Query 11 61.9 MiB 61.2 MiB -1.2%
Query 12 22.5 MiB 22.7 MiB +0.9%
Query 13 52.5 MiB 62.7 MiB +19.5%
Query 14 6.7 MiB 6.7 MiB +0.0%
Query 15 12.7 MiB 12.7 MiB +0.0%
Query 16 42.6 MiB 44.5 MiB +4.4%
Query 17 150.3 MiB 150.3 MiB +0.0%
Query 18 114.1 MiB 167.1 MiB +46.4%
Query 19 209.7 KiB 725.7 KiB +246.1%
Query 20 58.6 MiB 57.2 MiB -2.4%
Query 21 35.9 MiB 35.8 MiB -0.2%
Query 22 31.9 MiB 32.7 MiB +2.2%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpch base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 150.3 MiB 1.1 GiB 1011.2 MiB 7.7×
tpch changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 167.1 MiB 1.2 GiB 1.0 GiB 7.4×
Resource Usage

tpch — base (merge-base)

Metric Value
Wall time 5.0s
Peak memory 1.1 GiB
Avg memory 482.2 MiB
CPU user 20.0s
CPU sys 1.7s
Peak spill 0 B

tpch — branch

Metric Value
Wall time 5.0s
Peak memory 1.2 GiB
Avg memory 497.7 MiB
CPU user 19.2s
CPU sys 1.7s
Peak spill 0 B

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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃     HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 39.39 ms │                                    40.83 ms │     no change │
│ QQuery 2  │ 18.84 ms │                                    18.87 ms │     no change │
│ QQuery 3  │ 28.45 ms │                                    28.47 ms │     no change │
│ QQuery 4  │ 17.50 ms │                                    17.55 ms │     no change │
│ QQuery 5  │ 35.47 ms │                                    35.34 ms │     no change │
│ QQuery 6  │ 13.79 ms │                                    13.78 ms │     no change │
│ QQuery 7  │ 41.19 ms │                                    42.17 ms │     no change │
│ QQuery 8  │ 40.89 ms │                                    40.70 ms │     no change │
│ QQuery 9  │ 49.04 ms │                                    48.92 ms │     no change │
│ QQuery 10 │ 57.31 ms │                                    40.58 ms │ +1.41x faster │
│ QQuery 11 │ 13.66 ms │                                    13.67 ms │     no change │
│ QQuery 12 │ 20.12 ms │                                    17.79 ms │ +1.13x faster │
│ QQuery 13 │ 46.24 ms │                                    40.18 ms │ +1.15x faster │
│ QQuery 14 │ 23.84 ms │                                    24.27 ms │     no change │
│ QQuery 15 │ 31.21 ms │                                    31.41 ms │     no change │
│ QQuery 16 │ 13.63 ms │                                    13.57 ms │     no change │
│ QQuery 17 │ 67.47 ms │                                    68.94 ms │     no change │
│ QQuery 18 │ 64.53 ms │                                    41.03 ms │ +1.57x faster │
│ QQuery 19 │ 29.27 ms │                                    31.92 ms │  1.09x slower │
│ QQuery 20 │ 33.74 ms │                                    32.00 ms │ +1.05x faster │
│ QQuery 21 │ 52.47 ms │                                    53.41 ms │     no change │
│ QQuery 22 │ 15.45 ms │                                    13.58 ms │ +1.14x faster │
└───────────┴──────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 753.49ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 708.97ms │
│ Average Time (HEAD)                                        │  34.25ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │  32.23ms │
│ Queries Faster                                             │        6 │
│ Queries Slower                                             │        1 │
│ Queries with No Change                                     │       15 │
│ Queries with Failure                                       │        0 │
└────────────────────────────────────────────────────────────┴──────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 39.39 / 40.86 ±1.73 / 43.60 ms │              40.83 / 41.89 ±1.24 / 43.48 ms │     no change │
│ QQuery 2  │ 18.84 / 19.16 ±0.32 / 19.68 ms │              18.87 / 19.04 ±0.14 / 19.21 ms │     no change │
│ QQuery 3  │ 28.45 / 29.14 ±0.77 / 30.63 ms │              28.47 / 29.12 ±1.01 / 31.13 ms │     no change │
│ QQuery 4  │ 17.50 / 17.65 ±0.10 / 17.77 ms │              17.55 / 18.02 ±0.49 / 18.95 ms │     no change │
│ QQuery 5  │ 35.47 / 36.14 ±0.91 / 37.94 ms │              35.34 / 36.63 ±2.22 / 41.07 ms │     no change │
│ QQuery 6  │ 13.79 / 14.12 ±0.51 / 15.13 ms │              13.78 / 14.20 ±0.75 / 15.71 ms │     no change │
│ QQuery 7  │ 41.19 / 43.62 ±2.46 / 47.85 ms │              42.17 / 43.50 ±1.61 / 46.16 ms │     no change │
│ QQuery 8  │ 40.89 / 41.34 ±0.52 / 42.12 ms │              40.70 / 41.08 ±0.38 / 41.79 ms │     no change │
│ QQuery 9  │ 49.04 / 50.77 ±1.43 / 53.39 ms │              48.92 / 50.49 ±1.24 / 52.61 ms │     no change │
│ QQuery 10 │ 57.31 / 57.66 ±0.47 / 58.58 ms │              40.58 / 40.79 ±0.22 / 41.21 ms │ +1.41x faster │
│ QQuery 11 │ 13.66 / 13.83 ±0.14 / 14.01 ms │              13.67 / 14.03 ±0.49 / 14.97 ms │     no change │
│ QQuery 12 │ 20.12 / 20.46 ±0.19 / 20.69 ms │              17.79 / 18.01 ±0.22 / 18.37 ms │ +1.14x faster │
│ QQuery 13 │ 46.24 / 48.01 ±1.73 / 51.32 ms │              40.18 / 41.62 ±1.73 / 44.56 ms │ +1.15x faster │
│ QQuery 14 │ 23.84 / 24.12 ±0.15 / 24.25 ms │              24.27 / 24.33 ±0.06 / 24.45 ms │     no change │
│ QQuery 15 │ 31.21 / 31.73 ±0.67 / 32.93 ms │              31.41 / 32.00 ±0.49 / 32.68 ms │     no change │
│ QQuery 16 │ 13.63 / 13.82 ±0.18 / 14.06 ms │              13.57 / 13.64 ±0.04 / 13.70 ms │     no change │
│ QQuery 17 │ 67.47 / 68.08 ±0.41 / 68.50 ms │              68.94 / 71.03 ±1.68 / 73.05 ms │     no change │
│ QQuery 18 │ 64.53 / 67.36 ±2.12 / 70.24 ms │              41.03 / 41.64 ±0.34 / 42.03 ms │ +1.62x faster │
│ QQuery 19 │ 29.27 / 30.00 ±1.15 / 32.28 ms │              31.92 / 32.18 ±0.23 / 32.59 ms │  1.07x slower │
│ QQuery 20 │ 33.74 / 34.10 ±0.25 / 34.43 ms │              32.00 / 32.57 ±0.92 / 34.41 ms │     no change │
│ QQuery 21 │ 52.47 / 53.93 ±1.48 / 56.38 ms │              53.41 / 54.24 ±0.67 / 54.98 ms │     no change │
│ QQuery 22 │ 15.45 / 15.65 ±0.13 / 15.80 ms │              13.58 / 13.71 ±0.16 / 14.03 ms │ +1.14x faster │
└───────────┴────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃          ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 771.55ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 723.77ms │
│ Average Time (HEAD)                                        │  35.07ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │  32.90ms │
│ Queries Faster                                             │        5 │
│ Queries Slower                                             │        1 │
│ Queries with No Change                                     │       16 │
│ Queries with Failure                                       │        0 │
└────────────────────────────────────────────────────────────┴──────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

tpch — tpch_sf1

Query Base Changed Change
Query 1 30.2 MiB 31.4 MiB +4.0%
Query 2 23.1 MiB 24.7 MiB +7.0%
Query 3 10.6 MiB 10.9 MiB +2.4%
Query 4 32.2 MiB 23.4 MiB -27.5%
Query 5 42.8 MiB 42.9 MiB +0.3%
Query 6 832 B 832 B +0.0%
Query 7 108.5 MiB 103.6 MiB -4.5%
Query 8 39.8 MiB 39.8 MiB +0.0%
Query 9 147.5 MiB 141.3 MiB -4.2%
Query 10 53.8 MiB 36.9 MiB -31.3%
Query 11 41.7 MiB 51.5 MiB +23.5%
Query 12 22.5 MiB 22.6 MiB +0.7%
Query 13 72.3 MiB 52.6 MiB -27.3%
Query 14 6.7 MiB 6.7 MiB +0.0%
Query 15 12.8 MiB 12.7 MiB -0.8%
Query 16 43.3 MiB 34.1 MiB -21.1%
Query 17 150.3 MiB 150.3 MiB +0.0%
Query 18 108.8 MiB 167.1 MiB +53.5%
Query 19 209.7 KiB 725.7 KiB +246.1%
Query 20 60.5 MiB 60.5 MiB +0.0%
Query 21 36.0 MiB 35.7 MiB -0.6%
Query 22 22.0 MiB 23.0 MiB +4.7%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpch base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 150.3 MiB 1.1 GiB 1013.2 MiB 7.7×
tpch changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 167.1 MiB 1.1 GiB 1009.2 MiB 7.0×
Resource Usage

tpch — base (merge-base)

Metric Value
Wall time 5.0s
Peak memory 1.1 GiB
Avg memory 494.8 MiB
CPU user 20.0s
CPU sys 1.6s
Peak spill 0 B

tpch — branch

Metric Value
Wall time 5.0s
Peak memory 1.1 GiB
Avg memory 489.5 MiB
CPU user 19.4s
CPU sys 1.7s
Peak spill 0 B

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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │    5.76 ms │                                     5.71 ms │     no change │
│ QQuery 2  │   76.78 ms │                                    79.46 ms │     no change │
│ QQuery 3  │   29.33 ms │                                    29.36 ms │     no change │
│ QQuery 4  │  558.02 ms │                                   481.46 ms │ +1.16x faster │
│ QQuery 5  │   51.02 ms │                                    51.18 ms │     no change │
│ QQuery 6  │   34.16 ms │                                    35.93 ms │  1.05x slower │
│ QQuery 7  │   72.42 ms │                                    73.14 ms │     no change │
│ QQuery 8  │   38.64 ms │                                    37.13 ms │     no change │
│ QQuery 9  │   51.73 ms │                                    52.01 ms │     no change │
│ QQuery 10 │   59.45 ms │                                    60.17 ms │     no change │
│ QQuery 11 │  365.50 ms │                                   288.44 ms │ +1.27x faster │
│ QQuery 12 │   26.15 ms │                                    26.17 ms │     no change │
│ QQuery 13 │  112.73 ms │                                   112.94 ms │     no change │
│ QQuery 14 │  381.56 ms │                                   376.81 ms │     no change │
│ QQuery 15 │   50.92 ms │                                    53.91 ms │  1.06x slower │
│ QQuery 16 │    6.15 ms │                                     5.71 ms │ +1.08x faster │
│ QQuery 17 │   76.60 ms │                                    76.75 ms │     no change │
│ QQuery 18 │  103.46 ms │                                   101.79 ms │     no change │
│ QQuery 19 │   41.98 ms │                                    41.00 ms │     no change │
│ QQuery 20 │   31.58 ms │                                    31.43 ms │     no change │
│ QQuery 21 │   17.80 ms │                                    16.91 ms │     no change │
│ QQuery 22 │   62.41 ms │                                    60.21 ms │     no change │
│ QQuery 23 │  302.39 ms │                                   291.98 ms │     no change │
│ QQuery 24 │  199.96 ms │                                   201.69 ms │     no change │
│ QQuery 25 │  104.33 ms │                                   103.51 ms │     no change │
│ QQuery 26 │   44.92 ms │                                    46.59 ms │     no change │
│ QQuery 27 │    5.35 ms │                                     5.38 ms │     no change │
│ QQuery 28 │   60.97 ms │                                    61.11 ms │     no change │
│ QQuery 29 │   93.90 ms │                                    93.66 ms │     no change │
│ QQuery 30 │   36.74 ms │                                    32.89 ms │ +1.12x faster │
│ QQuery 31 │  110.88 ms │                                   108.78 ms │     no change │
│ QQuery 32 │   20.43 ms │                                    20.66 ms │     no change │
│ QQuery 33 │   37.41 ms │                                    37.28 ms │     no change │
│ QQuery 34 │    5.99 ms │                                     5.98 ms │     no change │
│ QQuery 35 │   72.26 ms │                                    73.05 ms │     no change │
│ QQuery 36 │    6.35 ms │                                     5.97 ms │ +1.06x faster │
│ QQuery 37 │    8.44 ms │                                     7.74 ms │ +1.09x faster │
│ QQuery 38 │   66.23 ms │                                    61.54 ms │ +1.08x faster │
│ QQuery 39 │   74.98 ms │                                    71.84 ms │     no change │
│ QQuery 40 │   23.69 ms │                                    22.71 ms │     no change │
│ QQuery 41 │   14.69 ms │                                    14.23 ms │     no change │
│ QQuery 42 │   24.12 ms │                                    23.95 ms │     no change │
│ QQuery 43 │    4.83 ms │                                     4.55 ms │ +1.06x faster │
│ QQuery 44 │    5.80 ms │                                     5.36 ms │ +1.08x faster │
│ QQuery 45 │   36.13 ms │                                    36.18 ms │     no change │
│ QQuery 46 │   11.84 ms │                                    11.87 ms │     no change │
│ QQuery 47 │  205.81 ms │                                   200.57 ms │     no change │
│ QQuery 48 │   90.23 ms │                                    91.05 ms │     no change │
│ QQuery 49 │   68.95 ms │                                    70.56 ms │     no change │
│ QQuery 50 │   56.73 ms │                                    56.53 ms │     no change │
│ QQuery 51 │   91.73 ms │                                    94.41 ms │     no change │
│ QQuery 52 │   23.92 ms │                                    24.47 ms │     no change │
│ QQuery 53 │   29.53 ms │                                    29.42 ms │     no change │
│ QQuery 54 │   53.13 ms │                                    54.02 ms │     no change │
│ QQuery 55 │   23.98 ms │                                    23.96 ms │     no change │
│ QQuery 56 │   38.56 ms │                                    38.46 ms │     no change │
│ QQuery 57 │  151.14 ms │                                   152.14 ms │     no change │
│ QQuery 58 │  103.65 ms │                                   102.52 ms │     no change │
│ QQuery 59 │  106.62 ms │                                   114.11 ms │  1.07x slower │
│ QQuery 60 │   38.56 ms │                                    38.20 ms │     no change │
│ QQuery 61 │    9.27 ms │                                     8.93 ms │     no change │
│ QQuery 62 │   39.03 ms │                                    37.86 ms │     no change │
│ QQuery 63 │   30.18 ms │                                    29.63 ms │     no change │
│ QQuery 64 │  362.97 ms │                                   352.22 ms │     no change │
│ QQuery 65 │   70.77 ms │                                    69.96 ms │     no change │
│ QQuery 66 │   73.12 ms │                                    72.39 ms │     no change │
│ QQuery 67 │  192.55 ms │                                   184.45 ms │     no change │
│ QQuery 68 │    7.02 ms │                                     6.26 ms │ +1.12x faster │
│ QQuery 69 │   55.06 ms │                                    55.61 ms │     no change │
│ QQuery 70 │   97.76 ms │                                    96.82 ms │     no change │
│ QQuery 71 │   35.62 ms │                                    35.71 ms │     no change │
│ QQuery 72 │ 1681.40 ms │                                  1571.89 ms │ +1.07x faster │
│ QQuery 73 │    5.60 ms │                                     5.37 ms │     no change │
│ QQuery 74 │  243.40 ms │                                   193.76 ms │ +1.26x faster │
│ QQuery 75 │  139.46 ms │                                   137.73 ms │     no change │
│ QQuery 76 │   34.85 ms │                                    33.88 ms │     no change │
│ QQuery 77 │   59.85 ms │                                    58.98 ms │     no change │
│ QQuery 78 │  269.86 ms │                                   258.56 ms │     no change │
│ QQuery 79 │   68.99 ms │                                    66.13 ms │     no change │
│ QQuery 80 │   94.67 ms │                                    94.12 ms │     no change │
│ QQuery 81 │   28.48 ms │                                    25.58 ms │ +1.11x faster │
│ QQuery 82 │   17.64 ms │                                    17.88 ms │     no change │
│ QQuery 83 │   32.67 ms │                                    32.54 ms │     no change │
│ QQuery 84 │   31.25 ms │                                    29.90 ms │     no change │
│ QQuery 85 │   95.34 ms │                                    98.24 ms │     no change │
│ QQuery 86 │   24.90 ms │                                    24.27 ms │     no change │
│ QQuery 87 │   65.92 ms │                                    61.59 ms │ +1.07x faster │
│ QQuery 88 │   60.34 ms │                                    59.87 ms │     no change │
│ QQuery 89 │   35.79 ms │                                    35.99 ms │     no change │
│ QQuery 90 │   16.76 ms │                                    16.50 ms │     no change │
│ QQuery 91 │   40.03 ms │                                    40.98 ms │     no change │
│ QQuery 92 │   29.69 ms │                                    29.08 ms │     no change │
│ QQuery 93 │   48.86 ms │                                    49.53 ms │     no change │
│ QQuery 94 │   39.01 ms │                                    38.01 ms │     no change │
│ QQuery 95 │   80.03 ms │                                    78.64 ms │     no change │
│ QQuery 96 │   23.72 ms │                                    23.84 ms │     no change │
│ QQuery 97 │   50.24 ms │                                    50.56 ms │     no change │
│ QQuery 98 │   38.14 ms │                                    38.34 ms │     no change │
│ QQuery 99 │   57.07 ms │                                    56.82 ms │     no change │
└───────────┴────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 9096.61ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 8714.94ms │
│ Average Time (HEAD)                                        │   91.88ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   88.03ms │
│ Queries Faster                                             │        14 │
│ Queries Slower                                             │         3 │
│ Queries with No Change                                     │        82 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │           5.76 / 6.33 ±0.98 / 8.28 ms │                 5.71 / 6.23 ±0.96 / 8.16 ms │     no change │
│ QQuery 2  │        76.78 / 77.05 ±0.20 / 77.40 ms │              79.46 / 80.24 ±0.44 / 80.76 ms │     no change │
│ QQuery 3  │        29.33 / 29.50 ±0.17 / 29.79 ms │              29.36 / 29.58 ±0.20 / 29.89 ms │     no change │
│ QQuery 4  │     558.02 / 562.21 ±3.63 / 566.85 ms │           481.46 / 486.06 ±3.72 / 492.25 ms │ +1.16x faster │
│ QQuery 5  │        51.02 / 51.65 ±0.62 / 52.41 ms │              51.18 / 52.27 ±1.37 / 54.89 ms │     no change │
│ QQuery 6  │        34.16 / 34.54 ±0.30 / 34.88 ms │              35.93 / 36.60 ±0.35 / 36.94 ms │  1.06x slower │
│ QQuery 7  │        72.42 / 74.22 ±2.88 / 79.85 ms │              73.14 / 73.59 ±0.41 / 74.11 ms │     no change │
│ QQuery 8  │        38.64 / 39.19 ±0.32 / 39.55 ms │              37.13 / 38.46 ±2.29 / 43.03 ms │     no change │
│ QQuery 9  │        51.73 / 54.02 ±1.46 / 55.64 ms │              52.01 / 55.75 ±4.42 / 63.83 ms │     no change │
│ QQuery 10 │        59.45 / 59.76 ±0.19 / 59.97 ms │              60.17 / 60.55 ±0.19 / 60.73 ms │     no change │
│ QQuery 11 │     365.50 / 374.20 ±5.20 / 379.75 ms │           288.44 / 295.94 ±6.13 / 303.67 ms │ +1.26x faster │
│ QQuery 12 │        26.15 / 26.53 ±0.30 / 26.94 ms │              26.17 / 26.78 ±0.53 / 27.57 ms │     no change │
│ QQuery 13 │     112.73 / 115.38 ±1.87 / 118.49 ms │           112.94 / 113.51 ±0.43 / 114.12 ms │     no change │
│ QQuery 14 │     381.56 / 382.75 ±1.02 / 384.27 ms │           376.81 / 381.18 ±3.62 / 386.76 ms │     no change │
│ QQuery 15 │        50.92 / 51.54 ±0.39 / 52.09 ms │              53.91 / 54.94 ±1.27 / 57.32 ms │  1.07x slower │
│ QQuery 16 │           6.15 / 6.31 ±0.18 / 6.67 ms │                 5.71 / 5.95 ±0.21 / 6.34 ms │ +1.06x faster │
│ QQuery 17 │        76.60 / 78.67 ±2.31 / 81.71 ms │              76.75 / 78.48 ±1.19 / 80.43 ms │     no change │
│ QQuery 18 │     103.46 / 104.55 ±0.94 / 105.95 ms │           101.79 / 104.68 ±2.45 / 108.86 ms │     no change │
│ QQuery 19 │        41.98 / 42.10 ±0.18 / 42.47 ms │              41.00 / 41.32 ±0.26 / 41.63 ms │     no change │
│ QQuery 20 │        31.58 / 32.19 ±0.44 / 32.80 ms │              31.43 / 31.78 ±0.36 / 32.39 ms │     no change │
│ QQuery 21 │        17.80 / 18.96 ±1.61 / 22.12 ms │              16.91 / 17.05 ±0.13 / 17.27 ms │ +1.11x faster │
│ QQuery 22 │        62.41 / 64.20 ±1.39 / 65.99 ms │              60.21 / 61.02 ±0.57 / 61.90 ms │     no change │
│ QQuery 23 │    302.39 / 314.79 ±15.13 / 340.18 ms │           291.98 / 297.85 ±5.97 / 308.87 ms │ +1.06x faster │
│ QQuery 24 │    199.96 / 208.82 ±13.84 / 236.42 ms │           201.69 / 205.47 ±2.22 / 207.92 ms │     no change │
│ QQuery 25 │     104.33 / 106.97 ±2.95 / 112.69 ms │           103.51 / 107.47 ±4.78 / 116.80 ms │     no change │
│ QQuery 26 │        44.92 / 45.14 ±0.21 / 45.50 ms │              46.59 / 46.80 ±0.17 / 47.03 ms │     no change │
│ QQuery 27 │           5.35 / 5.53 ±0.19 / 5.88 ms │                 5.38 / 5.53 ±0.08 / 5.62 ms │     no change │
│ QQuery 28 │        60.97 / 61.78 ±0.52 / 62.58 ms │              61.11 / 61.97 ±0.64 / 62.77 ms │     no change │
│ QQuery 29 │       93.90 / 96.56 ±2.46 / 100.22 ms │              93.66 / 95.81 ±2.42 / 99.66 ms │     no change │
│ QQuery 30 │        36.74 / 37.06 ±0.25 / 37.35 ms │              32.89 / 33.25 ±0.28 / 33.71 ms │ +1.11x faster │
│ QQuery 31 │     110.88 / 114.18 ±4.26 / 122.53 ms │           108.78 / 110.21 ±1.64 / 113.30 ms │     no change │
│ QQuery 32 │        20.43 / 20.82 ±0.24 / 21.11 ms │              20.66 / 22.35 ±2.66 / 27.66 ms │  1.07x slower │
│ QQuery 33 │        37.41 / 37.64 ±0.22 / 37.99 ms │              37.28 / 38.08 ±0.56 / 38.76 ms │     no change │
│ QQuery 34 │           5.99 / 6.13 ±0.11 / 6.26 ms │                 5.98 / 6.11 ±0.16 / 6.42 ms │     no change │
│ QQuery 35 │        72.26 / 72.97 ±0.66 / 74.01 ms │              73.05 / 73.96 ±0.85 / 75.36 ms │     no change │
│ QQuery 36 │           6.35 / 6.49 ±0.13 / 6.67 ms │                 5.97 / 6.25 ±0.19 / 6.52 ms │     no change │
│ QQuery 37 │          8.44 / 9.24 ±1.03 / 11.29 ms │                 7.74 / 7.86 ±0.10 / 8.00 ms │ +1.18x faster │
│ QQuery 38 │        66.23 / 66.85 ±0.77 / 68.34 ms │              61.54 / 63.59 ±2.10 / 67.62 ms │     no change │
│ QQuery 39 │        74.98 / 75.38 ±0.33 / 75.83 ms │              71.84 / 73.74 ±1.70 / 76.71 ms │     no change │
│ QQuery 40 │        23.69 / 24.35 ±0.48 / 24.93 ms │              22.71 / 23.04 ±0.27 / 23.42 ms │ +1.06x faster │
│ QQuery 41 │        14.69 / 14.88 ±0.19 / 15.11 ms │              14.23 / 14.88 ±0.91 / 16.67 ms │     no change │
│ QQuery 42 │        24.12 / 26.32 ±1.72 / 28.65 ms │              23.95 / 24.53 ±0.53 / 25.43 ms │ +1.07x faster │
│ QQuery 43 │           4.83 / 5.03 ±0.22 / 5.43 ms │                 4.55 / 4.73 ±0.24 / 5.22 ms │ +1.06x faster │
│ QQuery 44 │           5.80 / 5.93 ±0.07 / 5.98 ms │                 5.36 / 5.44 ±0.07 / 5.56 ms │ +1.09x faster │
│ QQuery 45 │        36.13 / 36.84 ±0.47 / 37.38 ms │              36.18 / 37.37 ±0.74 / 38.22 ms │     no change │
│ QQuery 46 │        11.84 / 12.19 ±0.20 / 12.40 ms │              11.87 / 11.95 ±0.08 / 12.06 ms │     no change │
│ QQuery 47 │     205.81 / 211.87 ±4.05 / 215.58 ms │           200.57 / 205.15 ±2.56 / 208.01 ms │     no change │
│ QQuery 48 │        90.23 / 90.83 ±0.41 / 91.30 ms │              91.05 / 91.51 ±0.35 / 91.87 ms │     no change │
│ QQuery 49 │        68.95 / 72.41 ±3.74 / 79.46 ms │              70.56 / 70.98 ±0.52 / 71.85 ms │     no change │
│ QQuery 50 │        56.73 / 57.60 ±0.58 / 58.51 ms │              56.53 / 58.89 ±3.36 / 65.55 ms │     no change │
│ QQuery 51 │        91.73 / 94.00 ±1.93 / 96.84 ms │              94.41 / 95.35 ±0.82 / 96.74 ms │     no change │
│ QQuery 52 │        23.92 / 25.65 ±2.20 / 30.01 ms │              24.47 / 24.76 ±0.21 / 25.10 ms │     no change │
│ QQuery 53 │        29.53 / 29.76 ±0.22 / 30.06 ms │              29.42 / 29.63 ±0.12 / 29.76 ms │     no change │
│ QQuery 54 │        53.13 / 53.91 ±0.61 / 54.80 ms │              54.02 / 55.86 ±3.05 / 61.93 ms │     no change │
│ QQuery 55 │        23.98 / 24.21 ±0.16 / 24.43 ms │              23.96 / 24.40 ±0.33 / 24.77 ms │     no change │
│ QQuery 56 │        38.56 / 39.11 ±0.48 / 39.82 ms │              38.46 / 39.21 ±0.49 / 39.92 ms │     no change │
│ QQuery 57 │     151.14 / 153.13 ±1.84 / 155.88 ms │           152.14 / 153.79 ±1.07 / 155.22 ms │     no change │
│ QQuery 58 │     103.65 / 105.41 ±2.08 / 109.37 ms │           102.52 / 103.31 ±0.56 / 104.20 ms │     no change │
│ QQuery 59 │     106.62 / 107.07 ±0.40 / 107.71 ms │           114.11 / 115.74 ±2.00 / 119.68 ms │  1.08x slower │
│ QQuery 60 │        38.56 / 38.94 ±0.32 / 39.49 ms │              38.20 / 38.52 ±0.38 / 39.27 ms │     no change │
│ QQuery 61 │           9.27 / 9.37 ±0.10 / 9.55 ms │                8.93 / 9.63 ±1.01 / 11.62 ms │     no change │
│ QQuery 62 │        39.03 / 40.63 ±1.40 / 42.70 ms │              37.86 / 38.45 ±0.47 / 38.87 ms │ +1.06x faster │
│ QQuery 63 │        30.18 / 30.42 ±0.37 / 31.16 ms │              29.63 / 29.84 ±0.21 / 30.22 ms │     no change │
│ QQuery 64 │     362.97 / 367.12 ±3.60 / 372.71 ms │           352.22 / 357.28 ±2.61 / 359.53 ms │     no change │
│ QQuery 65 │        70.77 / 72.10 ±1.02 / 73.18 ms │              69.96 / 70.90 ±0.58 / 71.40 ms │     no change │
│ QQuery 66 │        73.12 / 73.38 ±0.31 / 73.91 ms │              72.39 / 74.80 ±3.96 / 82.70 ms │     no change │
│ QQuery 67 │     192.55 / 197.92 ±4.79 / 206.86 ms │           184.45 / 196.35 ±6.19 / 201.25 ms │     no change │
│ QQuery 68 │           7.02 / 7.14 ±0.09 / 7.26 ms │                 6.26 / 6.50 ±0.28 / 7.04 ms │ +1.10x faster │
│ QQuery 69 │        55.06 / 57.24 ±3.51 / 64.25 ms │              55.61 / 56.79 ±1.19 / 59.09 ms │     no change │
│ QQuery 70 │       97.76 / 98.75 ±1.16 / 101.00 ms │             96.82 / 99.77 ±4.08 / 107.85 ms │     no change │
│ QQuery 71 │        35.62 / 35.78 ±0.18 / 36.11 ms │              35.71 / 36.50 ±0.57 / 37.24 ms │     no change │
│ QQuery 72 │ 1681.40 / 1702.49 ±23.48 / 1745.72 ms │       1571.89 / 1629.71 ±57.52 / 1732.04 ms │     no change │
│ QQuery 73 │           5.60 / 5.70 ±0.12 / 5.92 ms │                 5.37 / 6.24 ±1.57 / 9.37 ms │  1.09x slower │
│ QQuery 74 │     243.40 / 246.57 ±2.44 / 249.34 ms │           193.76 / 199.56 ±5.64 / 209.27 ms │ +1.24x faster │
│ QQuery 75 │     139.46 / 140.90 ±1.19 / 142.66 ms │           137.73 / 138.37 ±0.49 / 139.10 ms │     no change │
│ QQuery 76 │        34.85 / 35.27 ±0.32 / 35.78 ms │              33.88 / 34.28 ±0.34 / 34.78 ms │     no change │
│ QQuery 77 │        59.85 / 60.48 ±0.39 / 60.99 ms │              58.98 / 59.99 ±1.02 / 61.77 ms │     no change │
│ QQuery 78 │    269.86 / 282.23 ±12.07 / 303.61 ms │           258.56 / 273.69 ±9.61 / 285.38 ms │     no change │
│ QQuery 79 │        68.99 / 69.29 ±0.30 / 69.82 ms │              66.13 / 68.13 ±1.32 / 69.72 ms │     no change │
│ QQuery 80 │       94.67 / 97.41 ±2.24 / 101.36 ms │              94.12 / 95.47 ±1.11 / 97.04 ms │     no change │
│ QQuery 81 │        28.48 / 31.15 ±3.38 / 37.67 ms │              25.58 / 26.40 ±1.39 / 29.18 ms │ +1.18x faster │
│ QQuery 82 │        17.64 / 17.94 ±0.36 / 18.59 ms │              17.88 / 19.94 ±3.76 / 27.46 ms │  1.11x slower │
│ QQuery 83 │        32.67 / 32.89 ±0.14 / 33.04 ms │              32.54 / 32.99 ±0.28 / 33.38 ms │     no change │
│ QQuery 84 │        31.25 / 31.90 ±0.59 / 32.91 ms │              29.90 / 30.24 ±0.24 / 30.65 ms │ +1.05x faster │
│ QQuery 85 │        95.34 / 96.82 ±1.46 / 99.60 ms │             98.24 / 99.90 ±2.52 / 104.84 ms │     no change │
│ QQuery 86 │        24.90 / 26.79 ±2.97 / 32.71 ms │              24.27 / 24.50 ±0.28 / 25.05 ms │ +1.09x faster │
│ QQuery 87 │        65.92 / 66.76 ±0.43 / 67.13 ms │              61.59 / 63.54 ±1.85 / 66.97 ms │     no change │
│ QQuery 88 │        60.34 / 60.67 ±0.25 / 61.04 ms │              59.87 / 61.21 ±1.20 / 63.04 ms │     no change │
│ QQuery 89 │        35.79 / 36.43 ±0.62 / 37.60 ms │              35.99 / 36.55 ±0.31 / 36.86 ms │     no change │
│ QQuery 90 │        16.76 / 18.65 ±3.57 / 25.79 ms │              16.50 / 16.64 ±0.08 / 16.71 ms │ +1.12x faster │
│ QQuery 91 │        40.03 / 40.76 ±0.54 / 41.51 ms │              40.98 / 42.35 ±1.76 / 45.84 ms │     no change │
│ QQuery 92 │        29.69 / 30.23 ±0.35 / 30.80 ms │              29.08 / 29.67 ±0.56 / 30.72 ms │     no change │
│ QQuery 93 │        48.86 / 50.12 ±1.08 / 51.91 ms │              49.53 / 49.78 ±0.18 / 49.94 ms │     no change │
│ QQuery 94 │        39.01 / 39.55 ±0.29 / 39.83 ms │              38.01 / 38.68 ±0.53 / 39.59 ms │     no change │
│ QQuery 95 │        80.03 / 81.82 ±1.66 / 84.30 ms │              78.64 / 81.08 ±1.45 / 83.02 ms │     no change │
│ QQuery 96 │        23.72 / 23.95 ±0.14 / 24.13 ms │              23.84 / 24.16 ±0.26 / 24.63 ms │     no change │
│ QQuery 97 │        50.24 / 52.20 ±1.53 / 53.86 ms │              50.56 / 51.33 ±1.14 / 53.59 ms │     no change │
│ QQuery 98 │        38.14 / 38.38 ±0.17 / 38.54 ms │              38.34 / 39.29 ±0.88 / 40.77 ms │     no change │
│ QQuery 99 │        57.07 / 59.94 ±3.57 / 66.64 ms │              56.82 / 57.15 ±0.48 / 58.11 ms │     no change │
└───────────┴───────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 9268.80ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 8924.96ms │
│ Average Time (HEAD)                                        │   93.62ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   90.15ms │
│ Queries Faster                                             │        18 │
│ Queries Slower                                             │         6 │
│ Queries with No Change                                     │        75 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

tpcds — tpcds_sf1

Query Base Changed Change
Query 1 135.7 KiB 135.7 KiB -0.0%
Query 2 81.9 MiB 81.3 MiB -0.7%
Query 3 1.9 MiB 2.3 MiB +24.1%
Query 4 85.6 MiB 111.4 MiB +30.2%
Query 5 13.0 MiB 13.0 MiB +0.1%
Query 6 4.7 MiB 7.8 MiB +65.9%
Query 7 3.7 MiB 3.6 MiB -1.9%
Query 8 1.8 MiB 3.0 MiB +72.3%
Query 9 32.1 KiB 32.1 KiB +0.0%
Query 10 2.0 MiB 3.1 MiB +53.3%
Query 11 58.1 MiB 77.1 MiB +32.7%
Query 12 114.4 MiB 114.4 MiB +0.0%
Query 13 13.7 MiB 17.6 MiB +28.4%
Query 14 170.9 MiB 170.9 MiB +0.0%
Query 15 4.2 MiB 6.8 MiB +62.5%
Query 16 118.6 KiB 161.5 KiB +36.2%
Query 17 39.6 MiB 39.8 MiB +0.5%
Query 18 21.2 MiB 21.0 MiB -1.1%
Query 19 2.1 MiB 2.1 MiB +2.6%
Query 20 114.6 MiB 114.6 MiB -0.0%
Query 21 3.2 MiB 3.0 MiB -9.1%
Query 22 61.8 MiB 61.8 MiB +0.0%
Query 23 213.4 MiB 216.5 MiB +1.5%
Query 24 36.4 MiB 45.7 MiB +25.3%
Query 25 85.5 MiB 85.6 MiB +0.2%
Query 26 3.8 MiB 3.5 MiB -8.0%
Query 27 929.6 KiB 929.6 KiB +0.0%
Query 28 1.5 MiB 1.3 MiB -9.1%
Query 29 97.2 MiB 97.2 MiB +0.0%
Query 30 3.8 MiB 7.1 MiB +88.1%
Query 31 13.4 MiB 18.0 MiB +34.3%
Query 32 2.7 MiB 2.5 MiB -8.1%
Query 33 4.0 MiB 4.0 MiB +0.0%
Query 34 96.3 KiB 96.3 KiB +0.0%
Query 35 16.6 MiB 25.5 MiB +53.2%
Query 36 10.3 MiB 10.3 MiB +0.0%
Query 37 3.3 MiB 3.3 MiB +0.0%
Query 38 10.4 MiB 19.7 MiB +88.1%
Query 39 41.5 MiB 33.3 MiB -19.9%
Query 40 12.0 MiB 11.5 MiB -4.7%
Query 41 2.1 MiB 2.1 MiB +0.0%
Query 42 1.5 MiB 1.3 MiB -15.8%
Query 43 258.2 KiB 275.4 KiB +6.7%
Query 44 97.6 KiB 192.4 KiB +97.2%
Query 45 4.4 MiB 7.9 MiB +78.8%
Query 46 5.4 MiB 9.4 MiB +73.5%
Query 47 128.5 MiB 128.5 MiB -0.0%
Query 48 11.8 MiB 15.2 MiB +29.6%
Query 49 52.4 MiB 61.1 MiB +16.7%
Query 50 16.3 MiB 16.3 MiB +0.0%
Query 51 177.7 MiB 178.0 MiB +0.2%
Query 52 1.5 MiB 1.5 MiB +0.7%
Query 53 30.6 MiB 30.6 MiB +0.0%
Query 54 2.0 MiB 4.1 MiB +109.6%
Query 55 1.6 MiB 1.7 MiB +4.5%
Query 56 2.7 MiB 3.2 MiB +15.5%
Query 57 120.8 MiB 120.8 MiB -0.0%
Query 58 4.2 MiB 4.4 MiB +5.1%
Query 59 11.6 MiB 11.9 MiB +2.0%
Query 60 5.2 MiB 5.2 MiB -0.5%
Query 61 2.5 KiB 2.4 KiB -3.2%
Query 62 3.3 MiB 3.6 MiB +9.9%
Query 63 20.6 MiB 30.6 MiB +48.6%
Query 64 95.3 MiB 117.2 MiB +23.0%
Query 65 29.5 MiB 29.5 MiB +0.0%
Query 66 15.0 MiB 15.0 MiB +0.0%
Query 67 117.0 MiB 117.0 MiB -0.0%
Query 68 218.6 KiB 96.3 KiB -56.0%
Query 69 3.8 MiB 46.7 MiB +1125.5%
Query 70 32.5 MiB 32.5 MiB +0.0%
Query 71 50.4 MiB 40.6 MiB -19.4%
Query 72 91.2 MiB 91.2 MiB +0.0%
Query 73 96.3 KiB 96.3 KiB +0.0%
Query 74 35.2 MiB 48.9 MiB +38.8%
Query 75 22.0 MiB 22.2 MiB +0.9%
Query 76 4.2 MiB 4.2 MiB +1.8%
Query 77 4.1 MiB 4.1 MiB +0.0%
Query 78 316.9 MiB 316.9 MiB +0.0%
Query 79 8.7 MiB 12.9 MiB +49.1%
Query 80 41.6 MiB 41.2 MiB -1.0%
Query 81 4.1 MiB 11.4 MiB +181.3%
Query 82 4.9 MiB 4.9 MiB +0.0%
Query 83 8.3 MiB 8.3 MiB +0.0%
Query 84 11.6 MiB 12.0 MiB +3.1%
Query 85 17.0 MiB 21.1 MiB +23.7%
Query 86 82.1 MiB 82.1 MiB +0.0%
Query 87 12.9 MiB 20.6 MiB +59.8%
Query 88 1.6 MiB 1.5 MiB -5.3%
Query 89 112.2 MiB 112.0 MiB -0.2%
Query 90 348.6 KiB 348.6 KiB +0.0%
Query 91 31.3 MiB 31.6 MiB +0.9%
Query 92 2.5 MiB 2.7 MiB +8.9%
Query 93 15.4 MiB 15.1 MiB -1.6%
Query 94 3.7 MiB 3.4 MiB -7.6%
Query 95 19.4 MiB 19.4 MiB +0.0%
Query 96 224.7 KiB 224.7 KiB +0.0%
Query 97 33.1 MiB 33.1 MiB +0.0%
Query 98 114.7 MiB 114.7 MiB -0.0%
Query 99 1.7 MiB 2.1 MiB +26.6%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpcds base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 316.9 MiB 2.0 GiB 1.7 GiB 6.3×
tpcds changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 316.9 MiB 2.1 GiB 1.8 GiB 6.7×
Resource Usage

tpcds — base (merge-base)

Metric Value
Wall time 50.0s
Peak memory 2.0 GiB
Avg memory 1.3 GiB
CPU user 190.0s
CPU sys 5.5s
Peak spill 0 B

tpcds — branch

Metric Value
Wall time 45.0s
Peak memory 2.1 GiB
Avg memory 1.5 GiB
CPU user 185.8s
CPU sys 5.4s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃     HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 73.44 ms │                                    74.07 ms │     no change │
│ QQuery 2  │ 22.38 ms │                                    22.16 ms │     no change │
│ QQuery 3  │ 43.62 ms │                                    42.56 ms │     no change │
│ QQuery 4  │ 24.91 ms │                                    25.81 ms │     no change │
│ QQuery 5  │ 53.07 ms │                                    52.85 ms │     no change │
│ QQuery 6  │ 23.14 ms │                                    21.94 ms │ +1.06x faster │
│ QQuery 7  │ 64.10 ms │                                    60.35 ms │ +1.06x faster │
│ QQuery 8  │ 60.96 ms │                                    60.17 ms │     no change │
│ QQuery 9  │ 79.43 ms │                                    79.88 ms │     no change │
│ QQuery 10 │ 80.94 ms │                                    58.94 ms │ +1.37x faster │
│ QQuery 11 │ 16.73 ms │                                    15.67 ms │ +1.07x faster │
│ QQuery 12 │ 29.41 ms │                                    28.36 ms │     no change │
│ QQuery 13 │ 60.58 ms │                                    48.96 ms │ +1.24x faster │
│ QQuery 14 │ 35.62 ms │                                    36.45 ms │     no change │
│ QQuery 15 │ 48.24 ms │                                    49.44 ms │     no change │
│ QQuery 16 │ 15.16 ms │                                    16.89 ms │  1.11x slower │
│ QQuery 17 │ 94.96 ms │                                    94.30 ms │     no change │
│ QQuery 18 │ 86.40 ms │                                    64.01 ms │ +1.35x faster │
│ QQuery 19 │ 43.71 ms │                                    47.99 ms │  1.10x slower │
│ QQuery 20 │ 50.52 ms │                                    45.37 ms │ +1.11x faster │
│ QQuery 21 │ 80.74 ms │                                    89.32 ms │  1.11x slower │
│ QQuery 22 │ 19.26 ms │                                    16.03 ms │ +1.20x faster │
└───────────┴──────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 1107.33ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 1051.52ms │
│ Average Time (HEAD)                                        │   50.33ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   47.80ms │
│ Queries Faster                                             │         8 │
│ Queries Slower                                             │         3 │
│ Queries with No Change                                     │        11 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                           HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │ 73.44 / 76.28 ±3.05 / 82.02 ms │              74.07 / 78.37 ±4.19 / 84.32 ms │     no change │
│ QQuery 2  │ 22.38 / 24.02 ±1.14 / 25.95 ms │              22.16 / 23.62 ±1.63 / 26.28 ms │     no change │
│ QQuery 3  │ 43.62 / 44.72 ±0.66 / 45.63 ms │              42.56 / 44.37 ±1.37 / 46.78 ms │     no change │
│ QQuery 4  │ 24.91 / 26.13 ±0.63 / 26.64 ms │              25.81 / 26.35 ±0.28 / 26.59 ms │     no change │
│ QQuery 5  │ 53.07 / 55.25 ±1.68 / 57.56 ms │              52.85 / 54.67 ±1.47 / 56.82 ms │     no change │
│ QQuery 6  │ 23.14 / 24.45 ±1.37 / 26.97 ms │              21.94 / 23.46 ±0.95 / 24.65 ms │     no change │
│ QQuery 7  │ 64.10 / 64.60 ±0.47 / 65.41 ms │              60.35 / 65.61 ±3.32 / 70.79 ms │     no change │
│ QQuery 8  │ 60.96 / 62.24 ±0.86 / 63.33 ms │              60.17 / 62.01 ±1.96 / 65.25 ms │     no change │
│ QQuery 9  │ 79.43 / 84.79 ±5.17 / 93.74 ms │              79.88 / 84.15 ±4.11 / 91.78 ms │     no change │
│ QQuery 10 │ 80.94 / 84.30 ±2.12 / 87.01 ms │              58.94 / 61.63 ±1.74 / 64.38 ms │ +1.37x faster │
│ QQuery 11 │ 16.73 / 18.10 ±1.06 / 19.57 ms │              15.67 / 16.91 ±1.07 / 18.24 ms │ +1.07x faster │
│ QQuery 12 │ 29.41 / 30.59 ±0.84 / 31.55 ms │              28.36 / 29.10 ±0.46 / 29.71 ms │     no change │
│ QQuery 13 │ 60.58 / 63.98 ±3.22 / 69.61 ms │              48.96 / 52.77 ±5.31 / 63.13 ms │ +1.21x faster │
│ QQuery 14 │ 35.62 / 37.21 ±0.99 / 38.73 ms │              36.45 / 37.70 ±1.36 / 40.04 ms │     no change │
│ QQuery 15 │ 48.24 / 49.50 ±1.06 / 50.99 ms │              49.44 / 50.21 ±0.62 / 51.26 ms │     no change │
│ QQuery 16 │ 15.16 / 16.08 ±0.60 / 16.97 ms │              16.89 / 17.72 ±1.05 / 19.73 ms │  1.10x slower │
│ QQuery 17 │ 94.96 / 95.85 ±0.70 / 96.59 ms │             94.30 / 97.14 ±2.36 / 101.38 ms │     no change │
│ QQuery 18 │ 86.40 / 88.68 ±1.52 / 90.79 ms │              64.01 / 66.47 ±2.56 / 71.26 ms │ +1.33x faster │
│ QQuery 19 │ 43.71 / 46.60 ±1.47 / 47.82 ms │              47.99 / 50.99 ±2.81 / 56.33 ms │  1.09x slower │
│ QQuery 20 │ 50.52 / 51.61 ±1.17 / 53.89 ms │              45.37 / 46.41 ±0.63 / 47.16 ms │ +1.11x faster │
│ QQuery 21 │ 80.74 / 87.02 ±3.77 / 92.19 ms │              89.32 / 90.62 ±1.30 / 92.84 ms │     no change │
│ QQuery 22 │ 19.26 / 20.63 ±1.12 / 22.16 ms │              16.03 / 17.70 ±1.22 / 19.45 ms │ +1.17x faster │
└───────────┴────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 1152.64ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 1097.99ms │
│ Average Time (HEAD)                                        │   52.39ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   49.91ms │
│ Queries Faster                                             │         6 │
│ Queries Slower                                             │         2 │
│ Queries with No Change                                     │        14 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

tpch — tpch_sf1

Query Base Changed Change
Query 1 21.3 MiB 22.8 MiB +6.9%
Query 2 23.1 MiB 24.7 MiB +7.0%
Query 3 11.2 MiB 11.1 MiB -0.2%
Query 4 22.2 MiB 23.5 MiB +6.2%
Query 5 42.8 MiB 42.7 MiB -0.0%
Query 6 832 B 832 B +0.0%
Query 7 108.1 MiB 105.2 MiB -2.6%
Query 8 39.8 MiB 39.8 MiB -0.0%
Query 9 142.0 MiB 147.6 MiB +3.9%
Query 10 50.0 MiB 42.7 MiB -14.6%
Query 11 101.7 MiB 41.2 MiB -59.5%
Query 12 22.5 MiB 15.4 MiB -31.4%
Query 13 72.5 MiB 62.3 MiB -14.0%
Query 14 6.7 MiB 6.7 MiB +0.0%
Query 15 13.5 MiB 13.7 MiB +1.2%
Query 16 42.7 MiB 55.0 MiB +29.0%
Query 17 152.1 MiB 153.3 MiB +0.7%
Query 18 116.2 MiB 167.1 MiB +43.7%
Query 19 209.7 KiB 725.7 KiB +246.1%
Query 20 60.0 MiB 57.7 MiB -3.8%
Query 21 36.1 MiB 36.1 MiB +0.0%
Query 22 41.5 MiB 32.9 MiB -20.7%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpch base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 152.1 MiB 1.1 GiB 1009.5 MiB 7.6×
tpch changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 167.1 MiB 1.1 GiB 978.8 MiB 6.9×
Resource Usage

tpch — base (merge-base)

Metric Value
Wall time 10.0s
Peak memory 1.1 GiB
Avg memory 428.1 MiB
CPU user 36.6s
CPU sys 1.8s
Peak spill 0 B

tpch — branch

Metric Value
Wall time 10.0s
Peak memory 1.1 GiB
Avg memory 429.2 MiB
CPU user 35.4s
CPU sys 2.0s
Peak spill 0 B

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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │    5.43 ms │                                     5.63 ms │     no change │
│ QQuery 2  │   75.16 ms │                                    78.82 ms │     no change │
│ QQuery 3  │   28.92 ms │                                    29.12 ms │     no change │
│ QQuery 4  │  521.99 ms │                                   450.06 ms │ +1.16x faster │
│ QQuery 5  │   50.17 ms │                                    50.66 ms │     no change │
│ QQuery 6  │   33.15 ms │                                    35.27 ms │  1.06x slower │
│ QQuery 7  │   70.60 ms │                                    72.68 ms │     no change │
│ QQuery 8  │   38.63 ms │                                    36.61 ms │ +1.06x faster │
│ QQuery 9  │   52.38 ms │                                    54.17 ms │     no change │
│ QQuery 10 │   58.73 ms │                                    60.23 ms │     no change │
│ QQuery 11 │  344.90 ms │                                   293.88 ms │ +1.17x faster │
│ QQuery 12 │   26.29 ms │                                    27.04 ms │     no change │
│ QQuery 13 │  111.38 ms │                                   113.19 ms │     no change │
│ QQuery 14 │  378.73 ms │                                   375.35 ms │     no change │
│ QQuery 15 │   50.85 ms │                                    53.69 ms │  1.06x slower │
│ QQuery 16 │    6.05 ms │                                     5.98 ms │     no change │
│ QQuery 17 │   76.80 ms │                                    76.44 ms │     no change │
│ QQuery 18 │  103.07 ms │                                   103.52 ms │     no change │
│ QQuery 19 │   41.17 ms │                                    41.20 ms │     no change │
│ QQuery 20 │   30.81 ms │                                    31.24 ms │     no change │
│ QQuery 21 │   17.10 ms │                                    17.12 ms │     no change │
│ QQuery 22 │   60.46 ms │                                    61.23 ms │     no change │
│ QQuery 23 │  290.71 ms │                                   296.33 ms │     no change │
│ QQuery 24 │  198.14 ms │                                   197.10 ms │     no change │
│ QQuery 25 │  104.18 ms │                                   103.41 ms │     no change │
│ QQuery 26 │   44.47 ms │                                    45.86 ms │     no change │
│ QQuery 27 │    5.21 ms │                                     5.24 ms │     no change │
│ QQuery 28 │   61.06 ms │                                    57.02 ms │ +1.07x faster │
│ QQuery 29 │   93.02 ms │                                    92.80 ms │     no change │
│ QQuery 30 │   36.23 ms │                                    33.16 ms │ +1.09x faster │
│ QQuery 31 │  108.86 ms │                                   108.28 ms │     no change │
│ QQuery 32 │   20.26 ms │                                    20.31 ms │     no change │
│ QQuery 33 │   36.91 ms │                                    37.03 ms │     no change │
│ QQuery 34 │    5.56 ms │                                     5.77 ms │     no change │
│ QQuery 35 │   69.28 ms │                                    71.87 ms │     no change │
│ QQuery 36 │    5.80 ms │                                     6.10 ms │  1.05x slower │
│ QQuery 37 │    7.67 ms │                                     7.80 ms │     no change │
│ QQuery 38 │   64.37 ms │                                    61.71 ms │     no change │
│ QQuery 39 │   72.51 ms │                                    74.23 ms │     no change │
│ QQuery 40 │   22.65 ms │                                    23.58 ms │     no change │
│ QQuery 41 │   14.51 ms │                                    14.57 ms │     no change │
│ QQuery 42 │   23.88 ms │                                    24.09 ms │     no change │
│ QQuery 43 │    4.60 ms │                                     4.99 ms │  1.08x slower │
│ QQuery 44 │    5.38 ms │                                     5.65 ms │     no change │
│ QQuery 45 │   34.28 ms │                                    38.73 ms │  1.13x slower │
│ QQuery 46 │   11.20 ms │                                    12.05 ms │  1.08x slower │
│ QQuery 47 │  202.74 ms │                                   209.59 ms │     no change │
│ QQuery 48 │   90.30 ms │                                    91.01 ms │     no change │
│ QQuery 49 │   69.18 ms │                                    70.19 ms │     no change │
│ QQuery 50 │   57.26 ms │                                    56.99 ms │     no change │
│ QQuery 51 │   94.44 ms │                                    94.62 ms │     no change │
│ QQuery 52 │   24.26 ms │                                    24.62 ms │     no change │
│ QQuery 53 │   29.45 ms │                                    29.73 ms │     no change │
│ QQuery 54 │   53.99 ms │                                    54.27 ms │     no change │
│ QQuery 55 │   24.26 ms │                                    24.29 ms │     no change │
│ QQuery 56 │   38.28 ms │                                    38.62 ms │     no change │
│ QQuery 57 │  151.16 ms │                                   153.19 ms │     no change │
│ QQuery 58 │  103.31 ms │                                   103.05 ms │     no change │
│ QQuery 59 │  106.32 ms │                                   114.82 ms │  1.08x slower │
│ QQuery 60 │   38.30 ms │                                    38.68 ms │     no change │
│ QQuery 61 │    9.31 ms │                                     9.15 ms │     no change │
│ QQuery 62 │   38.34 ms │                                    38.11 ms │     no change │
│ QQuery 63 │   29.56 ms │                                    29.67 ms │     no change │
│ QQuery 64 │  354.77 ms │                                   350.77 ms │     no change │
│ QQuery 65 │   70.03 ms │                                    70.49 ms │     no change │
│ QQuery 66 │   72.19 ms │                                    71.98 ms │     no change │
│ QQuery 67 │  186.30 ms │                                   188.84 ms │     no change │
│ QQuery 68 │    6.57 ms │                                     6.81 ms │     no change │
│ QQuery 69 │   53.98 ms │                                    56.09 ms │     no change │
│ QQuery 70 │   95.71 ms │                                    96.57 ms │     no change │
│ QQuery 71 │   35.18 ms │                                    35.35 ms │     no change │
│ QQuery 72 │ 1393.32 ms │                                  1648.74 ms │  1.18x slower │
│ QQuery 73 │    5.42 ms │                                     5.58 ms │     no change │
│ QQuery 74 │  235.69 ms │                                   201.20 ms │ +1.17x faster │
│ QQuery 75 │  138.94 ms │                                   139.54 ms │     no change │
│ QQuery 76 │   34.48 ms │                                    34.54 ms │     no change │
│ QQuery 77 │   59.87 ms │                                    59.62 ms │     no change │
│ QQuery 78 │  266.67 ms │                                   268.02 ms │     no change │
│ QQuery 79 │   67.42 ms │                                    66.11 ms │     no change │
│ QQuery 80 │   94.67 ms │                                    94.71 ms │     no change │
│ QQuery 81 │   28.33 ms │                                    25.96 ms │ +1.09x faster │
│ QQuery 82 │   17.68 ms │                                    17.74 ms │     no change │
│ QQuery 83 │   33.00 ms │                                    33.09 ms │     no change │
│ QQuery 84 │   30.90 ms │                                    29.31 ms │ +1.05x faster │
│ QQuery 85 │   95.32 ms │                                    96.78 ms │     no change │
│ QQuery 86 │   24.39 ms │                                    24.39 ms │     no change │
│ QQuery 87 │   65.22 ms │                                    61.27 ms │ +1.06x faster │
│ QQuery 88 │   59.36 ms │                                    59.94 ms │     no change │
│ QQuery 89 │   36.14 ms │                                    36.18 ms │     no change │
│ QQuery 90 │   16.51 ms │                                    16.76 ms │     no change │
│ QQuery 91 │   39.83 ms │                                    41.29 ms │     no change │
│ QQuery 92 │   29.50 ms │                                    29.41 ms │     no change │
│ QQuery 93 │   48.89 ms │                                    49.13 ms │     no change │
│ QQuery 94 │   38.67 ms │                                    38.51 ms │     no change │
│ QQuery 95 │   79.62 ms │                                    79.79 ms │     no change │
│ QQuery 96 │   24.11 ms │                                    23.93 ms │     no change │
│ QQuery 97 │   50.16 ms │                                    50.69 ms │     no change │
│ QQuery 98 │   38.04 ms │                                    38.21 ms │     no change │
│ QQuery 99 │   56.24 ms │                                    56.87 ms │     no change │
└───────────┴────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 8667.10ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 8805.62ms │
│ Average Time (HEAD)                                        │   87.55ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   88.95ms │
│ Queries Faster                                             │         9 │
│ Queries Slower                                             │         8 │
│ Queries with No Change                                     │        82 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │           5.43 / 6.02 ±0.96 / 7.91 ms │                 5.63 / 6.13 ±0.93 / 7.98 ms │     no change │
│ QQuery 2  │        75.16 / 75.74 ±0.38 / 76.33 ms │              78.82 / 79.15 ±0.21 / 79.42 ms │     no change │
│ QQuery 3  │        28.92 / 29.11 ±0.13 / 29.31 ms │              29.12 / 29.20 ±0.06 / 29.29 ms │     no change │
│ QQuery 4  │     521.99 / 526.99 ±5.80 / 537.45 ms │           450.06 / 459.10 ±7.16 / 467.11 ms │ +1.15x faster │
│ QQuery 5  │        50.17 / 50.69 ±0.33 / 51.20 ms │              50.66 / 51.30 ±0.74 / 52.25 ms │     no change │
│ QQuery 6  │        33.15 / 33.94 ±0.57 / 34.85 ms │              35.27 / 35.66 ±0.43 / 36.47 ms │  1.05x slower │
│ QQuery 7  │        70.60 / 70.81 ±0.16 / 71.08 ms │              72.68 / 73.07 ±0.42 / 73.64 ms │     no change │
│ QQuery 8  │        38.63 / 39.95 ±1.98 / 43.83 ms │              36.61 / 37.00 ±0.36 / 37.61 ms │ +1.08x faster │
│ QQuery 9  │        52.38 / 53.75 ±0.91 / 55.17 ms │              54.17 / 55.40 ±1.29 / 57.87 ms │     no change │
│ QQuery 10 │        58.73 / 59.16 ±0.43 / 59.97 ms │              60.23 / 60.58 ±0.29 / 61.08 ms │     no change │
│ QQuery 11 │     344.90 / 348.26 ±2.71 / 352.41 ms │           293.88 / 298.61 ±3.98 / 303.87 ms │ +1.17x faster │
│ QQuery 12 │        26.29 / 26.38 ±0.07 / 26.47 ms │              27.04 / 28.52 ±1.46 / 30.91 ms │  1.08x slower │
│ QQuery 13 │     111.38 / 112.71 ±1.81 / 116.27 ms │           113.19 / 114.01 ±0.99 / 115.93 ms │     no change │
│ QQuery 14 │     378.73 / 380.39 ±1.72 / 383.33 ms │           375.35 / 380.06 ±2.84 / 382.67 ms │     no change │
│ QQuery 15 │        50.85 / 51.54 ±0.39 / 51.84 ms │              53.69 / 54.46 ±0.47 / 55.15 ms │  1.06x slower │
│ QQuery 16 │           6.05 / 6.17 ±0.16 / 6.48 ms │                 5.98 / 6.14 ±0.21 / 6.55 ms │     no change │
│ QQuery 17 │        76.80 / 77.86 ±1.05 / 79.84 ms │              76.44 / 77.98 ±2.34 / 82.63 ms │     no change │
│ QQuery 18 │     103.07 / 105.53 ±1.84 / 108.26 ms │           103.52 / 104.77 ±0.85 / 105.93 ms │     no change │
│ QQuery 19 │        41.17 / 42.09 ±0.59 / 42.82 ms │              41.20 / 41.79 ±0.41 / 42.25 ms │     no change │
│ QQuery 20 │        30.81 / 31.40 ±0.66 / 32.67 ms │              31.24 / 32.19 ±0.86 / 33.73 ms │     no change │
│ QQuery 21 │        17.10 / 17.42 ±0.26 / 17.75 ms │              17.12 / 17.34 ±0.16 / 17.54 ms │     no change │
│ QQuery 22 │        60.46 / 60.79 ±0.31 / 61.31 ms │              61.23 / 61.69 ±0.24 / 61.91 ms │     no change │
│ QQuery 23 │     290.71 / 294.83 ±4.26 / 302.43 ms │           296.33 / 300.08 ±4.28 / 308.29 ms │     no change │
│ QQuery 24 │     198.14 / 200.70 ±2.17 / 204.11 ms │           197.10 / 204.17 ±6.45 / 216.06 ms │     no change │
│ QQuery 25 │     104.18 / 105.58 ±1.65 / 108.69 ms │           103.41 / 105.55 ±1.75 / 107.91 ms │     no change │
│ QQuery 26 │        44.47 / 44.90 ±0.38 / 45.39 ms │              45.86 / 46.41 ±0.42 / 46.99 ms │     no change │
│ QQuery 27 │           5.21 / 5.34 ±0.17 / 5.68 ms │                 5.24 / 5.38 ±0.15 / 5.63 ms │     no change │
│ QQuery 28 │        61.06 / 61.72 ±0.35 / 62.08 ms │              57.02 / 59.86 ±2.22 / 62.08 ms │     no change │
│ QQuery 29 │        93.02 / 94.18 ±0.82 / 95.11 ms │             92.80 / 95.73 ±3.44 / 101.50 ms │     no change │
│ QQuery 30 │        36.23 / 36.76 ±0.31 / 37.19 ms │              33.16 / 33.78 ±0.79 / 35.33 ms │ +1.09x faster │
│ QQuery 31 │     108.86 / 109.52 ±0.58 / 110.56 ms │           108.28 / 109.29 ±1.04 / 111.10 ms │     no change │
│ QQuery 32 │        20.26 / 21.64 ±2.40 / 26.42 ms │              20.31 / 20.49 ±0.22 / 20.92 ms │ +1.06x faster │
│ QQuery 33 │        36.91 / 38.05 ±1.42 / 40.53 ms │              37.03 / 37.19 ±0.14 / 37.37 ms │     no change │
│ QQuery 34 │           5.56 / 5.72 ±0.16 / 6.01 ms │                 5.77 / 5.93 ±0.19 / 6.25 ms │     no change │
│ QQuery 35 │        69.28 / 69.61 ±0.20 / 69.82 ms │              71.87 / 72.78 ±0.56 / 73.39 ms │     no change │
│ QQuery 36 │           5.80 / 6.02 ±0.19 / 6.32 ms │                 6.10 / 6.27 ±0.16 / 6.53 ms │     no change │
│ QQuery 37 │           7.67 / 7.84 ±0.10 / 7.93 ms │                7.80 / 8.50 ±1.04 / 10.51 ms │  1.08x slower │
│ QQuery 38 │        64.37 / 64.73 ±0.25 / 65.10 ms │              61.71 / 62.24 ±0.62 / 63.40 ms │     no change │
│ QQuery 39 │        72.51 / 73.33 ±0.72 / 74.62 ms │              74.23 / 75.46 ±1.52 / 78.39 ms │     no change │
│ QQuery 40 │        22.65 / 23.29 ±0.61 / 24.44 ms │              23.58 / 24.16 ±0.49 / 24.98 ms │     no change │
│ QQuery 41 │        14.51 / 14.61 ±0.15 / 14.89 ms │              14.57 / 14.83 ±0.17 / 15.03 ms │     no change │
│ QQuery 42 │        23.88 / 24.26 ±0.37 / 24.93 ms │              24.09 / 24.53 ±0.51 / 25.44 ms │     no change │
│ QQuery 43 │           4.60 / 4.88 ±0.31 / 5.46 ms │                 4.99 / 5.11 ±0.14 / 5.38 ms │     no change │
│ QQuery 44 │           5.38 / 5.44 ±0.08 / 5.59 ms │                 5.65 / 5.75 ±0.07 / 5.87 ms │  1.06x slower │
│ QQuery 45 │        34.28 / 34.91 ±0.51 / 35.84 ms │              38.73 / 39.27 ±0.39 / 39.92 ms │  1.12x slower │
│ QQuery 46 │        11.20 / 11.37 ±0.20 / 11.73 ms │              12.05 / 12.34 ±0.37 / 13.06 ms │  1.09x slower │
│ QQuery 47 │     202.74 / 206.26 ±3.51 / 212.75 ms │           209.59 / 211.20 ±2.42 / 215.99 ms │     no change │
│ QQuery 48 │        90.30 / 90.93 ±0.72 / 92.31 ms │              91.01 / 93.04 ±2.26 / 97.43 ms │     no change │
│ QQuery 49 │        69.18 / 69.88 ±0.67 / 71.03 ms │              70.19 / 71.35 ±1.07 / 72.71 ms │     no change │
│ QQuery 50 │        57.26 / 58.94 ±3.20 / 65.35 ms │              56.99 / 59.37 ±4.13 / 67.62 ms │     no change │
│ QQuery 51 │        94.44 / 95.68 ±0.84 / 96.96 ms │              94.62 / 95.68 ±0.94 / 97.25 ms │     no change │
│ QQuery 52 │        24.26 / 24.57 ±0.39 / 25.33 ms │              24.62 / 25.06 ±0.73 / 26.51 ms │     no change │
│ QQuery 53 │        29.45 / 29.63 ±0.10 / 29.72 ms │              29.73 / 29.89 ±0.13 / 30.06 ms │     no change │
│ QQuery 54 │        53.99 / 55.43 ±2.23 / 59.86 ms │              54.27 / 55.94 ±2.70 / 61.32 ms │     no change │
│ QQuery 55 │        24.26 / 24.83 ±0.70 / 26.15 ms │              24.29 / 24.69 ±0.34 / 25.20 ms │     no change │
│ QQuery 56 │        38.28 / 38.75 ±0.33 / 39.28 ms │              38.62 / 38.84 ±0.25 / 39.28 ms │     no change │
│ QQuery 57 │     151.16 / 152.53 ±1.18 / 154.72 ms │           153.19 / 155.90 ±2.95 / 161.43 ms │     no change │
│ QQuery 58 │     103.31 / 104.33 ±1.32 / 106.80 ms │           103.05 / 104.97 ±1.66 / 107.46 ms │     no change │
│ QQuery 59 │     106.32 / 107.73 ±1.89 / 111.40 ms │           114.82 / 116.34 ±2.48 / 121.26 ms │  1.08x slower │
│ QQuery 60 │        38.30 / 39.12 ±0.64 / 40.24 ms │              38.68 / 39.42 ±0.66 / 40.45 ms │     no change │
│ QQuery 61 │           9.31 / 9.43 ±0.17 / 9.76 ms │                 9.15 / 9.45 ±0.20 / 9.77 ms │     no change │
│ QQuery 62 │        38.34 / 38.83 ±0.34 / 39.33 ms │              38.11 / 38.87 ±1.10 / 41.03 ms │     no change │
│ QQuery 63 │        29.56 / 29.91 ±0.18 / 30.04 ms │              29.67 / 30.00 ±0.31 / 30.49 ms │     no change │
│ QQuery 64 │     354.77 / 356.85 ±1.69 / 358.82 ms │           350.77 / 355.46 ±4.18 / 362.83 ms │     no change │
│ QQuery 65 │        70.03 / 71.00 ±1.00 / 72.71 ms │              70.49 / 70.76 ±0.22 / 71.13 ms │     no change │
│ QQuery 66 │        72.19 / 74.57 ±3.79 / 82.13 ms │              71.98 / 74.79 ±3.52 / 81.68 ms │     no change │
│ QQuery 67 │     186.30 / 192.51 ±3.99 / 196.85 ms │           188.84 / 192.32 ±4.49 / 201.18 ms │     no change │
│ QQuery 68 │           6.57 / 6.73 ±0.21 / 7.14 ms │                 6.81 / 6.90 ±0.06 / 6.99 ms │     no change │
│ QQuery 69 │        53.98 / 54.85 ±1.05 / 56.88 ms │              56.09 / 56.67 ±0.69 / 58.01 ms │     no change │
│ QQuery 70 │       95.71 / 99.24 ±4.21 / 107.48 ms │             96.57 / 98.17 ±2.29 / 102.65 ms │     no change │
│ QQuery 71 │        35.18 / 35.77 ±0.83 / 37.39 ms │              35.35 / 35.91 ±0.53 / 36.87 ms │     no change │
│ QQuery 72 │ 1393.32 / 1472.05 ±99.50 / 1654.24 ms │       1648.74 / 1703.14 ±40.31 / 1765.05 ms │  1.16x slower │
│ QQuery 73 │           5.42 / 5.58 ±0.19 / 5.93 ms │                 5.58 / 5.72 ±0.12 / 5.94 ms │     no change │
│ QQuery 74 │     235.69 / 239.95 ±2.33 / 242.68 ms │           201.20 / 206.76 ±6.33 / 218.14 ms │ +1.16x faster │
│ QQuery 75 │     138.94 / 139.47 ±0.64 / 140.44 ms │           139.54 / 141.52 ±2.04 / 145.44 ms │     no change │
│ QQuery 76 │        34.48 / 34.97 ±0.62 / 36.19 ms │              34.54 / 34.82 ±0.19 / 35.12 ms │     no change │
│ QQuery 77 │        59.87 / 62.18 ±3.69 / 69.52 ms │              59.62 / 60.21 ±0.58 / 61.19 ms │     no change │
│ QQuery 78 │    266.67 / 283.28 ±23.29 / 328.99 ms │           268.02 / 274.51 ±8.04 / 289.29 ms │     no change │
│ QQuery 79 │        67.42 / 67.96 ±0.41 / 68.64 ms │              66.11 / 67.57 ±2.31 / 72.16 ms │     no change │
│ QQuery 80 │       94.67 / 97.89 ±4.84 / 107.50 ms │              94.71 / 96.12 ±1.13 / 97.54 ms │     no change │
│ QQuery 81 │        28.33 / 28.82 ±0.27 / 29.09 ms │              25.96 / 27.10 ±1.66 / 30.40 ms │ +1.06x faster │
│ QQuery 82 │        17.68 / 17.90 ±0.17 / 18.13 ms │              17.74 / 17.91 ±0.11 / 18.06 ms │     no change │
│ QQuery 83 │        33.00 / 33.42 ±0.28 / 33.89 ms │              33.09 / 33.82 ±1.28 / 36.36 ms │     no change │
│ QQuery 84 │        30.90 / 31.00 ±0.10 / 31.18 ms │              29.31 / 29.75 ±0.33 / 30.12 ms │     no change │
│ QQuery 85 │       95.32 / 99.37 ±3.99 / 105.57 ms │             96.78 / 99.10 ±1.94 / 102.41 ms │     no change │
│ QQuery 86 │        24.39 / 24.82 ±0.29 / 25.23 ms │              24.39 / 24.77 ±0.37 / 25.21 ms │     no change │
│ QQuery 87 │        65.22 / 65.51 ±0.29 / 65.95 ms │              61.27 / 62.33 ±0.83 / 63.23 ms │     no change │
│ QQuery 88 │        59.36 / 61.25 ±2.59 / 66.34 ms │              59.94 / 60.27 ±0.21 / 60.54 ms │     no change │
│ QQuery 89 │        36.14 / 37.24 ±1.53 / 40.28 ms │              36.18 / 37.12 ±0.60 / 37.82 ms │     no change │
│ QQuery 90 │        16.51 / 16.83 ±0.18 / 17.01 ms │              16.76 / 16.83 ±0.09 / 17.00 ms │     no change │
│ QQuery 91 │        39.83 / 40.30 ±0.32 / 40.83 ms │              41.29 / 41.87 ±0.32 / 42.28 ms │     no change │
│ QQuery 92 │        29.50 / 29.86 ±0.30 / 30.25 ms │              29.41 / 30.67 ±1.42 / 33.30 ms │     no change │
│ QQuery 93 │        48.89 / 49.93 ±0.89 / 51.29 ms │              49.13 / 49.85 ±0.63 / 50.75 ms │     no change │
│ QQuery 94 │        38.67 / 40.26 ±1.85 / 43.77 ms │              38.51 / 40.13 ±1.86 / 43.76 ms │     no change │
│ QQuery 95 │        79.62 / 80.50 ±0.59 / 81.19 ms │              79.79 / 80.84 ±0.77 / 81.58 ms │     no change │
│ QQuery 96 │        24.11 / 24.17 ±0.05 / 24.23 ms │              23.93 / 24.81 ±1.12 / 27.03 ms │     no change │
│ QQuery 97 │        50.16 / 50.92 ±1.07 / 53.05 ms │              50.69 / 51.69 ±1.23 / 54.09 ms │     no change │
│ QQuery 98 │        38.04 / 39.08 ±1.29 / 41.63 ms │              38.21 / 38.63 ±0.35 / 39.22 ms │     no change │
│ QQuery 99 │        56.24 / 57.33 ±1.14 / 59.42 ms │              56.87 / 58.42 ±2.51 / 63.42 ms │     no change │
└───────────┴───────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 8866.06ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 8986.49ms │
│ Average Time (HEAD)                                        │   89.56ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   90.77ms │
│ Queries Faster                                             │         7 │
│ Queries Slower                                             │         9 │
│ Queries with No Change                                     │        83 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

tpcds — tpcds_sf1

Query Base Changed Change
Query 1 135.7 KiB 135.7 KiB -0.0%
Query 2 80.1 MiB 80.4 MiB +0.5%
Query 3 2.1 MiB 1.9 MiB -8.6%
Query 4 85.6 MiB 111.4 MiB +30.2%
Query 5 12.5 MiB 12.5 MiB +0.1%
Query 6 4.4 MiB 7.6 MiB +72.7%
Query 7 3.6 MiB 3.6 MiB +0.1%
Query 8 2.0 MiB 3.0 MiB +49.3%
Query 9 32.1 KiB 32.1 KiB +0.0%
Query 10 2.0 MiB 3.1 MiB +53.3%
Query 11 58.1 MiB 77.1 MiB +32.8%
Query 12 114.4 MiB 114.4 MiB +0.0%
Query 13 13.7 MiB 17.6 MiB +28.4%
Query 14 170.9 MiB 170.9 MiB +0.0%
Query 15 4.2 MiB 6.8 MiB +62.5%
Query 16 118.6 KiB 161.5 KiB +36.2%
Query 17 39.7 MiB 39.3 MiB -0.8%
Query 18 21.3 MiB 21.0 MiB -1.3%
Query 19 2.0 MiB 1.9 MiB -6.3%
Query 20 114.6 MiB 114.6 MiB -0.0%
Query 21 3.2 MiB 3.2 MiB +0.0%
Query 22 61.8 MiB 61.8 MiB +0.0%
Query 23 212.5 MiB 217.4 MiB +2.3%
Query 24 36.4 MiB 45.7 MiB +25.3%
Query 25 85.6 MiB 86.6 MiB +1.2%
Query 26 3.8 MiB 3.9 MiB +2.1%
Query 27 929.6 KiB 929.6 KiB +0.0%
Query 28 1.3 MiB 1.5 MiB +9.9%
Query 29 97.1 MiB 97.7 MiB +0.6%
Query 30 3.8 MiB 7.0 MiB +85.3%
Query 31 13.4 MiB 18.0 MiB +34.3%
Query 32 2.7 MiB 2.7 MiB +0.0%
Query 33 4.2 MiB 4.0 MiB -3.7%
Query 34 96.3 KiB 96.3 KiB +0.0%
Query 35 16.6 MiB 25.5 MiB +53.2%
Query 36 10.3 MiB 10.3 MiB +0.0%
Query 37 3.3 MiB 3.3 MiB +0.0%
Query 38 10.9 MiB 19.7 MiB +80.3%
Query 39 52.0 MiB 41.1 MiB -20.9%
Query 40 11.5 MiB 11.7 MiB +2.5%
Query 41 2.1 MiB 2.1 MiB +0.0%
Query 42 1.3 MiB 1.3 MiB -0.8%
Query 43 275.4 KiB 258.2 KiB -6.2%
Query 44 97.6 KiB 192.4 KiB +97.2%
Query 45 4.4 MiB 8.2 MiB +87.3%
Query 46 5.4 MiB 9.4 MiB +73.5%
Query 47 138.2 MiB 128.5 MiB -7.0%
Query 48 11.8 MiB 15.2 MiB +29.6%
Query 49 52.8 MiB 62.0 MiB +17.4%
Query 50 16.3 MiB 16.3 MiB +0.0%
Query 51 177.7 MiB 178.0 MiB +0.2%
Query 52 1.5 MiB 1.5 MiB +0.6%
Query 53 20.6 MiB 20.6 MiB +0.0%
Query 54 2.0 MiB 4.1 MiB +109.6%
Query 55 1.7 MiB 1.4 MiB -16.6%
Query 56 3.2 MiB 2.8 MiB -10.0%
Query 57 120.8 MiB 120.8 MiB +0.0%
Query 58 4.4 MiB 4.4 MiB +0.0%
Query 59 11.6 MiB 11.9 MiB +2.0%
Query 60 5.4 MiB 5.3 MiB -3.3%
Query 61 2.4 KiB 2.4 KiB +0.0%
Query 62 3.3 MiB 3.7 MiB +10.0%
Query 63 30.6 MiB 30.3 MiB -1.0%
Query 64 106.2 MiB 127.2 MiB +19.8%
Query 65 29.5 MiB 29.5 MiB -0.0%
Query 66 11.9 MiB 15.0 MiB +26.1%
Query 67 117.0 MiB 117.0 MiB +0.0%
Query 68 218.6 KiB 96.4 KiB -55.9%
Query 69 4.1 MiB 46.7 MiB +1045.6%
Query 70 32.5 MiB 32.5 MiB +0.0%
Query 71 40.4 MiB 40.4 MiB -0.1%
Query 72 91.2 MiB 91.2 MiB +0.0%
Query 73 96.3 KiB 96.3 KiB +0.0%
Query 74 35.2 MiB 49.4 MiB +40.4%
Query 75 23.0 MiB 22.6 MiB -1.6%
Query 76 4.2 MiB 4.2 MiB +1.1%
Query 77 4.1 MiB 4.1 MiB +0.0%
Query 78 316.9 MiB 316.9 MiB +0.0%
Query 79 8.0 MiB 12.1 MiB +51.6%
Query 80 41.6 MiB 42.0 MiB +1.0%
Query 81 4.1 MiB 11.4 MiB +180.7%
Query 82 4.9 MiB 4.9 MiB +0.0%
Query 83 8.3 MiB 8.2 MiB -1.3%
Query 84 11.6 MiB 12.0 MiB +3.1%
Query 85 17.0 MiB 21.1 MiB +23.7%
Query 86 82.1 MiB 82.1 MiB +0.0%
Query 87 12.9 MiB 20.6 MiB +59.8%
Query 88 1.6 MiB 1.6 MiB -4.0%
Query 89 112.2 MiB 112.2 MiB +0.0%
Query 90 348.6 KiB 357.5 KiB +2.5%
Query 91 31.3 MiB 21.6 MiB -31.0%
Query 92 2.7 MiB 2.7 MiB +0.0%
Query 93 15.1 MiB 15.1 MiB +0.0%
Query 94 3.4 MiB 3.7 MiB +8.2%
Query 95 19.4 MiB 19.4 MiB +0.0%
Query 96 224.7 KiB 224.7 KiB +0.0%
Query 97 33.1 MiB 33.1 MiB -0.2%
Query 98 114.7 MiB 114.7 MiB +0.0%
Query 99 1.7 MiB 1.7 MiB +0.5%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpcds base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 316.9 MiB 2.2 GiB 1.9 GiB 7.1×
tpcds changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 316.9 MiB 2.2 GiB 1.9 GiB 7.1×
Resource Usage

tpcds — base (merge-base)

Metric Value
Wall time 45.0s
Peak memory 2.2 GiB
Avg memory 1.6 GiB
CPU user 178.1s
CPU sys 5.3s
Peak spill 0 B

tpcds — branch

Metric Value
Wall time 50.0s
Peak memory 2.2 GiB
Avg memory 1.4 GiB
CPU user 186.2s
CPU sys 5.5s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991648869-5460-jg95k 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │    5.47 ms │                                     6.01 ms │  1.10x slower │
│ QQuery 2  │   85.81 ms │                                    93.14 ms │  1.09x slower │
│ QQuery 3  │   35.41 ms │                                    35.61 ms │     no change │
│ QQuery 4  │  575.83 ms │                                   483.00 ms │ +1.19x faster │
│ QQuery 5  │   58.51 ms │                                    61.27 ms │     no change │
│ QQuery 6  │   38.77 ms │                                    40.59 ms │     no change │
│ QQuery 7  │   87.41 ms │                                    89.16 ms │     no change │
│ QQuery 8  │   44.52 ms │                                    44.18 ms │     no change │
│ QQuery 9  │   83.74 ms │                                    79.69 ms │     no change │
│ QQuery 10 │   66.34 ms │                                    70.03 ms │  1.06x slower │
│ QQuery 11 │  360.96 ms │                                   295.01 ms │ +1.22x faster │
│ QQuery 12 │   32.49 ms │                                    32.27 ms │     no change │
│ QQuery 13 │  134.50 ms │                                   138.86 ms │     no change │
│ QQuery 14 │  516.65 ms │                                   516.57 ms │     no change │
│ QQuery 15 │   52.24 ms │                                    56.18 ms │  1.08x slower │
│ QQuery 16 │    5.98 ms │                                     6.12 ms │     no change │
│ QQuery 17 │   97.57 ms │                                   101.79 ms │     no change │
│ QQuery 18 │  120.91 ms │                                   120.72 ms │     no change │
│ QQuery 19 │   50.70 ms │                                    52.22 ms │     no change │
│ QQuery 20 │   37.43 ms │                                    38.66 ms │     no change │
│ QQuery 21 │   21.90 ms │                                    21.94 ms │     no change │
│ QQuery 22 │   95.61 ms │                                    97.15 ms │     no change │
│ QQuery 23 │  396.08 ms │                                   391.36 ms │     no change │
│ QQuery 24 │  248.05 ms │                                   258.91 ms │     no change │
│ QQuery 25 │  138.11 ms │                                   138.06 ms │     no change │
│ QQuery 26 │   53.65 ms │                                    56.05 ms │     no change │
│ QQuery 27 │    5.70 ms │                                     5.37 ms │ +1.06x faster │
│ QQuery 28 │   94.61 ms │                                    85.83 ms │ +1.10x faster │
│ QQuery 29 │  126.74 ms │                                   128.82 ms │     no change │
│ QQuery 30 │   43.35 ms │                                    38.01 ms │ +1.14x faster │
│ QQuery 31 │  138.23 ms │                                   131.89 ms │     no change │
│ QQuery 32 │   23.75 ms │                                    24.40 ms │     no change │
│ QQuery 33 │   43.69 ms │                                    45.53 ms │     no change │
│ QQuery 34 │    5.69 ms │                                     5.75 ms │     no change │
│ QQuery 35 │   76.11 ms │                                    80.53 ms │  1.06x slower │
│ QQuery 36 │    6.13 ms │                                     6.09 ms │     no change │
│ QQuery 37 │    8.03 ms │                                     8.22 ms │     no change │
│ QQuery 38 │   72.34 ms │                                    74.22 ms │     no change │
│ QQuery 39 │   88.83 ms │                                    93.85 ms │  1.06x slower │
│ QQuery 40 │   27.31 ms │                                    29.88 ms │  1.09x slower │
│ QQuery 41 │   15.23 ms │                                    15.07 ms │     no change │
│ QQuery 42 │   29.11 ms │                                    29.92 ms │     no change │
│ QQuery 43 │    4.83 ms │                                     4.90 ms │     no change │
│ QQuery 44 │    5.87 ms │                                     5.94 ms │     no change │
│ QQuery 45 │   38.39 ms │                                    41.37 ms │  1.08x slower │
│ QQuery 46 │   12.50 ms │                                    14.31 ms │  1.15x slower │
│ QQuery 47 │  248.92 ms │                                   268.70 ms │  1.08x slower │
│ QQuery 48 │  109.24 ms │                                   112.62 ms │     no change │
│ QQuery 49 │   86.02 ms │                                    91.82 ms │  1.07x slower │
│ QQuery 50 │   73.25 ms │                                    76.57 ms │     no change │
│ QQuery 51 │  121.99 ms │                                   121.41 ms │     no change │
│ QQuery 52 │   30.37 ms │                                    31.83 ms │     no change │
│ QQuery 53 │   36.50 ms │                                    38.56 ms │  1.06x slower │
│ QQuery 54 │   59.97 ms │                                    63.74 ms │  1.06x slower │
│ QQuery 55 │   31.38 ms │                                    30.90 ms │     no change │
│ QQuery 56 │   46.21 ms │                                    47.95 ms │     no change │
│ QQuery 57 │  182.31 ms │                                   185.17 ms │     no change │
│ QQuery 58 │  118.66 ms │                                   121.28 ms │     no change │
│ QQuery 59 │  128.94 ms │                                   142.99 ms │  1.11x slower │
│ QQuery 60 │   47.54 ms │                                    47.74 ms │     no change │
│ QQuery 61 │    9.02 ms │                                     9.01 ms │     no change │
│ QQuery 62 │   44.07 ms │                                    44.52 ms │     no change │
│ QQuery 63 │   37.78 ms │                                    37.89 ms │     no change │
│ QQuery 64 │  435.21 ms │                                   440.88 ms │     no change │
│ QQuery 65 │   90.08 ms │                                    91.01 ms │     no change │
│ QQuery 66 │   80.35 ms │                                    83.09 ms │     no change │
│ QQuery 67 │  209.82 ms │                                   223.17 ms │  1.06x slower │
│ QQuery 68 │    6.68 ms │                                     6.46 ms │     no change │
│ QQuery 69 │   61.99 ms │                                    64.42 ms │     no change │
│ QQuery 70 │  114.30 ms │                                   114.67 ms │     no change │
│ QQuery 71 │   41.65 ms │                                    44.13 ms │  1.06x slower │
│ QQuery 72 │ 2153.11 ms │                                  2123.89 ms │     no change │
│ QQuery 73 │    5.54 ms │                                     5.35 ms │     no change │
│ QQuery 74 │  261.47 ms │                                   205.01 ms │ +1.28x faster │
│ QQuery 75 │  175.54 ms │                                   174.49 ms │     no change │
│ QQuery 76 │   42.63 ms │                                    47.96 ms │  1.12x slower │
│ QQuery 77 │   70.74 ms │                                    71.26 ms │     no change │
│ QQuery 78 │  371.70 ms │                                   370.03 ms │     no change │
│ QQuery 79 │   78.72 ms │                                    78.84 ms │     no change │
│ QQuery 80 │  127.61 ms │                                   125.14 ms │     no change │
│ QQuery 81 │   34.24 ms │                                    30.47 ms │ +1.12x faster │
│ QQuery 82 │   24.37 ms │                                    21.73 ms │ +1.12x faster │
│ QQuery 83 │   35.05 ms │                                    35.64 ms │     no change │
│ QQuery 84 │   36.13 ms │                                    37.39 ms │     no change │
│ QQuery 85 │  116.92 ms │                                   118.75 ms │     no change │
│ QQuery 86 │   29.97 ms │                                    29.83 ms │     no change │
│ QQuery 87 │   75.12 ms │                                    71.81 ms │     no change │
│ QQuery 88 │   87.56 ms │                                    89.39 ms │     no change │
│ QQuery 89 │   44.34 ms │                                    44.21 ms │     no change │
│ QQuery 90 │   18.68 ms │                                    18.33 ms │     no change │
│ QQuery 91 │   46.46 ms │                                    47.34 ms │     no change │
│ QQuery 92 │   37.77 ms │                                    37.49 ms │     no change │
│ QQuery 93 │   69.90 ms │                                    70.73 ms │     no change │
│ QQuery 94 │   47.19 ms │                                    47.66 ms │     no change │
│ QQuery 95 │  101.42 ms │                                   101.75 ms │     no change │
│ QQuery 96 │   28.05 ms │                                    27.80 ms │     no change │
│ QQuery 97 │   67.95 ms │                                    65.77 ms │     no change │
│ QQuery 98 │   46.93 ms │                                    47.18 ms │     no change │
│ QQuery 99 │   64.88 ms │                                    65.44 ms │     no change │
└───────────┴────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 11063.32ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 10945.67ms │
│ Average Time (HEAD)                                        │   111.75ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   110.56ms │
│ Queries Faster                                             │          8 │
│ Queries Slower                                             │         17 │
│ Queries with No Change                                     │         74 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │          5.47 / 7.10 ±2.29 / 11.61 ms │                6.01 / 7.12 ±1.97 / 11.05 ms │     no change │
│ QQuery 2  │       85.81 / 91.29 ±7.90 / 106.96 ms │              93.14 / 95.02 ±1.67 / 97.68 ms │     no change │
│ QQuery 3  │        35.41 / 38.76 ±5.04 / 48.78 ms │              35.61 / 40.76 ±5.77 / 49.14 ms │  1.05x slower │
│ QQuery 4  │    575.83 / 599.12 ±18.48 / 625.97 ms │          483.00 / 500.92 ±10.37 / 513.93 ms │ +1.20x faster │
│ QQuery 5  │        58.51 / 67.65 ±5.29 / 73.80 ms │              61.27 / 64.48 ±2.90 / 68.63 ms │     no change │
│ QQuery 6  │        38.77 / 42.68 ±5.91 / 54.43 ms │              40.59 / 43.93 ±5.60 / 55.08 ms │     no change │
│ QQuery 7  │       87.41 / 95.45 ±8.79 / 107.03 ms │             89.16 / 96.70 ±7.72 / 109.30 ms │     no change │
│ QQuery 8  │        44.52 / 52.45 ±5.81 / 58.06 ms │              44.18 / 44.79 ±0.47 / 45.42 ms │ +1.17x faster │
│ QQuery 9  │        83.74 / 85.88 ±1.59 / 87.98 ms │              79.69 / 84.38 ±2.99 / 88.79 ms │     no change │
│ QQuery 10 │        66.34 / 71.41 ±4.93 / 78.94 ms │              70.03 / 76.71 ±3.88 / 82.01 ms │  1.07x slower │
│ QQuery 11 │    360.96 / 380.36 ±12.75 / 399.93 ms │          295.01 / 309.35 ±10.41 / 323.66 ms │ +1.23x faster │
│ QQuery 12 │        32.49 / 33.37 ±0.87 / 34.97 ms │              32.27 / 33.14 ±0.78 / 34.53 ms │     no change │
│ QQuery 13 │     134.50 / 144.30 ±8.24 / 155.09 ms │           138.86 / 146.85 ±7.56 / 160.16 ms │     no change │
│ QQuery 14 │    516.65 / 532.92 ±10.13 / 546.05 ms │          516.57 / 531.84 ±12.34 / 553.43 ms │     no change │
│ QQuery 15 │        52.24 / 52.94 ±0.45 / 53.52 ms │              56.18 / 60.95 ±4.88 / 67.85 ms │  1.15x slower │
│ QQuery 16 │           5.98 / 6.47 ±0.35 / 7.04 ms │                 6.12 / 6.59 ±0.41 / 7.24 ms │     no change │
│ QQuery 17 │      97.57 / 102.86 ±4.72 / 111.65 ms │          101.79 / 108.96 ±10.74 / 130.32 ms │  1.06x slower │
│ QQuery 18 │     120.91 / 127.97 ±9.89 / 147.56 ms │           120.72 / 127.88 ±6.20 / 136.48 ms │     no change │
│ QQuery 19 │        50.70 / 54.80 ±5.36 / 65.23 ms │              52.22 / 54.82 ±4.77 / 64.36 ms │     no change │
│ QQuery 20 │        37.43 / 39.46 ±1.20 / 40.94 ms │              38.66 / 39.05 ±0.35 / 39.55 ms │     no change │
│ QQuery 21 │        21.90 / 22.47 ±0.48 / 23.23 ms │              21.94 / 22.75 ±0.51 / 23.52 ms │     no change │
│ QQuery 22 │       95.61 / 98.18 ±2.35 / 101.26 ms │             97.15 / 99.29 ±2.77 / 104.38 ms │     no change │
│ QQuery 23 │     396.08 / 402.59 ±6.06 / 413.18 ms │          391.36 / 412.19 ±14.13 / 431.70 ms │     no change │
│ QQuery 24 │    248.05 / 265.38 ±11.89 / 279.88 ms │          258.91 / 277.70 ±13.09 / 299.24 ms │     no change │
│ QQuery 25 │     138.11 / 143.39 ±6.76 / 156.03 ms │           138.06 / 144.51 ±4.97 / 152.82 ms │     no change │
│ QQuery 26 │        53.65 / 56.93 ±4.38 / 65.38 ms │              56.05 / 56.69 ±0.76 / 58.17 ms │     no change │
│ QQuery 27 │           5.70 / 5.98 ±0.29 / 6.41 ms │                 5.37 / 5.77 ±0.36 / 6.43 ms │     no change │
│ QQuery 28 │       94.61 / 97.41 ±2.43 / 100.43 ms │            85.83 / 98.03 ±10.26 / 112.94 ms │     no change │
│ QQuery 29 │     126.74 / 133.11 ±8.93 / 150.73 ms │           128.82 / 131.19 ±2.32 / 134.56 ms │     no change │
│ QQuery 30 │        43.35 / 43.81 ±0.25 / 44.12 ms │              38.01 / 39.46 ±1.57 / 42.37 ms │ +1.11x faster │
│ QQuery 31 │     138.23 / 142.11 ±3.35 / 147.28 ms │          131.89 / 144.37 ±10.24 / 157.46 ms │     no change │
│ QQuery 32 │        23.75 / 24.42 ±0.51 / 25.17 ms │              24.40 / 24.82 ±0.48 / 25.45 ms │     no change │
│ QQuery 33 │        43.69 / 50.44 ±4.89 / 58.17 ms │              45.53 / 48.41 ±4.39 / 57.14 ms │     no change │
│ QQuery 34 │           5.69 / 5.88 ±0.22 / 6.31 ms │                 5.75 / 6.11 ±0.48 / 7.05 ms │     no change │
│ QQuery 35 │        76.11 / 80.79 ±4.15 / 86.89 ms │              80.53 / 87.07 ±4.30 / 93.17 ms │  1.08x slower │
│ QQuery 36 │           6.13 / 6.51 ±0.64 / 7.79 ms │                 6.09 / 6.49 ±0.57 / 7.60 ms │     no change │
│ QQuery 37 │           8.03 / 8.34 ±0.31 / 8.85 ms │                8.22 / 9.21 ±0.82 / 10.57 ms │  1.10x slower │
│ QQuery 38 │        72.34 / 78.31 ±5.70 / 87.28 ms │              74.22 / 78.97 ±4.47 / 86.23 ms │     no change │
│ QQuery 39 │        88.83 / 92.02 ±2.67 / 95.57 ms │             93.85 / 98.58 ±2.66 / 102.09 ms │  1.07x slower │
│ QQuery 40 │        27.31 / 29.22 ±1.54 / 31.47 ms │              29.88 / 31.47 ±1.21 / 33.18 ms │  1.08x slower │
│ QQuery 41 │        15.23 / 16.20 ±1.43 / 19.03 ms │              15.07 / 15.97 ±0.58 / 16.75 ms │     no change │
│ QQuery 42 │        29.11 / 31.98 ±2.71 / 35.53 ms │              29.92 / 32.23 ±3.23 / 38.54 ms │     no change │
│ QQuery 43 │           4.83 / 4.96 ±0.10 / 5.08 ms │                 4.90 / 5.19 ±0.39 / 5.95 ms │     no change │
│ QQuery 44 │           5.87 / 6.07 ±0.23 / 6.36 ms │                 5.94 / 6.02 ±0.13 / 6.28 ms │     no change │
│ QQuery 45 │        38.39 / 38.93 ±0.43 / 39.68 ms │              41.37 / 42.02 ±0.56 / 43.02 ms │  1.08x slower │
│ QQuery 46 │        12.50 / 12.96 ±0.43 / 13.73 ms │              14.31 / 14.98 ±0.65 / 16.17 ms │  1.16x slower │
│ QQuery 47 │    248.92 / 270.54 ±15.54 / 293.15 ms │          268.70 / 289.10 ±25.53 / 338.85 ms │  1.07x slower │
│ QQuery 48 │     109.24 / 120.76 ±7.98 / 130.29 ms │           112.62 / 121.69 ±9.22 / 134.56 ms │     no change │
│ QQuery 49 │        86.02 / 92.48 ±4.66 / 98.90 ms │             91.82 / 98.90 ±5.09 / 105.45 ms │  1.07x slower │
│ QQuery 50 │        73.25 / 83.27 ±8.67 / 96.93 ms │              76.57 / 80.40 ±6.05 / 92.46 ms │     no change │
│ QQuery 51 │     121.99 / 127.45 ±3.62 / 131.98 ms │           121.41 / 128.38 ±6.16 / 139.16 ms │     no change │
│ QQuery 52 │        30.37 / 32.36 ±3.05 / 38.32 ms │              31.83 / 34.03 ±3.36 / 40.73 ms │  1.05x slower │
│ QQuery 53 │        36.50 / 40.12 ±4.21 / 47.34 ms │              38.56 / 41.77 ±5.38 / 52.50 ms │     no change │
│ QQuery 54 │        59.97 / 62.77 ±2.98 / 68.52 ms │              63.74 / 67.36 ±5.13 / 77.47 ms │  1.07x slower │
│ QQuery 55 │        31.38 / 34.13 ±3.00 / 39.22 ms │              30.90 / 33.57 ±3.72 / 40.97 ms │     no change │
│ QQuery 56 │        46.21 / 53.02 ±4.12 / 58.09 ms │              47.95 / 52.62 ±4.86 / 61.17 ms │     no change │
│ QQuery 57 │     182.31 / 187.63 ±6.50 / 199.29 ms │           185.17 / 189.90 ±5.26 / 200.12 ms │     no change │
│ QQuery 58 │     118.66 / 124.10 ±6.28 / 135.64 ms │           121.28 / 125.73 ±4.97 / 135.31 ms │     no change │
│ QQuery 59 │    128.94 / 150.43 ±13.07 / 169.67 ms │           142.99 / 154.41 ±9.52 / 165.94 ms │     no change │
│ QQuery 60 │        47.54 / 52.02 ±3.41 / 57.26 ms │              47.74 / 49.30 ±0.86 / 50.32 ms │ +1.06x faster │
│ QQuery 61 │          9.02 / 9.89 ±1.31 / 12.50 ms │                9.01 / 9.44 ±0.48 / 10.35 ms │     no change │
│ QQuery 62 │        44.07 / 44.85 ±0.60 / 45.73 ms │              44.52 / 45.33 ±0.67 / 46.29 ms │     no change │
│ QQuery 63 │        37.78 / 41.16 ±4.41 / 49.78 ms │              37.89 / 38.93 ±0.76 / 39.73 ms │ +1.06x faster │
│ QQuery 64 │     435.21 / 446.07 ±7.24 / 454.28 ms │          440.88 / 458.68 ±12.56 / 477.36 ms │     no change │
│ QQuery 65 │       90.08 / 99.57 ±7.87 / 111.79 ms │             91.01 / 98.21 ±6.22 / 107.05 ms │     no change │
│ QQuery 66 │        80.35 / 83.01 ±3.22 / 89.17 ms │             83.09 / 87.52 ±7.22 / 101.93 ms │  1.05x slower │
│ QQuery 67 │     209.82 / 220.10 ±7.89 / 234.17 ms │           223.17 / 227.09 ±2.93 / 230.44 ms │     no change │
│ QQuery 68 │           6.68 / 7.23 ±0.58 / 7.99 ms │                 6.46 / 6.90 ±0.45 / 7.77 ms │     no change │
│ QQuery 69 │        61.99 / 66.33 ±4.94 / 75.33 ms │              64.42 / 69.44 ±7.94 / 85.27 ms │     no change │
│ QQuery 70 │     114.30 / 119.05 ±4.86 / 128.32 ms │           114.67 / 127.18 ±6.72 / 133.82 ms │  1.07x slower │
│ QQuery 71 │        41.65 / 46.55 ±5.18 / 53.65 ms │              44.13 / 49.46 ±3.68 / 55.47 ms │  1.06x slower │
│ QQuery 72 │ 2153.11 / 2211.46 ±58.34 / 2312.25 ms │       2123.89 / 2243.43 ±71.67 / 2307.51 ms │     no change │
│ QQuery 73 │           5.54 / 5.91 ±0.65 / 7.19 ms │                 5.35 / 5.68 ±0.47 / 6.62 ms │     no change │
│ QQuery 74 │     261.47 / 272.77 ±6.63 / 279.12 ms │          205.01 / 221.18 ±11.11 / 238.55 ms │ +1.23x faster │
│ QQuery 75 │    175.54 / 186.50 ±13.56 / 211.53 ms │           174.49 / 181.99 ±7.72 / 192.67 ms │     no change │
│ QQuery 76 │        42.63 / 47.90 ±3.66 / 52.25 ms │              47.96 / 49.72 ±1.00 / 51.03 ms │     no change │
│ QQuery 77 │        70.74 / 75.58 ±5.37 / 84.49 ms │              71.26 / 73.38 ±1.93 / 76.57 ms │     no change │
│ QQuery 78 │    371.70 / 394.56 ±19.30 / 419.11 ms │          370.03 / 391.06 ±18.89 / 422.73 ms │     no change │
│ QQuery 79 │        78.72 / 81.27 ±2.48 / 85.58 ms │             78.84 / 84.10 ±8.13 / 100.27 ms │     no change │
│ QQuery 80 │     127.61 / 134.17 ±4.39 / 139.44 ms │           125.14 / 132.11 ±4.61 / 137.58 ms │     no change │
│ QQuery 81 │        34.24 / 36.51 ±2.14 / 40.34 ms │              30.47 / 32.45 ±2.50 / 36.91 ms │ +1.13x faster │
│ QQuery 82 │        24.37 / 24.95 ±0.49 / 25.87 ms │              21.73 / 24.86 ±2.15 / 28.38 ms │     no change │
│ QQuery 83 │        35.05 / 36.17 ±1.33 / 38.63 ms │              35.64 / 36.36 ±0.64 / 37.36 ms │     no change │
│ QQuery 84 │        36.13 / 37.47 ±1.64 / 40.65 ms │              37.39 / 39.72 ±2.19 / 42.53 ms │  1.06x slower │
│ QQuery 85 │     116.92 / 119.76 ±4.66 / 129.05 ms │           118.75 / 126.12 ±6.53 / 137.91 ms │  1.05x slower │
│ QQuery 86 │        29.97 / 30.75 ±0.67 / 31.64 ms │              29.83 / 30.61 ±0.65 / 31.59 ms │     no change │
│ QQuery 87 │        75.12 / 79.00 ±3.05 / 83.30 ms │              71.81 / 78.79 ±5.66 / 85.04 ms │     no change │
│ QQuery 88 │        87.56 / 89.76 ±2.36 / 94.18 ms │              89.39 / 90.49 ±0.71 / 91.24 ms │     no change │
│ QQuery 89 │        44.34 / 45.81 ±1.08 / 47.61 ms │              44.21 / 47.15 ±4.15 / 55.39 ms │     no change │
│ QQuery 90 │        18.68 / 19.05 ±0.32 / 19.63 ms │              18.33 / 18.60 ±0.31 / 19.06 ms │     no change │
│ QQuery 91 │        46.46 / 51.39 ±5.15 / 58.44 ms │              47.34 / 48.27 ±0.85 / 49.79 ms │ +1.06x faster │
│ QQuery 92 │        37.77 / 38.75 ±0.61 / 39.49 ms │              37.49 / 39.13 ±1.04 / 40.38 ms │     no change │
│ QQuery 93 │        69.90 / 76.02 ±6.13 / 84.50 ms │              70.73 / 78.62 ±6.41 / 85.29 ms │     no change │
│ QQuery 94 │        47.19 / 48.02 ±0.77 / 49.00 ms │              47.66 / 48.02 ±0.30 / 48.45 ms │     no change │
│ QQuery 95 │     101.42 / 103.78 ±1.96 / 106.97 ms │           101.75 / 104.36 ±2.75 / 107.95 ms │     no change │
│ QQuery 96 │        28.05 / 32.61 ±5.20 / 39.92 ms │              27.80 / 32.08 ±4.41 / 37.77 ms │     no change │
│ QQuery 97 │        67.95 / 70.17 ±1.59 / 72.42 ms │              65.77 / 73.41 ±4.54 / 79.30 ms │     no change │
│ QQuery 98 │        46.93 / 55.87 ±6.97 / 63.19 ms │              47.18 / 50.26 ±4.38 / 58.82 ms │ +1.11x faster │
│ QQuery 99 │        64.88 / 69.87 ±5.83 / 80.71 ms │              65.44 / 67.09 ±1.55 / 69.83 ms │     no change │
└───────────┴───────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 11616.33ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 11548.09ms │
│ Average Time (HEAD)                                        │   117.34ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   116.65ms │
│ Queries Faster                                             │         10 │
│ Queries Slower                                             │         19 │
│ Queries with No Change                                     │         70 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

tpcds — tpcds_sf1

Query Base Changed Change
Query 1 135.9 KiB 254.5 KiB +87.3%
Query 2 47.9 MiB 60.5 MiB +26.2%
Query 3 2.5 MiB 2.5 MiB +0.0%
Query 4 85.6 MiB 113.4 MiB +32.5%
Query 5 14.9 MiB 14.0 MiB -5.4%
Query 6 4.3 MiB 7.6 MiB +77.9%
Query 7 10.0 MiB 8.7 MiB -13.4%
Query 8 2.0 MiB 3.2 MiB +58.6%
Query 9 32.1 KiB 32.1 KiB +0.0%
Query 10 2.0 MiB 3.1 MiB +53.3%
Query 11 58.1 MiB 78.1 MiB +34.5%
Query 12 114.4 MiB 124.1 MiB +8.5%
Query 13 15.7 MiB 19.6 MiB +24.8%
Query 14 170.9 MiB 170.9 MiB +0.0%
Query 15 4.3 MiB 7.1 MiB +64.9%
Query 16 118.8 KiB 161.5 KiB +36.0%
Query 17 39.9 MiB 39.7 MiB -0.6%
Query 18 21.2 MiB 22.4 MiB +5.3%
Query 19 2.3 MiB 3.1 MiB +34.7%
Query 20 114.3 MiB 114.6 MiB +0.3%
Query 21 3.6 MiB 3.1 MiB -15.9%
Query 22 61.8 MiB 61.8 MiB +0.0%
Query 23 211.7 MiB 215.5 MiB +1.8%
Query 24 36.4 MiB 45.7 MiB +25.3%
Query 25 86.0 MiB 85.5 MiB -0.6%
Query 26 9.9 MiB 9.9 MiB +0.0%
Query 27 929.6 KiB 929.6 KiB +0.0%
Query 28 1.3 MiB 1.6 MiB +20.0%
Query 29 97.2 MiB 97.5 MiB +0.3%
Query 30 3.8 MiB 7.3 MiB +92.0%
Query 31 13.4 MiB 18.0 MiB +34.3%
Query 32 2.9 MiB 2.9 MiB +0.0%
Query 33 4.4 MiB 4.3 MiB -2.6%
Query 34 12.7 KiB 12.7 KiB +0.0%
Query 35 19.3 MiB 29.0 MiB +50.6%
Query 36 10.3 MiB 10.3 MiB +0.0%
Query 37 3.3 MiB 3.3 MiB +0.0%
Query 38 10.4 MiB 20.2 MiB +93.3%
Query 39 52.9 MiB 43.7 MiB -17.3%
Query 40 12.0 MiB 12.0 MiB +0.0%
Query 41 2.1 MiB 2.1 MiB +0.0%
Query 42 1.6 MiB 1.6 MiB -0.0%
Query 43 275.4 KiB 275.4 KiB +0.0%
Query 44 97.6 KiB 192.4 KiB +97.2%
Query 45 4.7 MiB 8.2 MiB +76.7%
Query 46 5.4 MiB 9.4 MiB +73.5%
Query 47 128.5 MiB 128.5 MiB -0.0%
Query 48 12.9 MiB 16.0 MiB +24.1%
Query 49 34.0 MiB 51.1 MiB +50.4%
Query 50 16.7 MiB 16.5 MiB -1.0%
Query 51 177.7 MiB 177.7 MiB +0.0%
Query 52 2.3 MiB 2.5 MiB +8.0%
Query 53 30.6 MiB 20.6 MiB -32.7%
Query 54 2.0 MiB 4.0 MiB +98.7%
Query 55 2.1 MiB 2.1 MiB +1.3%
Query 56 3.0 MiB 3.1 MiB +2.6%
Query 57 120.8 MiB 120.8 MiB +0.0%
Query 58 4.6 MiB 4.6 MiB +0.0%
Query 59 12.6 MiB 12.1 MiB -3.8%
Query 60 5.7 MiB 5.5 MiB -2.9%
Query 61 2.4 KiB 2.5 KiB +3.3%
Query 62 4.3 MiB 4.3 MiB +0.0%
Query 63 30.6 MiB 30.4 MiB -0.5%
Query 64 123.4 MiB 129.2 MiB +4.7%
Query 65 29.5 MiB 29.5 MiB +0.0%
Query 66 20.9 MiB 18.3 MiB -12.3%
Query 67 117.0 MiB 116.2 MiB -0.7%
Query 68 218.6 KiB 96.3 KiB -56.0%
Query 69 3.9 MiB 47.3 MiB +1106.0%
Query 70 32.5 MiB 32.5 MiB +0.0%
Query 71 50.4 MiB 50.6 MiB +0.4%
Query 72 110.2 MiB 108.3 MiB -1.7%
Query 73 12.7 KiB 12.7 KiB +0.0%
Query 74 35.2 MiB 50.0 MiB +41.9%
Query 75 21.5 MiB 21.7 MiB +1.0%
Query 76 5.6 MiB 5.7 MiB +1.9%
Query 77 4.5 MiB 4.9 MiB +7.6%
Query 78 316.9 MiB 316.9 MiB +0.0%
Query 79 5.8 MiB 9.8 MiB +68.8%
Query 80 41.2 MiB 43.6 MiB +6.0%
Query 81 4.0 MiB 11.4 MiB +187.9%
Query 82 4.9 MiB 4.9 MiB +0.0%
Query 83 8.4 MiB 8.4 MiB +0.0%
Query 84 11.6 MiB 12.0 MiB +3.1%
Query 85 17.2 MiB 21.3 MiB +24.0%
Query 86 82.1 MiB 81.9 MiB -0.2%
Query 87 12.9 MiB 21.0 MiB +62.7%
Query 88 2.0 MiB 1.8 MiB -12.0%
Query 89 112.2 MiB 112.0 MiB -0.2%
Query 90 348.6 KiB 338.7 KiB -2.8%
Query 91 21.1 MiB 22.1 MiB +4.4%
Query 92 2.9 MiB 2.9 MiB +0.0%
Query 93 14.9 MiB 15.6 MiB +5.1%
Query 94 3.7 MiB 3.7 MiB +0.0%
Query 95 19.4 MiB 19.4 MiB +0.0%
Query 96 224.7 KiB 224.7 KiB +0.0%
Query 97 33.2 MiB 33.1 MiB -0.2%
Query 98 114.7 MiB 124.4 MiB +8.4%
Query 99 4.1 MiB 4.1 MiB +0.0%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpcds base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 316.9 MiB 2.0 GiB 1.7 GiB 6.6×
tpcds changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 316.9 MiB 2.1 GiB 1.8 GiB 6.8×
Resource Usage

tpcds — base (merge-base)

Metric Value
Wall time 60.0s
Peak memory 2.0 GiB
Avg memory 1.4 GiB
CPU user 249.5s
CPU sys 6.2s
Peak spill 0 B

tpcds — branch

Metric Value
Wall time 60.0s
Peak memory 2.1 GiB
Avg memory 1.5 GiB
CPU user 249.5s
CPU sys 6.2s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991649714-5461-ksft5 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991649714-5463-r44tq 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

Results will be posted here when complete


File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark running (GKE) | trigger
Instance: c4-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c6991649714-5462-sf4hc 6.12.105+ #1 SMP Sat Sep 19 07:57:03 UTC 2026 x86_64 GNU/Linux

CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"

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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │    1.14 ms │                                     1.18 ms │     no change │
│ QQuery 1  │   11.48 ms │                                    11.90 ms │     no change │
│ QQuery 2  │   35.17 ms │                                    35.10 ms │     no change │
│ QQuery 3  │   31.12 ms │                                    31.47 ms │     no change │
│ QQuery 4  │  247.95 ms │                                   247.31 ms │     no change │
│ QQuery 5  │  757.01 ms │                                   278.61 ms │ +2.72x faster │
│ QQuery 6  │    1.18 ms │                                     1.21 ms │     no change │
│ QQuery 7  │   13.23 ms │                                    13.61 ms │     no change │
│ QQuery 8  │  346.84 ms │                                   343.46 ms │     no change │
│ QQuery 9  │  485.80 ms │                                   472.06 ms │     no change │
│ QQuery 10 │  203.81 ms │                                    65.02 ms │ +3.13x faster │
│ QQuery 11 │  214.39 ms │                                    76.62 ms │ +2.80x faster │
│ QQuery 12 │  769.71 ms │                                   272.39 ms │ +2.83x faster │
│ QQuery 13 │ 1084.31 ms │                                   368.21 ms │ +2.94x faster │
│ QQuery 14 │  876.36 ms │                                   288.19 ms │ +3.04x faster │
│ QQuery 15 │  299.30 ms │                                   293.57 ms │     no change │
│ QQuery 16 │ 1329.73 ms │                                   631.78 ms │ +2.10x faster │
│ QQuery 17 │ 1291.47 ms │                                   639.58 ms │ +2.02x faster │
│ QQuery 18 │ 2102.81 ms │                                  1246.83 ms │ +1.69x faster │
│ QQuery 19 │   27.66 ms │                                    27.43 ms │     no change │
│ QQuery 20 │ 1069.73 ms │                                   518.96 ms │ +2.06x faster │
│ QQuery 21 │ 1383.37 ms │                                   512.58 ms │ +2.70x faster │
│ QQuery 22 │ 3247.69 ms │                                   987.87 ms │ +3.29x faster │
│ QQuery 23 │ 9315.98 ms │                                  3131.03 ms │ +2.98x faster │
│ QQuery 24 │   97.08 ms │                                    41.22 ms │ +2.35x faster │
│ QQuery 25 │  356.68 ms │                                   106.91 ms │ +3.34x faster │
│ QQuery 26 │   98.94 ms │                                    41.60 ms │ +2.38x faster │
│ QQuery 27 │ 1165.21 ms │                                   522.07 ms │ +2.23x faster │
│ QQuery 28 │ 2518.63 ms │                                  2963.67 ms │  1.18x slower │
│ QQuery 29 │   38.13 ms │                                    39.20 ms │     no change │
│ QQuery 30 │  558.86 ms │                                   316.92 ms │ +1.76x faster │
│ QQuery 31 │  551.56 ms │                                   289.57 ms │ +1.90x faster │
│ QQuery 32 │ 1095.53 ms │                                   978.92 ms │ +1.12x faster │
│ QQuery 33 │ 3379.15 ms │                                  1537.53 ms │ +2.20x faster │
│ QQuery 34 │ 3430.06 ms │                                  1511.36 ms │ +2.27x faster │
│ QQuery 35 │  312.02 ms │                                   300.90 ms │     no change │
│ QQuery 36 │  138.42 ms │                                    64.89 ms │ +2.13x faster │
│ QQuery 37 │   97.26 ms │                                    34.18 ms │ +2.85x faster │
│ QQuery 38 │   81.29 ms │                                    38.89 ms │ +2.09x faster │
│ QQuery 39 │  320.16 ms │                                   139.08 ms │ +2.30x faster │
│ QQuery 40 │   13.35 ms │                                    13.63 ms │     no change │
│ QQuery 41 │   13.03 ms │                                    13.04 ms │     no change │
│ QQuery 42 │   12.95 ms │                                    12.25 ms │ +1.06x faster │
└───────────┴────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 39425.58ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 19461.80ms │
│ Average Time (HEAD)                                        │   916.87ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   452.60ms │
│ Queries Faster                                             │         27 │
│ Queries Slower                                             │          1 │
│ Queries with No Change                                     │         15 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                   HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │            1.14 / 1.75 ±1.10 / 3.95 ms │                 1.18 / 1.80 ±1.09 / 3.97 ms │     no change │
│ QQuery 1  │         11.48 / 11.77 ±0.15 / 11.93 ms │              11.90 / 12.24 ±0.19 / 12.43 ms │     no change │
│ QQuery 2  │         35.17 / 35.74 ±0.63 / 36.90 ms │              35.10 / 35.39 ±0.27 / 35.83 ms │     no change │
│ QQuery 3  │         31.12 / 31.81 ±0.76 / 33.28 ms │              31.47 / 31.90 ±0.25 / 32.22 ms │     no change │
│ QQuery 4  │      247.95 / 254.08 ±5.60 / 261.62 ms │           247.31 / 252.35 ±4.80 / 260.77 ms │     no change │
│ QQuery 5  │     757.01 / 769.04 ±10.75 / 788.42 ms │           278.61 / 283.59 ±4.06 / 288.17 ms │ +2.71x faster │
│ QQuery 6  │            1.18 / 1.34 ±0.23 / 1.79 ms │                 1.21 / 1.37 ±0.26 / 1.88 ms │     no change │
│ QQuery 7  │         13.23 / 15.63 ±4.51 / 24.64 ms │              13.61 / 14.70 ±1.86 / 18.41 ms │ +1.06x faster │
│ QQuery 8  │      346.84 / 351.13 ±5.10 / 360.90 ms │           343.46 / 348.82 ±4.30 / 356.42 ms │     no change │
│ QQuery 9  │      485.80 / 497.43 ±7.43 / 506.59 ms │          472.06 / 486.45 ±11.07 / 500.27 ms │     no change │
│ QQuery 10 │      203.81 / 205.81 ±2.80 / 211.35 ms │              65.02 / 66.78 ±1.65 / 69.29 ms │ +3.08x faster │
│ QQuery 11 │      214.39 / 215.12 ±0.38 / 215.50 ms │              76.62 / 77.47 ±0.58 / 78.43 ms │ +2.78x faster │
│ QQuery 12 │      769.71 / 778.17 ±6.24 / 787.72 ms │           272.39 / 279.72 ±8.01 / 294.11 ms │ +2.78x faster │
│ QQuery 13 │  1084.31 / 1101.48 ±15.08 / 1122.92 ms │          368.21 / 380.09 ±10.03 / 397.40 ms │ +2.90x faster │
│ QQuery 14 │     876.36 / 896.03 ±13.61 / 907.43 ms │           288.19 / 293.66 ±6.40 / 305.72 ms │ +3.05x faster │
│ QQuery 15 │      299.30 / 310.48 ±6.02 / 315.63 ms │           293.57 / 300.89 ±7.77 / 313.52 ms │     no change │
│ QQuery 16 │  1329.73 / 1348.99 ±12.38 / 1362.54 ms │           631.78 / 644.71 ±9.86 / 661.73 ms │ +2.09x faster │
│ QQuery 17 │  1291.47 / 1316.10 ±19.19 / 1348.03 ms │           639.58 / 647.93 ±9.77 / 662.67 ms │ +2.03x faster │
│ QQuery 18 │  2102.81 / 2125.09 ±22.16 / 2156.17 ms │       1246.83 / 1268.54 ±17.27 / 1289.33 ms │ +1.68x faster │
│ QQuery 19 │         27.66 / 28.71 ±1.41 / 31.48 ms │              27.43 / 29.39 ±2.83 / 34.95 ms │     no change │
│ QQuery 20 │  1069.73 / 1084.90 ±10.73 / 1103.10 ms │           518.96 / 530.52 ±8.16 / 543.67 ms │ +2.04x faster │
│ QQuery 21 │   1383.37 / 1392.32 ±5.22 / 1397.01 ms │          512.58 / 529.99 ±12.98 / 547.62 ms │ +2.63x faster │
│ QQuery 22 │  3247.69 / 3325.07 ±67.54 / 3440.69 ms │        987.87 / 1007.55 ±26.24 / 1058.69 ms │ +3.30x faster │
│ QQuery 23 │ 9315.98 / 9588.41 ±190.89 / 9834.17 ms │       3131.03 / 3199.05 ±43.59 / 3247.75 ms │ +3.00x faster │
│ QQuery 24 │      97.08 / 105.34 ±11.85 / 128.72 ms │              41.22 / 44.65 ±5.49 / 55.54 ms │ +2.36x faster │
│ QQuery 25 │      356.68 / 360.71 ±4.11 / 368.38 ms │           106.91 / 109.88 ±2.56 / 112.95 ms │ +3.28x faster │
│ QQuery 26 │       98.94 / 101.40 ±2.36 / 105.57 ms │              41.60 / 42.63 ±0.81 / 43.97 ms │ +2.38x faster │
│ QQuery 27 │   1165.21 / 1171.88 ±5.31 / 1180.23 ms │           522.07 / 532.65 ±9.33 / 548.34 ms │ +2.20x faster │
│ QQuery 28 │  2518.63 / 2560.47 ±32.71 / 2596.60 ms │       2963.67 / 3000.66 ±21.13 / 3025.85 ms │  1.17x slower │
│ QQuery 29 │         38.13 / 42.49 ±8.09 / 58.66 ms │             39.20 / 50.80 ±16.90 / 83.10 ms │  1.20x slower │
│ QQuery 30 │      558.86 / 564.53 ±5.37 / 574.62 ms │           316.92 / 324.79 ±5.57 / 333.53 ms │ +1.74x faster │
│ QQuery 31 │      551.56 / 561.56 ±9.34 / 578.51 ms │           289.57 / 297.68 ±6.30 / 308.66 ms │ +1.89x faster │
│ QQuery 32 │  1095.53 / 1141.08 ±70.61 / 1280.94 ms │        978.92 / 1019.62 ±40.40 / 1097.20 ms │ +1.12x faster │
│ QQuery 33 │  3379.15 / 3509.82 ±71.90 / 3592.30 ms │       1537.53 / 1560.12 ±22.36 / 1602.66 ms │ +2.25x faster │
│ QQuery 34 │  3430.06 / 3469.01 ±49.50 / 3563.68 ms │       1511.36 / 1526.40 ±12.87 / 1550.00 ms │ +2.27x faster │
│ QQuery 35 │     312.02 / 335.41 ±28.45 / 389.13 ms │          300.90 / 329.38 ±28.05 / 365.28 ms │     no change │
│ QQuery 36 │     138.42 / 146.16 ±10.20 / 166.10 ms │              64.89 / 71.75 ±8.24 / 87.86 ms │ +2.04x faster │
│ QQuery 37 │       97.26 / 102.37 ±6.37 / 114.35 ms │              34.18 / 40.53 ±3.42 / 44.22 ms │ +2.53x faster │
│ QQuery 38 │         81.29 / 85.10 ±3.47 / 91.03 ms │              38.89 / 44.96 ±4.10 / 48.73 ms │ +1.89x faster │
│ QQuery 39 │      320.16 / 325.95 ±4.16 / 332.78 ms │           139.08 / 145.30 ±3.81 / 150.12 ms │ +2.24x faster │
│ QQuery 40 │         13.35 / 14.11 ±0.53 / 14.94 ms │              13.63 / 17.12 ±3.90 / 22.45 ms │  1.21x slower │
│ QQuery 41 │         13.03 / 15.08 ±3.46 / 21.97 ms │              13.04 / 15.18 ±3.76 / 22.67 ms │     no change │
│ QQuery 42 │         12.95 / 15.11 ±3.56 / 22.16 ms │              12.25 / 13.81 ±2.60 / 18.99 ms │ +1.09x faster │
└───────────┴────────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 40313.98ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 19912.82ms │
│ Average Time (HEAD)                                        │   937.53ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   463.09ms │
│ Queries Faster                                             │         28 │
│ Queries Slower                                             │          3 │
│ Queries with No Change                                     │         12 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

clickbench_partitioned

Query Base Changed Change
Query 0 0 B 0 B 0.0%
Query 1 104 B 104 B +0.0%
Query 2 936 B 936 B +0.0%
Query 3 312 B 312 B +0.0%
Query 4 832.1 MiB 844.7 MiB +1.5%
Query 5 1.4 GiB 1.2 GiB -8.1%
Query 6 0 B 0 B 0.0%
Query 7 60.1 MiB 60.3 MiB +0.3%
Query 8 946.3 MiB 943.6 MiB -0.3%
Query 9 552.2 MiB 594.0 MiB +7.6%
Query 10 72.9 MiB 114.7 MiB +57.4%
Query 11 75.2 MiB 124.2 MiB +65.0%
Query 12 1.5 GiB 1.4 GiB -8.1%
Query 13 1.5 GiB 1.6 GiB +1.2%
Query 14 1.6 GiB 1.3 GiB -16.9%
Query 15 1.2 GiB 1.2 GiB +0.1%
Query 16 3.3 GiB 3.3 GiB +0.5%
Query 17 3.2 GiB 3.5 GiB +7.2%
Query 18 7.4 GiB 8.6 GiB +16.6%
Query 19 0 B 0 B 0.0%
Query 20 104 B 104 B +0.0%
Query 21 1.3 MiB 3.7 MiB +177.8%
Query 22 2.5 MiB 2.9 MiB +14.4%
Query 23 10.8 MiB 28.3 MiB +161.8%
Query 24 6.9 MiB 61.3 MiB +792.1%
Query 25 6.9 MiB 167.3 MiB +2315.8%
Query 26 7.5 MiB 62.3 MiB +731.4%
Query 27 2.5 MiB 2.5 MiB +0.0%
Query 28 1.8 GiB 1.8 GiB -2.0%
Query 29 624 B 624 B +0.0%
Query 30 668.8 MiB 708.9 MiB +6.0%
Query 31 1.5 GiB 1.5 GiB +4.8%
Query 32 7.2 GiB 7.1 GiB -0.3%
Query 33 9.9 GiB 7.9 GiB -20.2%
Query 34 9.9 GiB 7.9 GiB -20.2%
Query 35 665.0 MiB 669.0 MiB +0.6%
Query 36 100.1 MiB 123.9 MiB +23.7%
Query 37 8.6 MiB 7.9 MiB -8.7%
Query 38 5.4 MiB 5.0 MiB -7.1%
Query 39 213.4 MiB 308.8 MiB +44.7%
Query 40 2.3 MiB 2.1 MiB -8.7%
Query 41 3.3 MiB 3.3 MiB +0.0%
Query 42 1.6 MiB 1.8 MiB +12.0%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
clickbench_partitioned base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 9.9 GiB 13.0 GiB 3.2 GiB 1.3×
clickbench_partitioned changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 8.6 GiB 12.0 GiB 3.4 GiB 1.4×
Resource Usage

clickbench_partitioned — base (merge-base)

Metric Value
Wall time 205.0s
Peak memory 13.0 GiB
Avg memory 5.5 GiB
CPU user 1938.3s
CPU sys 133.8s
Peak spill 0 B

clickbench_partitioned — branch

Metric Value
Wall time 100.0s
Peak memory 12.0 GiB
Avg memory 4.9 GiB
CPU user 1012.8s
CPU sys 75.9s
Peak spill 0 B

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 f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "arm64"
CPU Details (lscpu)
Architecture:                            aarch64
CPU op-mode(s):                          64-bit
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               ARM
Model name:                              Neoverse-V2
Model:                                   1
Thread(s) per core:                      1
Core(s) per cluster:                     16
Socket(s):                               -
Cluster(s):                              1
Stepping:                                r0p1
BogoMIPS:                                2000.00
Flags:                                   fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm sb paca pacg dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh rng bti
L1d cache:                               1 MiB (16 instances)
L1i cache:                               1 MiB (16 instances)
L2 cache:                                32 MiB (16 instances)
L3 cache:                                80 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; CSV2, BHB
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │    1.16 ms │                                     1.21 ms │     no change │
│ QQuery 1  │   11.97 ms │                                    12.12 ms │     no change │
│ QQuery 2  │   35.21 ms │                                    35.32 ms │     no change │
│ QQuery 3  │   31.82 ms │                                    32.43 ms │     no change │
│ QQuery 4  │  258.09 ms │                                   253.36 ms │     no change │
│ QQuery 5  │  788.24 ms │                                   287.81 ms │ +2.74x faster │
│ QQuery 6  │    1.22 ms │                                     1.28 ms │  1.05x slower │
│ QQuery 7  │   13.40 ms │                                    13.85 ms │     no change │
│ QQuery 8  │  359.86 ms │                                   355.39 ms │     no change │
│ QQuery 9  │  510.74 ms │                                   510.57 ms │     no change │
│ QQuery 10 │  204.71 ms │                                    67.15 ms │ +3.05x faster │
│ QQuery 11 │  217.13 ms │                                    78.23 ms │ +2.78x faster │
│ QQuery 12 │  784.87 ms │                                   282.67 ms │ +2.78x faster │
│ QQuery 13 │ 1111.16 ms │                                   398.56 ms │ +2.79x faster │
│ QQuery 14 │  876.78 ms │                                   302.37 ms │ +2.90x faster │
│ QQuery 15 │  318.58 ms │                                   309.56 ms │     no change │
│ QQuery 16 │ 1392.76 ms │                                   647.47 ms │ +2.15x faster │
│ QQuery 17 │ 1363.30 ms │                                   658.08 ms │ +2.07x faster │
│ QQuery 18 │ 2170.12 ms │                                  1294.05 ms │ +1.68x faster │
│ QQuery 19 │   28.32 ms │                                    28.69 ms │     no change │
│ QQuery 20 │ 1091.83 ms │                                   530.27 ms │ +2.06x faster │
│ QQuery 21 │ 1412.11 ms │                                   523.50 ms │ +2.70x faster │
│ QQuery 22 │ 3338.97 ms │                                  1014.50 ms │ +3.29x faster │
│ QQuery 23 │ 9427.38 ms │                                  3234.29 ms │ +2.91x faster │
│ QQuery 24 │   98.52 ms │                                    41.55 ms │ +2.37x faster │
│ QQuery 25 │  360.47 ms │                                   108.23 ms │ +3.33x faster │
│ QQuery 26 │  101.02 ms │                                    42.44 ms │ +2.38x faster │
│ QQuery 27 │ 1194.34 ms │                                   527.56 ms │ +2.26x faster │
│ QQuery 28 │ 2573.63 ms │                                  3017.99 ms │  1.17x slower │
│ QQuery 29 │   39.08 ms │                                    39.93 ms │     no change │
│ QQuery 30 │  581.24 ms │                                   328.08 ms │ +1.77x faster │
│ QQuery 31 │  552.30 ms │                                   292.49 ms │ +1.89x faster │
│ QQuery 32 │ 1149.53 ms │                                  1014.15 ms │ +1.13x faster │
│ QQuery 33 │ 3563.93 ms │                                  1559.38 ms │ +2.29x faster │
│ QQuery 34 │ 3696.39 ms │                                  1563.57 ms │ +2.36x faster │
│ QQuery 35 │  356.14 ms │                                   313.52 ms │ +1.14x faster │
│ QQuery 36 │  138.34 ms │                                    66.22 ms │ +2.09x faster │
│ QQuery 37 │  100.57 ms │                                    34.91 ms │ +2.88x faster │
│ QQuery 38 │   78.93 ms │                                    41.34 ms │ +1.91x faster │
│ QQuery 39 │  310.64 ms │                                   150.65 ms │ +2.06x faster │
│ QQuery 40 │   14.21 ms │                                    14.02 ms │     no change │
│ QQuery 41 │   13.07 ms │                                    13.08 ms │     no change │
│ QQuery 42 │   12.93 ms │                                    12.64 ms │     no change │
└───────────┴────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 40685.02ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 20054.50ms │
│ Average Time (HEAD)                                        │   946.16ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   466.38ms │
│ Queries Faster                                             │         27 │
│ Queries Slower                                             │          2 │
│ Queries with No Change                                     │         14 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                   HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │            1.16 / 1.80 ±1.16 / 4.11 ms │                 1.21 / 1.84 ±1.08 / 4.00 ms │     no change │
│ QQuery 1  │         11.97 / 12.29 ±0.35 / 12.87 ms │              12.12 / 12.33 ±0.12 / 12.49 ms │     no change │
│ QQuery 2  │         35.21 / 35.61 ±0.27 / 35.88 ms │              35.32 / 35.56 ±0.26 / 36.07 ms │     no change │
│ QQuery 3  │         31.82 / 32.68 ±0.68 / 33.62 ms │              32.43 / 32.83 ±0.65 / 34.11 ms │     no change │
│ QQuery 4  │      258.09 / 261.59 ±1.85 / 263.54 ms │           253.36 / 264.17 ±5.90 / 270.03 ms │     no change │
│ QQuery 5  │      788.24 / 802.44 ±9.01 / 816.31 ms │           287.81 / 293.33 ±2.83 / 295.72 ms │ +2.74x faster │
│ QQuery 6  │            1.22 / 1.37 ±0.21 / 1.78 ms │                 1.28 / 1.42 ±0.23 / 1.88 ms │     no change │
│ QQuery 7  │         13.40 / 13.54 ±0.10 / 13.67 ms │              13.85 / 14.26 ±0.22 / 14.44 ms │  1.05x slower │
│ QQuery 8  │      359.86 / 365.56 ±5.14 / 375.07 ms │           355.39 / 359.78 ±4.33 / 367.97 ms │     no change │
│ QQuery 9  │     510.74 / 526.01 ±10.98 / 538.84 ms │          510.57 / 522.53 ±12.01 / 543.61 ms │     no change │
│ QQuery 10 │      204.71 / 207.50 ±4.46 / 216.34 ms │              67.15 / 70.43 ±5.21 / 80.73 ms │ +2.95x faster │
│ QQuery 11 │      217.13 / 219.37 ±2.41 / 223.98 ms │              78.23 / 79.70 ±0.90 / 80.92 ms │ +2.75x faster │
│ QQuery 12 │     784.87 / 807.83 ±12.89 / 822.66 ms │           282.67 / 291.44 ±7.23 / 302.02 ms │ +2.77x faster │
│ QQuery 13 │  1111.16 / 1154.93 ±41.43 / 1227.31 ms │          398.56 / 408.90 ±11.03 / 429.72 ms │ +2.82x faster │
│ QQuery 14 │     876.78 / 907.79 ±20.88 / 934.06 ms │           302.37 / 306.29 ±2.92 / 309.85 ms │ +2.96x faster │
│ QQuery 15 │      318.58 / 323.83 ±4.82 / 332.90 ms │          309.56 / 335.67 ±20.45 / 370.81 ms │     no change │
│ QQuery 16 │   1392.76 / 1406.56 ±8.58 / 1419.87 ms │           647.47 / 659.84 ±8.49 / 672.07 ms │ +2.13x faster │
│ QQuery 17 │   1363.30 / 1371.07 ±7.47 / 1381.48 ms │          658.08 / 678.87 ±19.43 / 715.56 ms │ +2.02x faster │
│ QQuery 18 │  2170.12 / 2199.25 ±26.37 / 2241.12 ms │       1294.05 / 1325.10 ±26.09 / 1373.20 ms │ +1.66x faster │
│ QQuery 19 │         28.32 / 29.11 ±0.83 / 30.46 ms │              28.69 / 34.93 ±6.87 / 44.86 ms │  1.20x slower │
│ QQuery 20 │  1091.83 / 1103.75 ±15.73 / 1133.80 ms │           530.27 / 545.26 ±9.18 / 554.52 ms │ +2.02x faster │
│ QQuery 21 │  1412.11 / 1425.68 ±11.27 / 1441.00 ms │           523.50 / 530.53 ±5.11 / 539.34 ms │ +2.69x faster │
│ QQuery 22 │  3338.97 / 3375.37 ±20.86 / 3394.03 ms │       1014.50 / 1032.10 ±13.00 / 1048.97 ms │ +3.27x faster │
│ QQuery 23 │ 9427.38 / 9624.63 ±127.25 / 9806.81 ms │       3234.29 / 3281.72 ±40.51 / 3329.06 ms │ +2.93x faster │
│ QQuery 24 │      98.52 / 104.88 ±10.74 / 126.27 ms │              41.55 / 43.09 ±1.80 / 46.53 ms │ +2.43x faster │
│ QQuery 25 │     360.47 / 367.92 ±13.01 / 393.87 ms │           108.23 / 117.00 ±9.86 / 134.94 ms │ +3.14x faster │
│ QQuery 26 │      101.02 / 101.56 ±0.52 / 102.36 ms │              42.44 / 42.87 ±0.41 / 43.55 ms │ +2.37x faster │
│ QQuery 27 │  1194.34 / 1208.30 ±12.49 / 1230.13 ms │           527.56 / 533.68 ±4.27 / 539.13 ms │ +2.26x faster │
│ QQuery 28 │  2573.63 / 2630.97 ±54.33 / 2703.19 ms │       3017.99 / 3042.39 ±26.28 / 3076.39 ms │  1.16x slower │
│ QQuery 29 │        39.08 / 45.95 ±13.01 / 71.98 ms │            39.93 / 57.62 ±25.44 / 107.65 ms │  1.25x slower │
│ QQuery 30 │      581.24 / 589.44 ±6.62 / 596.48 ms │           328.08 / 339.62 ±9.62 / 355.34 ms │ +1.74x faster │
│ QQuery 31 │     552.30 / 571.96 ±15.64 / 595.32 ms │           292.49 / 308.18 ±8.48 / 318.01 ms │ +1.86x faster │
│ QQuery 32 │ 1149.53 / 1314.85 ±306.29 / 1927.17 ms │       1014.15 / 1068.15 ±91.48 / 1250.05 ms │ +1.23x faster │
│ QQuery 33 │  3563.93 / 3676.74 ±63.67 / 3743.52 ms │       1559.38 / 1580.19 ±15.70 / 1604.69 ms │ +2.33x faster │
│ QQuery 34 │  3696.39 / 3779.74 ±81.12 / 3931.51 ms │       1563.57 / 1616.86 ±56.78 / 1722.58 ms │ +2.34x faster │
│ QQuery 35 │     356.14 / 386.66 ±19.31 / 406.42 ms │          313.52 / 329.51 ±18.03 / 363.05 ms │ +1.17x faster │
│ QQuery 36 │      138.34 / 142.07 ±2.60 / 144.80 ms │              66.22 / 69.86 ±2.75 / 74.65 ms │ +2.03x faster │
│ QQuery 37 │      100.57 / 107.84 ±6.44 / 117.07 ms │             34.91 / 43.50 ±12.12 / 67.13 ms │ +2.48x faster │
│ QQuery 38 │         78.93 / 83.83 ±2.89 / 86.73 ms │              41.34 / 43.68 ±2.60 / 48.34 ms │ +1.92x faster │
│ QQuery 39 │     310.64 / 339.41 ±25.33 / 372.56 ms │           150.65 / 155.78 ±4.35 / 160.79 ms │ +2.18x faster │
│ QQuery 40 │         14.21 / 15.03 ±0.49 / 15.70 ms │              14.02 / 14.39 ±0.66 / 15.70 ms │     no change │
│ QQuery 41 │         13.07 / 13.48 ±0.32 / 13.82 ms │              13.08 / 13.30 ±0.13 / 13.44 ms │     no change │
│ QQuery 42 │         12.93 / 13.13 ±0.18 / 13.44 ms │              12.64 / 13.99 ±2.42 / 18.82 ms │  1.07x slower │
└───────────┴────────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 41703.33ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 20552.51ms │
│ Average Time (HEAD)                                        │   969.84ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   477.97ms │
│ Queries Faster                                             │         27 │
│ Queries Slower                                             │          5 │
│ Queries with No Change                                     │         11 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

clickbench_partitioned

Query Base Changed Change
Query 0 0 B 0 B 0.0%
Query 1 104 B 104 B +0.0%
Query 2 936 B 936 B +0.0%
Query 3 312 B 312 B +0.0%
Query 4 843.5 MiB 842.3 MiB -0.1%
Query 5 1.4 GiB 1.2 GiB -10.5%
Query 6 0 B 0 B 0.0%
Query 7 70.0 MiB 50.1 MiB -28.4%
Query 8 972.0 MiB 944.9 MiB -2.8%
Query 9 552.2 MiB 593.9 MiB +7.5%
Query 10 72.2 MiB 118.9 MiB +64.7%
Query 11 77.1 MiB 117.9 MiB +52.9%
Query 12 1.5 GiB 1.3 GiB -10.2%
Query 13 1.5 GiB 1.6 GiB +2.1%
Query 14 1.6 GiB 1.2 GiB -22.6%
Query 15 1.2 GiB 1.3 GiB +4.0%
Query 16 3.3 GiB 3.4 GiB +5.1%
Query 17 3.2 GiB 3.4 GiB +5.9%
Query 18 7.1 GiB 8.0 GiB +12.4%
Query 19 0 B 0 B 0.0%
Query 20 104 B 104 B +0.0%
Query 21 1.4 MiB 3.7 MiB +170.7%
Query 22 2.5 MiB 4.2 MiB +64.9%
Query 23 12.6 MiB 27.9 MiB +122.1%
Query 24 6.9 MiB 61.8 MiB +797.6%
Query 25 8.1 MiB 172.8 MiB +2022.3%
Query 26 8.2 MiB 59.9 MiB +626.9%
Query 27 2.7 MiB 2.2 MiB -16.3%
Query 28 1.8 GiB 1.7 GiB -3.2%
Query 29 624 B 624 B +0.0%
Query 30 651.1 MiB 696.2 MiB +6.9%
Query 31 1.5 GiB 1.4 GiB -4.6%
Query 32 7.2 GiB 7.1 GiB -0.2%
Query 33 9.6 GiB 7.7 GiB -20.3%
Query 34 9.9 GiB 8.1 GiB -17.6%
Query 35 688.3 MiB 665.5 MiB -3.3%
Query 36 100.1 MiB 118.9 MiB +18.7%
Query 37 8.4 MiB 7.3 MiB -13.5%
Query 38 5.3 MiB 6.0 MiB +11.9%
Query 39 213.4 MiB 308.8 MiB +44.7%
Query 40 2.6 MiB 2.3 MiB -10.7%
Query 41 3.1 MiB 3.3 MiB +9.1%
Query 42 1.7 MiB 1.6 MiB -8.4%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
clickbench_partitioned base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 9.9 GiB 13.3 GiB 3.4 GiB 1.3×
clickbench_partitioned changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 8.1 GiB 10.9 GiB 2.7 GiB 1.3×
Resource Usage

clickbench_partitioned — base (merge-base)

Metric Value
Wall time 210.1s
Peak memory 13.3 GiB
Avg memory 5.5 GiB
CPU user 1991.9s
CPU sys 140.0s
Peak spill 0 B

clickbench_partitioned — branch

Metric Value
Wall time 105.0s
Peak memory 10.9 GiB
Avg memory 4.5 GiB
CPU user 1042.5s
CPU sys 80.5s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpch
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃      HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │  75.95 ms │                                    79.52 ms │     no change │
│ QQuery 2  │  23.77 ms │                                    24.84 ms │     no change │
│ QQuery 3  │  46.77 ms │                                    44.38 ms │ +1.05x faster │
│ QQuery 4  │  26.96 ms │                                    27.38 ms │     no change │
│ QQuery 5  │  54.87 ms │                                    55.05 ms │     no change │
│ QQuery 6  │  24.80 ms │                                    24.47 ms │     no change │
│ QQuery 7  │  66.68 ms │                                    66.82 ms │     no change │
│ QQuery 8  │  65.18 ms │                                    65.73 ms │     no change │
│ QQuery 9  │  87.09 ms │                                    85.13 ms │     no change │
│ QQuery 10 │  86.85 ms │                                    64.73 ms │ +1.34x faster │
│ QQuery 11 │  18.33 ms │                                    17.16 ms │ +1.07x faster │
│ QQuery 12 │  33.06 ms │                                    30.60 ms │ +1.08x faster │
│ QQuery 13 │  65.73 ms │                                    53.92 ms │ +1.22x faster │
│ QQuery 14 │  37.78 ms │                                    38.00 ms │     no change │
│ QQuery 15 │  52.40 ms │                                    53.30 ms │     no change │
│ QQuery 16 │  17.04 ms │                                    17.81 ms │     no change │
│ QQuery 17 │ 104.14 ms │                                   105.25 ms │     no change │
│ QQuery 18 │  90.59 ms │                                    69.88 ms │ +1.30x faster │
│ QQuery 19 │  48.17 ms │                                    54.24 ms │  1.13x slower │
│ QQuery 20 │  52.17 ms │                                    48.92 ms │ +1.07x faster │
│ QQuery 21 │  91.62 ms │                                    90.58 ms │     no change │
│ QQuery 22 │  20.40 ms │                                    17.41 ms │ +1.17x faster │
└───────────┴───────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 1190.35ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 1135.13ms │
│ Average Time (HEAD)                                        │   54.11ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   51.60ms │
│ Queries Faster                                             │         8 │
│ Queries Slower                                             │         1 │
│ Queries with No Change                                     │        13 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                              HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │    75.95 / 81.08 ±3.09 / 84.31 ms │              79.52 / 85.42 ±4.76 / 93.57 ms │  1.05x slower │
│ QQuery 2  │    23.77 / 25.44 ±1.39 / 27.88 ms │              24.84 / 26.21 ±1.59 / 29.24 ms │     no change │
│ QQuery 3  │    46.77 / 47.87 ±0.96 / 49.29 ms │              44.38 / 46.88 ±1.68 / 49.65 ms │     no change │
│ QQuery 4  │    26.96 / 27.61 ±0.38 / 28.05 ms │              27.38 / 28.07 ±0.54 / 28.63 ms │     no change │
│ QQuery 5  │    54.87 / 56.35 ±1.12 / 57.51 ms │              55.05 / 57.02 ±2.70 / 62.28 ms │     no change │
│ QQuery 6  │    24.80 / 25.31 ±0.34 / 25.85 ms │              24.47 / 25.80 ±1.06 / 27.37 ms │     no change │
│ QQuery 7  │    66.68 / 70.99 ±2.93 / 75.19 ms │              66.82 / 71.93 ±3.34 / 76.71 ms │     no change │
│ QQuery 8  │    65.18 / 68.08 ±1.86 / 70.30 ms │              65.73 / 68.74 ±2.79 / 73.56 ms │     no change │
│ QQuery 9  │    87.09 / 88.49 ±1.27 / 90.77 ms │              85.13 / 87.97 ±1.67 / 90.06 ms │     no change │
│ QQuery 10 │    86.85 / 90.86 ±3.11 / 94.43 ms │              64.73 / 67.68 ±3.32 / 74.11 ms │ +1.34x faster │
│ QQuery 11 │    18.33 / 20.86 ±2.04 / 23.78 ms │              17.16 / 18.99 ±1.40 / 21.38 ms │ +1.10x faster │
│ QQuery 12 │    33.06 / 34.83 ±1.35 / 36.68 ms │              30.60 / 32.99 ±2.21 / 37.09 ms │ +1.06x faster │
│ QQuery 13 │    65.73 / 68.10 ±2.92 / 73.85 ms │              53.92 / 57.87 ±3.17 / 60.67 ms │ +1.18x faster │
│ QQuery 14 │    37.78 / 39.91 ±1.32 / 41.63 ms │              38.00 / 40.46 ±1.97 / 43.41 ms │     no change │
│ QQuery 15 │    52.40 / 54.47 ±1.20 / 56.02 ms │              53.30 / 54.16 ±0.66 / 54.92 ms │     no change │
│ QQuery 16 │    17.04 / 19.10 ±2.73 / 24.44 ms │              17.81 / 18.76 ±0.85 / 20.09 ms │     no change │
│ QQuery 17 │ 104.14 / 113.03 ±6.08 / 122.96 ms │           105.25 / 112.53 ±6.36 / 123.75 ms │     no change │
│ QQuery 18 │   90.59 / 95.95 ±5.65 / 106.39 ms │              69.88 / 71.46 ±1.21 / 72.90 ms │ +1.34x faster │
│ QQuery 19 │    48.17 / 50.74 ±2.25 / 54.65 ms │              54.24 / 57.15 ±2.18 / 59.47 ms │  1.13x slower │
│ QQuery 20 │    52.17 / 56.12 ±2.88 / 60.35 ms │              48.92 / 50.97 ±1.08 / 52.03 ms │ +1.10x faster │
│ QQuery 21 │    91.62 / 93.96 ±2.03 / 96.93 ms │              90.58 / 91.35 ±0.78 / 92.43 ms │     no change │
│ QQuery 22 │    20.40 / 22.63 ±1.87 / 25.99 ms │              17.41 / 19.46 ±1.13 / 20.82 ms │ +1.16x faster │
└───────────┴───────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃           ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 1251.81ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 1191.87ms │
│ Average Time (HEAD)                                        │   56.90ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   54.18ms │
│ Queries Faster                                             │         7 │
│ Queries Slower                                             │         2 │
│ Queries with No Change                                     │        13 │
│ Queries with Failure                                       │         0 │
└────────────────────────────────────────────────────────────┴───────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

tpch — tpch_sf1

Query Base Changed Change
Query 1 30.2 MiB 40.0 MiB +32.5%
Query 2 23.1 MiB 24.7 MiB +7.0%
Query 3 11.1 MiB 11.1 MiB +0.0%
Query 4 22.3 MiB 23.2 MiB +4.3%
Query 5 42.7 MiB 42.9 MiB +0.3%
Query 6 832 B 832 B +0.0%
Query 7 103.6 MiB 106.5 MiB +2.7%
Query 8 33.1 MiB 33.1 MiB +0.0%
Query 9 147.7 MiB 147.6 MiB -0.1%
Query 10 46.3 MiB 41.0 MiB -11.4%
Query 11 41.2 MiB 31.5 MiB -23.4%
Query 12 22.4 MiB 22.7 MiB +1.1%
Query 13 52.5 MiB 72.6 MiB +38.3%
Query 14 6.7 MiB 6.7 MiB +0.0%
Query 15 13.4 MiB 12.7 MiB -5.1%
Query 16 33.3 MiB 44.8 MiB +34.6%
Query 17 154.1 MiB 155.1 MiB +0.7%
Query 18 118.0 MiB 167.1 MiB +41.6%
Query 19 209.7 KiB 725.7 KiB +246.1%
Query 20 56.0 MiB 58.7 MiB +4.7%
Query 21 36.1 MiB 36.4 MiB +0.9%
Query 22 22.1 MiB 31.8 MiB +43.6%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpch base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 154.1 MiB 1.2 GiB 1.0 GiB 7.7×
tpch changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 167.1 MiB 965.7 MiB 798.6 MiB 5.8×
Resource Usage

tpch — base (merge-base)

Metric Value
Wall time 10.0s
Peak memory 1.2 GiB
Avg memory 496.7 MiB
CPU user 40.4s
CPU sys 2.2s
Peak spill 0 B

tpch — branch

Metric Value
Wall time 10.0s
Peak memory 965.7 MiB
Avg memory 383.7 MiB
CPU user 39.4s
CPU sys 2.2s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark tpcds
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃       HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │    5.24 ms │                                     5.07 ms │     no change │
│ QQuery 2  │   81.73 ms │                                    85.46 ms │     no change │
│ QQuery 3  │   33.60 ms │                                    33.17 ms │     no change │
│ QQuery 4  │  559.04 ms │                                   463.10 ms │ +1.21x faster │
│ QQuery 5  │   59.08 ms │                                    55.05 ms │ +1.07x faster │
│ QQuery 6  │   36.59 ms │                                    36.64 ms │     no change │
│ QQuery 7  │   83.53 ms │                                    83.32 ms │     no change │
│ QQuery 8  │   42.65 ms │                                    39.97 ms │ +1.07x faster │
│ QQuery 9  │   74.70 ms │                                    71.82 ms │     no change │
│ QQuery 10 │   64.54 ms │                                    75.36 ms │  1.17x slower │
│ QQuery 11 │  347.47 ms │                                   265.91 ms │ +1.31x faster │
│ QQuery 12 │   30.61 ms │                                    29.30 ms │     no change │
│ QQuery 13 │  127.63 ms │                                   129.85 ms │     no change │
│ QQuery 14 │  477.55 ms │                                   492.93 ms │     no change │
│ QQuery 15 │   48.57 ms │                                    52.92 ms │  1.09x slower │
│ QQuery 16 │    5.99 ms │                                     5.91 ms │     no change │
│ QQuery 17 │   95.27 ms │                                    96.47 ms │     no change │
│ QQuery 18 │  114.57 ms │                                   115.95 ms │     no change │
│ QQuery 19 │   46.62 ms │                                    45.81 ms │     no change │
│ QQuery 20 │   34.65 ms │                                    33.66 ms │     no change │
│ QQuery 21 │   20.67 ms │                                    21.17 ms │     no change │
│ QQuery 22 │   91.02 ms │                                    91.29 ms │     no change │
│ QQuery 23 │  373.43 ms │                                   370.22 ms │     no change │
│ QQuery 24 │  246.57 ms │                                   244.82 ms │     no change │
│ QQuery 25 │  134.42 ms │                                   130.54 ms │     no change │
│ QQuery 26 │   50.77 ms │                                    52.32 ms │     no change │
│ QQuery 27 │    5.33 ms │                                     5.52 ms │     no change │
│ QQuery 28 │   87.91 ms │                                    85.52 ms │     no change │
│ QQuery 29 │  116.61 ms │                                   120.32 ms │     no change │
│ QQuery 30 │   41.02 ms │                                    35.15 ms │ +1.17x faster │
│ QQuery 31 │  126.48 ms │                                   124.72 ms │     no change │
│ QQuery 32 │   21.80 ms │                                    22.28 ms │     no change │
│ QQuery 33 │   38.46 ms │                                    40.52 ms │  1.05x slower │
│ QQuery 34 │    5.45 ms │                                     5.52 ms │     no change │
│ QQuery 35 │   72.13 ms │                                    74.53 ms │     no change │
│ QQuery 36 │    5.65 ms │                                     5.76 ms │     no change │
│ QQuery 37 │    7.54 ms │                                     7.56 ms │     no change │
│ QQuery 38 │   68.06 ms │                                    64.92 ms │     no change │
│ QQuery 39 │   81.56 ms │                                    83.06 ms │     no change │
│ QQuery 40 │   27.42 ms │                                    26.53 ms │     no change │
│ QQuery 41 │   14.49 ms │                                    14.63 ms │     no change │
│ QQuery 42 │   28.39 ms │                                    27.95 ms │     no change │
│ QQuery 43 │    4.58 ms │                                     4.64 ms │     no change │
│ QQuery 44 │    5.53 ms │                                     5.60 ms │     no change │
│ QQuery 45 │   35.67 ms │                                    36.35 ms │     no change │
│ QQuery 46 │   11.85 ms │                                    12.94 ms │  1.09x slower │
│ QQuery 47 │  237.85 ms │                                   236.98 ms │     no change │
│ QQuery 48 │  105.72 ms │                                   106.79 ms │     no change │
│ QQuery 49 │   86.51 ms │                                    80.72 ms │ +1.07x faster │
│ QQuery 50 │   72.06 ms │                                    70.59 ms │     no change │
│ QQuery 51 │  117.18 ms │                                   112.80 ms │     no change │
│ QQuery 52 │   28.11 ms │                                    27.97 ms │     no change │
│ QQuery 53 │   35.67 ms │                                    35.94 ms │     no change │
│ QQuery 54 │   58.64 ms │                                    57.45 ms │     no change │
│ QQuery 55 │   27.96 ms │                                    26.29 ms │ +1.06x faster │
│ QQuery 56 │   42.49 ms │                                    42.54 ms │     no change │
│ QQuery 57 │  174.49 ms │                                   170.67 ms │     no change │
│ QQuery 58 │  111.64 ms │                                   112.46 ms │     no change │
│ QQuery 59 │  122.54 ms │                                   128.97 ms │  1.05x slower │
│ QQuery 60 │   43.30 ms │                                    43.28 ms │     no change │
│ QQuery 61 │    8.26 ms │                                     8.55 ms │     no change │
│ QQuery 62 │   41.45 ms │                                    40.14 ms │     no change │
│ QQuery 63 │   34.64 ms │                                    35.05 ms │     no change │
│ QQuery 64 │  419.92 ms │                                   424.08 ms │     no change │
│ QQuery 65 │   85.17 ms │                                    84.42 ms │     no change │
│ QQuery 66 │   74.99 ms │                                    74.24 ms │     no change │
│ QQuery 67 │  203.35 ms │                                   204.79 ms │     no change │
│ QQuery 68 │    6.61 ms │                                     6.24 ms │ +1.06x faster │
│ QQuery 69 │   58.46 ms │                                    60.49 ms │     no change │
│ QQuery 70 │  109.33 ms │                                   115.78 ms │  1.06x slower │
│ QQuery 71 │   40.45 ms │                                    39.62 ms │     no change │
│ QQuery 72 │ 2101.66 ms │                                  2058.89 ms │     no change │
│ QQuery 73 │    5.13 ms │                                     5.16 ms │     no change │
│ QQuery 74 │  245.27 ms │                                   198.44 ms │ +1.24x faster │
│ QQuery 75 │  167.34 ms │                                   166.35 ms │     no change │
│ QQuery 76 │   40.85 ms │                                    40.65 ms │     no change │
│ QQuery 77 │   65.50 ms │                                    71.09 ms │  1.09x slower │
│ QQuery 78 │  344.81 ms │                                   357.29 ms │     no change │
│ QQuery 79 │   73.50 ms │                                    77.42 ms │  1.05x slower │
│ QQuery 80 │  119.99 ms │                                   117.55 ms │     no change │
│ QQuery 81 │   30.28 ms │                                    28.20 ms │ +1.07x faster │
│ QQuery 82 │   20.40 ms │                                    22.91 ms │  1.12x slower │
│ QQuery 83 │   33.70 ms │                                    32.63 ms │     no change │
│ QQuery 84 │   33.66 ms │                                    31.46 ms │ +1.07x faster │
│ QQuery 85 │  108.20 ms │                                   114.89 ms │  1.06x slower │
│ QQuery 86 │   27.30 ms │                                    26.19 ms │     no change │
│ QQuery 87 │   70.84 ms │                                    70.83 ms │     no change │
│ QQuery 88 │   82.62 ms │                                    82.27 ms │     no change │
│ QQuery 89 │   40.04 ms │                                    41.24 ms │     no change │
│ QQuery 90 │   17.34 ms │                                    18.14 ms │     no change │
│ QQuery 91 │   43.18 ms │                                    45.33 ms │     no change │
│ QQuery 92 │   35.25 ms │                                    36.41 ms │     no change │
│ QQuery 93 │   61.51 ms │                                    63.04 ms │     no change │
│ QQuery 94 │   43.87 ms │                                    46.11 ms │  1.05x slower │
│ QQuery 95 │   95.78 ms │                                    96.92 ms │     no change │
│ QQuery 96 │   26.38 ms │                                    26.37 ms │     no change │
│ QQuery 97 │   63.20 ms │                                    62.73 ms │     no change │
│ QQuery 98 │   44.71 ms │                                    44.16 ms │     no change │
│ QQuery 99 │   61.10 ms │                                    63.01 ms │     no change │
└───────────┴────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 10542.65ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 10324.54ms │
│ Average Time (HEAD)                                        │   106.49ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   104.29ms │
│ Queries Faster                                             │         11 │
│ Queries Slower                                             │         11 │
│ Queries with No Change                                     │         77 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                  HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1  │          5.24 / 7.04 ±2.39 / 11.73 ms │                5.07 / 6.44 ±2.10 / 10.60 ms │ +1.09x faster │
│ QQuery 2  │        81.73 / 86.38 ±5.67 / 93.40 ms │             85.46 / 90.54 ±5.98 / 102.02 ms │     no change │
│ QQuery 3  │        33.60 / 37.35 ±3.99 / 42.97 ms │              33.17 / 36.30 ±5.35 / 46.90 ms │     no change │
│ QQuery 4  │    559.04 / 570.74 ±10.33 / 587.40 ms │          463.10 / 482.02 ±14.02 / 500.01 ms │ +1.18x faster │
│ QQuery 5  │        59.08 / 63.82 ±3.72 / 69.77 ms │              55.05 / 58.18 ±1.94 / 60.64 ms │ +1.10x faster │
│ QQuery 6  │        36.59 / 41.10 ±4.51 / 48.67 ms │              36.64 / 40.25 ±5.35 / 50.91 ms │     no change │
│ QQuery 7  │        83.53 / 85.15 ±1.39 / 86.86 ms │             83.32 / 88.06 ±6.39 / 100.62 ms │     no change │
│ QQuery 8  │        42.65 / 42.92 ±0.17 / 43.14 ms │              39.97 / 43.55 ±5.14 / 53.71 ms │     no change │
│ QQuery 9  │        74.70 / 81.58 ±4.12 / 85.92 ms │              71.82 / 80.04 ±5.35 / 86.80 ms │     no change │
│ QQuery 10 │        64.54 / 67.29 ±2.93 / 70.95 ms │              75.36 / 77.10 ±1.04 / 78.19 ms │  1.15x slower │
│ QQuery 11 │    347.47 / 377.10 ±17.55 / 399.80 ms │          265.91 / 280.54 ±12.08 / 297.36 ms │ +1.34x faster │
│ QQuery 12 │        30.61 / 31.20 ±0.54 / 32.11 ms │              29.30 / 30.38 ±1.09 / 32.43 ms │     no change │
│ QQuery 13 │     127.63 / 132.51 ±5.15 / 142.06 ms │           129.85 / 139.47 ±8.02 / 150.68 ms │  1.05x slower │
│ QQuery 14 │    477.55 / 502.41 ±18.71 / 530.19 ms │           492.93 / 505.26 ±8.57 / 518.46 ms │     no change │
│ QQuery 15 │        48.57 / 49.75 ±0.84 / 50.74 ms │              52.92 / 53.57 ±0.82 / 55.10 ms │  1.08x slower │
│ QQuery 16 │           5.99 / 6.27 ±0.21 / 6.56 ms │                5.91 / 7.00 ±1.59 / 10.16 ms │  1.12x slower │
│ QQuery 17 │       95.27 / 98.01 ±1.94 / 100.36 ms │            96.47 / 104.22 ±4.81 / 109.81 ms │  1.06x slower │
│ QQuery 18 │     114.57 / 116.30 ±1.68 / 118.86 ms │           115.95 / 122.00 ±5.68 / 129.45 ms │     no change │
│ QQuery 19 │        46.62 / 48.22 ±1.16 / 49.59 ms │              45.81 / 52.91 ±6.60 / 61.50 ms │  1.10x slower │
│ QQuery 20 │        34.65 / 35.77 ±0.67 / 36.59 ms │              33.66 / 35.58 ±2.39 / 40.29 ms │     no change │
│ QQuery 21 │        20.67 / 21.96 ±0.85 / 23.17 ms │              21.17 / 22.07 ±0.70 / 23.08 ms │     no change │
│ QQuery 22 │        91.02 / 93.50 ±1.62 / 95.49 ms │              91.29 / 92.43 ±0.67 / 93.14 ms │     no change │
│ QQuery 23 │    373.43 / 399.89 ±20.91 / 430.70 ms │          370.22 / 385.00 ±11.07 / 403.86 ms │     no change │
│ QQuery 24 │    246.57 / 258.18 ±14.21 / 285.23 ms │          244.82 / 255.53 ±10.15 / 274.55 ms │     no change │
│ QQuery 25 │     134.42 / 138.07 ±2.92 / 141.66 ms │           130.54 / 133.15 ±3.44 / 139.79 ms │     no change │
│ QQuery 26 │        50.77 / 53.31 ±2.12 / 57.13 ms │              52.32 / 54.83 ±3.59 / 61.89 ms │     no change │
│ QQuery 27 │           5.33 / 6.07 ±0.78 / 7.03 ms │                 5.52 / 6.39 ±0.67 / 7.20 ms │  1.05x slower │
│ QQuery 28 │       87.91 / 94.35 ±6.97 / 106.62 ms │             85.52 / 92.93 ±7.04 / 104.81 ms │     no change │
│ QQuery 29 │     116.61 / 123.52 ±6.44 / 135.67 ms │           120.32 / 127.17 ±5.44 / 134.76 ms │     no change │
│ QQuery 30 │        41.02 / 43.06 ±2.30 / 46.87 ms │              35.15 / 36.78 ±1.22 / 38.53 ms │ +1.17x faster │
│ QQuery 31 │     126.48 / 134.78 ±7.08 / 147.72 ms │           124.72 / 131.82 ±8.12 / 146.76 ms │     no change │
│ QQuery 32 │        21.80 / 22.73 ±0.71 / 23.76 ms │              22.28 / 23.10 ±0.64 / 24.24 ms │     no change │
│ QQuery 33 │        38.46 / 42.54 ±3.75 / 49.63 ms │              40.52 / 43.16 ±3.70 / 50.50 ms │     no change │
│ QQuery 34 │           5.45 / 5.62 ±0.24 / 6.08 ms │                 5.52 / 5.69 ±0.24 / 6.13 ms │     no change │
│ QQuery 35 │        72.13 / 75.97 ±4.79 / 85.42 ms │              74.53 / 76.53 ±2.20 / 80.58 ms │     no change │
│ QQuery 36 │           5.65 / 5.92 ±0.23 / 6.24 ms │                 5.76 / 6.44 ±0.87 / 8.06 ms │  1.09x slower │
│ QQuery 37 │           7.54 / 8.07 ±0.48 / 8.88 ms │                 7.56 / 8.84 ±0.93 / 9.91 ms │  1.10x slower │
│ QQuery 38 │        68.06 / 71.15 ±2.96 / 76.72 ms │              64.92 / 71.70 ±6.37 / 80.27 ms │     no change │
│ QQuery 39 │        81.56 / 87.33 ±3.00 / 89.73 ms │              83.06 / 86.59 ±2.94 / 90.44 ms │     no change │
│ QQuery 40 │        27.42 / 29.03 ±1.10 / 30.46 ms │              26.53 / 28.03 ±0.80 / 28.92 ms │     no change │
│ QQuery 41 │        14.49 / 14.75 ±0.20 / 15.01 ms │              14.63 / 16.18 ±2.02 / 19.92 ms │  1.10x slower │
│ QQuery 42 │        28.39 / 30.66 ±2.73 / 35.85 ms │              27.95 / 31.37 ±2.60 / 35.09 ms │     no change │
│ QQuery 43 │           4.58 / 4.78 ±0.21 / 5.16 ms │                 4.64 / 4.76 ±0.12 / 4.99 ms │     no change │
│ QQuery 44 │           5.53 / 5.65 ±0.11 / 5.83 ms │                 5.60 / 5.69 ±0.11 / 5.91 ms │     no change │
│ QQuery 45 │        35.67 / 36.61 ±0.65 / 37.18 ms │              36.35 / 38.77 ±1.48 / 40.65 ms │  1.06x slower │
│ QQuery 46 │        11.85 / 12.33 ±0.33 / 12.79 ms │              12.94 / 13.68 ±0.42 / 14.27 ms │  1.11x slower │
│ QQuery 47 │     237.85 / 245.17 ±7.41 / 259.24 ms │          236.98 / 252.86 ±12.23 / 265.93 ms │     no change │
│ QQuery 48 │     105.72 / 108.39 ±3.52 / 115.07 ms │           106.79 / 111.35 ±7.36 / 126.01 ms │     no change │
│ QQuery 49 │        86.51 / 91.52 ±3.76 / 97.34 ms │              80.72 / 85.44 ±4.50 / 92.76 ms │ +1.07x faster │
│ QQuery 50 │        72.06 / 75.09 ±4.81 / 84.68 ms │              70.59 / 77.69 ±7.85 / 89.76 ms │     no change │
│ QQuery 51 │     117.18 / 118.78 ±2.56 / 123.90 ms │           112.80 / 118.41 ±5.76 / 128.41 ms │     no change │
│ QQuery 52 │        28.11 / 30.91 ±4.20 / 39.21 ms │              27.97 / 28.87 ±0.74 / 29.67 ms │ +1.07x faster │
│ QQuery 53 │        35.67 / 37.74 ±3.76 / 45.25 ms │              35.94 / 39.44 ±4.44 / 47.71 ms │     no change │
│ QQuery 54 │        58.64 / 63.52 ±5.20 / 70.81 ms │              57.45 / 58.68 ±0.86 / 59.94 ms │ +1.08x faster │
│ QQuery 55 │        27.96 / 31.92 ±4.72 / 38.72 ms │              26.29 / 27.37 ±0.67 / 28.06 ms │ +1.17x faster │
│ QQuery 56 │        42.49 / 51.29 ±4.90 / 56.66 ms │              42.54 / 45.53 ±3.09 / 50.35 ms │ +1.13x faster │
│ QQuery 57 │     174.49 / 180.01 ±6.10 / 189.25 ms │           170.67 / 177.40 ±6.47 / 186.82 ms │     no change │
│ QQuery 58 │     111.64 / 113.87 ±1.84 / 116.60 ms │           112.46 / 121.77 ±6.33 / 131.93 ms │  1.07x slower │
│ QQuery 59 │     122.54 / 132.98 ±8.83 / 144.35 ms │          128.97 / 143.36 ±10.91 / 154.71 ms │  1.08x slower │
│ QQuery 60 │        43.30 / 49.48 ±6.22 / 57.46 ms │              43.28 / 46.01 ±3.93 / 53.80 ms │ +1.08x faster │
│ QQuery 61 │           8.26 / 8.50 ±0.16 / 8.74 ms │                 8.55 / 8.75 ±0.20 / 9.10 ms │     no change │
│ QQuery 62 │        41.45 / 42.03 ±0.36 / 42.58 ms │              40.14 / 41.10 ±0.54 / 41.82 ms │     no change │
│ QQuery 63 │        34.64 / 37.70 ±3.50 / 44.26 ms │              35.05 / 37.83 ±4.07 / 45.84 ms │     no change │
│ QQuery 64 │    419.92 / 437.07 ±17.92 / 469.15 ms │          424.08 / 435.23 ±11.58 / 454.86 ms │     no change │
│ QQuery 65 │       85.17 / 94.68 ±7.35 / 103.05 ms │             84.42 / 91.86 ±5.77 / 100.93 ms │     no change │
│ QQuery 66 │        74.99 / 83.79 ±9.06 / 95.77 ms │              74.24 / 76.56 ±1.28 / 78.05 ms │ +1.09x faster │
│ QQuery 67 │     203.35 / 210.25 ±5.55 / 219.05 ms │           204.79 / 211.10 ±5.01 / 218.76 ms │     no change │
│ QQuery 68 │           6.61 / 7.00 ±0.35 / 7.56 ms │                 6.24 / 6.70 ±0.49 / 7.59 ms │     no change │
│ QQuery 69 │        58.46 / 63.48 ±3.14 / 67.60 ms │              60.49 / 63.73 ±3.69 / 68.81 ms │     no change │
│ QQuery 70 │     109.33 / 114.14 ±4.38 / 120.43 ms │           115.78 / 119.17 ±2.07 / 121.78 ms │     no change │
│ QQuery 71 │        40.45 / 46.28 ±4.04 / 51.63 ms │              39.62 / 46.74 ±3.65 / 49.90 ms │     no change │
│ QQuery 72 │ 2101.66 / 2176.27 ±39.55 / 2210.01 ms │       2058.89 / 2147.72 ±54.33 / 2213.77 ms │     no change │
│ QQuery 73 │           5.13 / 5.46 ±0.40 / 6.25 ms │                 5.16 / 5.49 ±0.43 / 6.31 ms │     no change │
│ QQuery 74 │    245.27 / 257.45 ±10.56 / 271.79 ms │          198.44 / 211.59 ±12.82 / 228.20 ms │ +1.22x faster │
│ QQuery 75 │     167.34 / 175.55 ±6.75 / 186.37 ms │           166.35 / 175.81 ±9.64 / 191.54 ms │     no change │
│ QQuery 76 │        40.85 / 44.99 ±2.65 / 47.41 ms │              40.65 / 45.45 ±4.36 / 53.06 ms │     no change │
│ QQuery 77 │        65.50 / 66.19 ±0.65 / 67.17 ms │              71.09 / 76.76 ±4.00 / 81.08 ms │  1.16x slower │
│ QQuery 78 │    344.81 / 360.88 ±18.52 / 393.26 ms │          357.29 / 374.42 ±20.60 / 414.48 ms │     no change │
│ QQuery 79 │        73.50 / 75.57 ±1.74 / 78.57 ms │              77.42 / 82.34 ±7.35 / 96.58 ms │  1.09x slower │
│ QQuery 80 │     119.99 / 124.38 ±3.52 / 130.11 ms │           117.55 / 126.72 ±6.76 / 134.56 ms │     no change │
│ QQuery 81 │        30.28 / 31.77 ±1.24 / 33.44 ms │              28.20 / 31.49 ±3.94 / 39.05 ms │     no change │
│ QQuery 82 │        20.40 / 22.12 ±1.00 / 23.20 ms │              22.91 / 23.81 ±0.61 / 24.51 ms │  1.08x slower │
│ QQuery 83 │        33.70 / 35.00 ±1.71 / 38.25 ms │              32.63 / 33.85 ±0.73 / 34.85 ms │     no change │
│ QQuery 84 │        33.66 / 34.95 ±1.32 / 37.48 ms │              31.46 / 32.57 ±1.05 / 34.27 ms │ +1.07x faster │
│ QQuery 85 │     108.20 / 111.77 ±5.43 / 122.43 ms │           114.89 / 119.80 ±7.03 / 133.65 ms │  1.07x slower │
│ QQuery 86 │        27.30 / 28.09 ±0.46 / 28.53 ms │              26.19 / 27.86 ±1.41 / 30.09 ms │     no change │
│ QQuery 87 │        70.84 / 77.53 ±4.34 / 81.99 ms │              70.83 / 77.75 ±3.98 / 81.68 ms │     no change │
│ QQuery 88 │        82.62 / 83.51 ±0.81 / 84.84 ms │              82.27 / 84.26 ±1.56 / 86.93 ms │     no change │
│ QQuery 89 │        40.04 / 44.80 ±3.89 / 50.70 ms │              41.24 / 42.44 ±0.72 / 43.24 ms │ +1.06x faster │
│ QQuery 90 │        17.34 / 18.44 ±1.58 / 21.57 ms │              18.14 / 19.55 ±2.47 / 24.48 ms │  1.06x slower │
│ QQuery 91 │        43.18 / 46.16 ±4.62 / 55.35 ms │              45.33 / 48.28 ±4.76 / 57.72 ms │     no change │
│ QQuery 92 │        35.25 / 36.95 ±0.94 / 38.00 ms │              36.41 / 37.15 ±0.74 / 38.56 ms │     no change │
│ QQuery 93 │        61.51 / 69.69 ±6.28 / 77.58 ms │              63.04 / 70.12 ±8.47 / 86.72 ms │     no change │
│ QQuery 94 │        43.87 / 45.44 ±1.61 / 48.27 ms │              46.11 / 46.79 ±0.37 / 47.07 ms │     no change │
│ QQuery 95 │       95.78 / 97.90 ±1.45 / 100.24 ms │            96.92 / 100.35 ±2.13 / 103.37 ms │     no change │
│ QQuery 96 │        26.38 / 32.47 ±5.22 / 38.51 ms │              26.37 / 29.87 ±6.08 / 42.02 ms │ +1.09x faster │
│ QQuery 97 │        63.20 / 67.63 ±2.75 / 71.89 ms │              62.73 / 70.79 ±6.56 / 80.39 ms │     no change │
│ QQuery 98 │        44.71 / 52.35 ±5.94 / 60.17 ms │              44.16 / 49.06 ±6.17 / 60.39 ms │ +1.07x faster │
│ QQuery 99 │        61.10 / 64.86 ±5.85 / 76.51 ms │              63.01 / 65.00 ±2.33 / 68.25 ms │     no change │
└───────────┴───────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 11064.48ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 10862.10ms │
│ Average Time (HEAD)                                        │   111.76ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   109.72ms │
│ Queries Faster                                             │         17 │
│ Queries Slower                                             │         19 │
│ Queries with No Change                                     │         63 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

tpcds — tpcds_sf1

Query Base Changed Change
Query 1 137.0 KiB 135.9 KiB -0.8%
Query 2 50.5 MiB 43.1 MiB -14.5%
Query 3 2.5 MiB 2.5 MiB +1.3%
Query 4 85.5 MiB 113.4 MiB +32.6%
Query 5 14.0 MiB 14.0 MiB +0.0%
Query 6 4.3 MiB 7.6 MiB +78.3%
Query 7 7.3 MiB 7.3 MiB -0.1%
Query 8 2.0 MiB 3.2 MiB +59.0%
Query 9 32.1 KiB 32.1 KiB +0.0%
Query 10 2.0 MiB 3.1 MiB +53.3%
Query 11 58.1 MiB 79.1 MiB +36.1%
Query 12 114.4 MiB 114.4 MiB -0.0%
Query 13 15.7 MiB 18.9 MiB +20.6%
Query 14 170.9 MiB 170.9 MiB +0.0%
Query 15 4.3 MiB 7.1 MiB +64.9%
Query 16 118.8 KiB 161.5 KiB +36.0%
Query 17 39.4 MiB 39.3 MiB -0.2%
Query 18 21.2 MiB 25.6 MiB +20.6%
Query 19 2.0 MiB 1.9 MiB -7.6%
Query 20 114.6 MiB 114.6 MiB -0.0%
Query 21 2.7 MiB 2.6 MiB -4.9%
Query 22 61.8 MiB 61.8 MiB +0.0%
Query 23 211.6 MiB 215.7 MiB +1.9%
Query 24 36.4 MiB 45.7 MiB +25.3%
Query 25 85.5 MiB 85.9 MiB +0.5%
Query 26 9.9 MiB 7.4 MiB -24.8%
Query 27 929.6 KiB 929.6 KiB +0.0%
Query 28 1.5 MiB 1.3 MiB -9.1%
Query 29 97.2 MiB 97.6 MiB +0.4%
Query 30 3.7 MiB 7.3 MiB +97.9%
Query 31 13.4 MiB 18.0 MiB +34.3%
Query 32 2.9 MiB 2.9 MiB +0.0%
Query 33 4.3 MiB 4.2 MiB -1.4%
Query 34 96.3 KiB 96.3 KiB +0.0%
Query 35 20.2 MiB 26.6 MiB +31.7%
Query 36 10.3 MiB 10.3 MiB +0.0%
Query 37 3.3 MiB 3.3 MiB +0.0%
Query 38 11.3 MiB 20.2 MiB +79.1%
Query 39 32.1 MiB 32.0 MiB -0.4%
Query 40 12.0 MiB 12.0 MiB +0.0%
Query 41 2.1 MiB 2.1 MiB +0.0%
Query 42 1.6 MiB 1.5 MiB -2.1%
Query 43 275.4 KiB 292.6 KiB +6.2%
Query 44 97.6 KiB 192.4 KiB +97.2%
Query 45 4.8 MiB 8.2 MiB +72.6%
Query 46 5.4 MiB 9.4 MiB +73.5%
Query 47 132.6 MiB 138.2 MiB +4.3%
Query 48 12.5 MiB 16.4 MiB +30.9%
Query 49 34.0 MiB 34.0 MiB -0.0%
Query 50 16.3 MiB 16.7 MiB +2.0%
Query 51 178.0 MiB 177.8 MiB -0.1%
Query 52 2.5 MiB 2.5 MiB +1.4%
Query 53 30.6 MiB 30.6 MiB +0.0%
Query 54 2.0 MiB 4.0 MiB +98.7%
Query 55 2.1 MiB 2.2 MiB +3.1%
Query 56 3.1 MiB 3.1 MiB +2.7%
Query 57 120.8 MiB 120.8 MiB +0.0%
Query 58 4.6 MiB 4.6 MiB +0.0%
Query 59 12.6 MiB 13.0 MiB +3.7%
Query 60 5.5 MiB 5.9 MiB +6.9%
Query 61 2.4 KiB 2.4 KiB +2.7%
Query 62 4.3 MiB 4.3 MiB +0.0%
Query 63 20.4 MiB 20.6 MiB +0.8%
Query 64 126.2 MiB 125.3 MiB -0.7%
Query 65 29.5 MiB 29.5 MiB -0.0%
Query 66 18.0 MiB 18.3 MiB +1.4%
Query 67 117.0 MiB 116.3 MiB -0.6%
Query 68 218.6 KiB 96.3 KiB -56.0%
Query 69 4.0 MiB 47.9 MiB +1088.2%
Query 70 32.5 MiB 32.5 MiB +0.0%
Query 71 40.4 MiB 40.4 MiB +0.0%
Query 72 104.4 MiB 106.6 MiB +2.2%
Query 73 12.7 KiB 96.3 KiB +656.9%
Query 74 35.2 MiB 50.0 MiB +41.9%
Query 75 21.7 MiB 21.7 MiB +0.0%
Query 76 4.9 MiB 4.5 MiB -8.7%
Query 77 4.4 MiB 5.3 MiB +21.2%
Query 78 316.9 MiB 316.9 MiB -0.0%
Query 79 6.1 MiB 8.7 MiB +42.5%
Query 80 41.2 MiB 41.6 MiB +1.0%
Query 81 4.4 MiB 11.4 MiB +158.9%
Query 82 4.9 MiB 4.9 MiB +0.0%
Query 83 8.4 MiB 8.4 MiB +0.0%
Query 84 11.6 MiB 12.0 MiB +3.1%
Query 85 17.4 MiB 21.6 MiB +24.2%
Query 86 82.1 MiB 92.1 MiB +12.2%
Query 87 13.1 MiB 21.0 MiB +59.9%
Query 88 1.6 MiB 1.5 MiB -2.7%
Query 89 112.0 MiB 112.0 MiB -0.0%
Query 90 394.0 KiB 356.5 KiB -9.5%
Query 91 21.3 MiB 22.1 MiB +3.8%
Query 92 2.9 MiB 2.9 MiB +0.0%
Query 93 15.1 MiB 15.1 MiB +0.0%
Query 94 3.7 MiB 4.0 MiB +7.6%
Query 95 19.4 MiB 19.4 MiB +0.0%
Query 96 224.7 KiB 224.7 KiB +0.0%
Query 97 33.4 MiB 33.2 MiB -0.5%
Query 98 114.7 MiB 114.7 MiB -0.0%
Query 99 4.1 MiB 4.1 MiB +0.0%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
tpcds base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 316.9 MiB 2.0 GiB 1.7 GiB 6.5×
tpcds changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 316.9 MiB 1.9 GiB 1.6 GiB 6.3×
Resource Usage

tpcds — base (merge-base)

Metric Value
Wall time 60.0s
Peak memory 2.0 GiB
Avg memory 1.3 GiB
CPU user 236.8s
CPU sys 5.6s
Peak spill 0 B

tpcds — branch

Metric Value
Wall time 55.0s
Peak memory 1.9 GiB
Avg memory 1.4 GiB
CPU user 234.7s
CPU sys 5.9s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "20000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃        HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │     1.27 ms │                                     1.63 ms │  1.29x slower │
│ QQuery 1  │    18.79 ms │                                    20.63 ms │  1.10x slower │
│ QQuery 2  │    53.25 ms │                                    59.55 ms │  1.12x slower │
│ QQuery 3  │    57.06 ms │                                    59.20 ms │     no change │
│ QQuery 4  │   383.85 ms │                                   389.27 ms │     no change │
│ QQuery 5  │  1057.11 ms │                                   454.27 ms │ +2.33x faster │
│ QQuery 6  │     1.30 ms │                                     1.71 ms │  1.32x slower │
│ QQuery 7  │    20.79 ms │                                    22.29 ms │  1.07x slower │
│ QQuery 8  │   558.38 ms │                                   579.28 ms │     no change │
│ QQuery 9  │   688.87 ms │                                   706.50 ms │     no change │
│ QQuery 10 │   295.11 ms │                                   117.28 ms │ +2.52x faster │
│ QQuery 11 │   315.15 ms │                                   137.75 ms │ +2.29x faster │
│ QQuery 12 │  1027.61 ms │                                   433.35 ms │ +2.37x faster │
│ QQuery 13 │  1500.06 ms │                                   626.26 ms │ +2.40x faster │
│ QQuery 14 │  1216.41 ms │                                   444.82 ms │ +2.73x faster │
│ QQuery 15 │   441.98 ms │                                   461.34 ms │     no change │
│ QQuery 16 │  1891.90 ms │                                  1063.26 ms │ +1.78x faster │
│ QQuery 17 │  1888.28 ms │                                  1077.49 ms │ +1.75x faster │
│ QQuery 18 │  3045.32 ms │                                  2110.29 ms │ +1.44x faster │
│ QQuery 19 │    52.63 ms │                                    55.54 ms │  1.06x slower │
│ QQuery 20 │  1647.67 ms │                                   856.40 ms │ +1.92x faster │
│ QQuery 21 │  2155.32 ms │                                   884.29 ms │ +2.44x faster │
│ QQuery 22 │  5344.94 ms │                                  1670.98 ms │ +3.20x faster │
│ QQuery 23 │ 23506.51 ms │                                  5639.41 ms │ +4.17x faster │
│ QQuery 24 │   148.28 ms │                                    62.63 ms │ +2.37x faster │
│ QQuery 25 │   551.71 ms │                                   164.85 ms │ +3.35x faster │
│ QQuery 26 │   139.52 ms │                                    61.03 ms │ +2.29x faster │
│ QQuery 27 │  1762.30 ms │                                   859.86 ms │ +2.05x faster │
│ QQuery 28 │  3504.04 ms │                                  3740.56 ms │  1.07x slower │
│ QQuery 29 │    55.39 ms │                                    55.82 ms │     no change │
│ QQuery 30 │   876.16 ms │                                   496.98 ms │ +1.76x faster │
│ QQuery 31 │   902.84 ms │                                   519.18 ms │ +1.74x faster │
│ QQuery 32 │  1730.81 ms │                                  1709.70 ms │     no change │
│ QQuery 33 │  4486.49 ms │                                  2317.95 ms │ +1.94x faster │
│ QQuery 34 │  4459.45 ms │                                  2333.85 ms │ +1.91x faster │
│ QQuery 35 │   391.30 ms │                                   387.96 ms │     no change │
│ QQuery 36 │   151.50 ms │                                    92.99 ms │ +1.63x faster │
│ QQuery 37 │   131.16 ms │                                    43.49 ms │ +3.02x faster │
│ QQuery 38 │    98.37 ms │                                    60.88 ms │ +1.62x faster │
│ QQuery 39 │   383.27 ms │                                   191.79 ms │ +2.00x faster │
│ QQuery 40 │    20.06 ms │                                    19.18 ms │     no change │
│ QQuery 41 │    17.34 ms │                                    17.47 ms │     no change │
│ QQuery 42 │    13.59 ms │                                    13.75 ms │     no change │
└───────────┴─────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 66993.13ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 31022.72ms │
│ Average Time (HEAD)                                        │  1557.98ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   721.46ms │
│ Queries Faster                                             │         25 │
│ Queries Slower                                             │          7 │
│ Queries with No Change                                     │         11 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                      HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │               1.27 / 2.41 ±2.08 / 6.56 ms │                 1.63 / 2.67 ±2.00 / 6.68 ms │  1.11x slower │
│ QQuery 1  │            18.79 / 20.41 ±0.97 / 21.84 ms │              20.63 / 21.52 ±0.80 / 22.86 ms │  1.05x slower │
│ QQuery 2  │            53.25 / 57.01 ±2.32 / 59.57 ms │              59.55 / 60.83 ±1.51 / 62.80 ms │  1.07x slower │
│ QQuery 3  │            57.06 / 58.86 ±1.57 / 61.07 ms │              59.20 / 60.32 ±1.43 / 63.10 ms │     no change │
│ QQuery 4  │         383.85 / 395.14 ±7.87 / 404.15 ms │          389.27 / 414.85 ±14.55 / 432.65 ms │     no change │
│ QQuery 5  │      1057.11 / 1067.47 ±8.15 / 1081.37 ms │          454.27 / 475.89 ±11.06 / 485.57 ms │ +2.24x faster │
│ QQuery 6  │               1.30 / 1.58 ±0.35 / 2.23 ms │                 1.71 / 1.94 ±0.35 / 2.61 ms │  1.23x slower │
│ QQuery 7  │            20.79 / 21.09 ±0.24 / 21.49 ms │              22.29 / 25.09 ±3.58 / 32.08 ms │  1.19x slower │
│ QQuery 8  │         558.38 / 574.69 ±9.98 / 586.47 ms │          579.28 / 586.18 ±10.22 / 606.43 ms │     no change │
│ QQuery 9  │         688.87 / 698.87 ±9.24 / 715.27 ms │          706.50 / 731.61 ±15.53 / 749.83 ms │     no change │
│ QQuery 10 │         295.11 / 307.83 ±9.82 / 321.79 ms │           117.28 / 119.98 ±1.96 / 123.22 ms │ +2.57x faster │
│ QQuery 11 │         315.15 / 324.52 ±9.68 / 342.49 ms │           137.75 / 139.09 ±0.73 / 139.78 ms │ +2.33x faster │
│ QQuery 12 │     1027.61 / 1049.37 ±19.38 / 1082.15 ms │          433.35 / 443.74 ±11.81 / 465.24 ms │ +2.36x faster │
│ QQuery 13 │     1500.06 / 1539.17 ±28.84 / 1587.92 ms │          626.26 / 643.76 ±20.38 / 677.55 ms │ +2.39x faster │
│ QQuery 14 │     1216.41 / 1233.07 ±13.50 / 1250.68 ms │           444.82 / 451.50 ±5.86 / 461.69 ms │ +2.73x faster │
│ QQuery 15 │        441.98 / 461.36 ±11.88 / 473.29 ms │          461.34 / 486.08 ±18.25 / 507.27 ms │  1.05x slower │
│ QQuery 16 │     1891.90 / 1957.86 ±34.36 / 1992.45 ms │       1063.26 / 1077.80 ±12.26 / 1092.86 ms │ +1.82x faster │
│ QQuery 17 │     1888.28 / 1919.32 ±17.12 / 1934.95 ms │       1077.49 / 1098.48 ±19.43 / 1134.35 ms │ +1.75x faster │
│ QQuery 18 │     3045.32 / 3079.81 ±30.11 / 3128.16 ms │       2110.29 / 2146.93 ±23.63 / 2174.05 ms │ +1.43x faster │
│ QQuery 19 │            52.63 / 54.79 ±1.51 / 57.25 ms │             55.54 / 64.89 ±14.59 / 93.80 ms │  1.18x slower │
│ QQuery 20 │     1647.67 / 1668.10 ±16.80 / 1688.48 ms │           856.40 / 868.58 ±9.21 / 881.81 ms │ +1.92x faster │
│ QQuery 21 │      2155.32 / 2164.89 ±8.63 / 2176.04 ms │           884.29 / 893.66 ±9.11 / 910.60 ms │ +2.42x faster │
│ QQuery 22 │     5344.94 / 5421.76 ±51.83 / 5506.49 ms │       1670.98 / 1693.82 ±12.59 / 1704.94 ms │ +3.20x faster │
│ QQuery 23 │ 23506.51 / 23850.97 ±282.21 / 24368.42 ms │       5639.41 / 5696.68 ±43.57 / 5760.17 ms │ +4.19x faster │
│ QQuery 24 │        148.28 / 162.11 ±11.69 / 183.03 ms │              62.63 / 66.72 ±4.40 / 74.81 ms │ +2.43x faster │
│ QQuery 25 │         551.71 / 563.61 ±9.96 / 579.16 ms │           164.85 / 175.39 ±9.67 / 191.12 ms │ +3.21x faster │
│ QQuery 26 │         139.52 / 147.54 ±7.14 / 160.64 ms │              61.03 / 65.20 ±3.40 / 70.24 ms │ +2.26x faster │
│ QQuery 27 │     1762.30 / 1777.49 ±11.16 / 1796.36 ms │          859.86 / 874.63 ±10.81 / 889.74 ms │ +2.03x faster │
│ QQuery 28 │     3504.04 / 3525.18 ±25.84 / 3575.23 ms │       3740.56 / 3796.00 ±48.39 / 3875.68 ms │  1.08x slower │
│ QQuery 29 │            55.39 / 57.20 ±1.59 / 59.93 ms │              55.82 / 59.17 ±3.98 / 66.73 ms │     no change │
│ QQuery 30 │        876.16 / 887.44 ±11.49 / 908.31 ms │           496.98 / 500.86 ±2.63 / 504.87 ms │ +1.77x faster │
│ QQuery 31 │        902.84 / 937.83 ±18.19 / 953.17 ms │          519.18 / 541.48 ±13.19 / 554.07 ms │ +1.73x faster │
│ QQuery 32 │     1730.81 / 1767.16 ±32.74 / 1810.83 ms │       1709.70 / 1750.37 ±51.59 / 1850.24 ms │     no change │
│ QQuery 33 │     4486.49 / 4562.51 ±43.91 / 4609.83 ms │       2317.95 / 2363.72 ±26.11 / 2397.32 ms │ +1.93x faster │
│ QQuery 34 │     4459.45 / 4563.82 ±74.22 / 4638.77 ms │       2333.85 / 2380.95 ±31.99 / 2420.14 ms │ +1.92x faster │
│ QQuery 35 │         391.30 / 403.54 ±8.31 / 415.52 ms │          387.96 / 411.18 ±32.32 / 474.10 ms │     no change │
│ QQuery 36 │        151.50 / 169.08 ±14.23 / 188.80 ms │             92.99 / 97.28 ±3.56 / 102.74 ms │ +1.74x faster │
│ QQuery 37 │         131.16 / 141.04 ±9.42 / 158.45 ms │              43.49 / 50.45 ±5.66 / 60.69 ms │ +2.80x faster │
│ QQuery 38 │          98.37 / 103.81 ±3.25 / 107.12 ms │              60.88 / 67.69 ±5.14 / 73.65 ms │ +1.53x faster │
│ QQuery 39 │         383.27 / 393.39 ±9.35 / 408.58 ms │           191.79 / 205.76 ±8.10 / 217.03 ms │ +1.91x faster │
│ QQuery 40 │            20.06 / 22.89 ±1.94 / 25.51 ms │              19.18 / 19.98 ±0.66 / 20.70 ms │ +1.15x faster │
│ QQuery 41 │            17.34 / 18.39 ±0.81 / 19.28 ms │              17.47 / 19.97 ±1.32 / 21.13 ms │  1.09x slower │
│ QQuery 42 │            13.59 / 15.97 ±2.59 / 20.01 ms │              13.75 / 14.59 ±1.31 / 17.19 ms │ +1.09x faster │
└───────────┴───────────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 68150.34ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 31667.31ms │
│ Average Time (HEAD)                                        │  1584.89ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   736.45ms │
│ Queries Faster                                             │         27 │
│ Queries Slower                                             │          9 │
│ Queries with No Change                                     │          7 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

clickbench_partitioned

Query Base Changed Change
Query 0 0 B 0 B 0.0%
Query 1 104 B 104 B +0.0%
Query 2 936 B 936 B +0.0%
Query 3 312 B 312 B +0.0%
Query 4 798.8 MiB 832.3 MiB +4.2%
Query 5 1.4 GiB 1.2 GiB -11.1%
Query 6 0 B 0 B 0.0%
Query 7 40.5 MiB 40.0 MiB -1.2%
Query 8 925.7 MiB 921.6 MiB -0.4%
Query 9 636.0 MiB 678.0 MiB +6.6%
Query 10 74.4 MiB 120.3 MiB +61.6%
Query 11 76.6 MiB 112.2 MiB +46.5%
Query 12 1.5 GiB 1.4 GiB -8.6%
Query 13 1.5 GiB 1.6 GiB +0.4%
Query 14 1.6 GiB 1.3 GiB -19.9%
Query 15 1.2 GiB 1.1 GiB -3.7%
Query 16 3.1 GiB 3.4 GiB +9.0%
Query 17 3.2 GiB 3.3 GiB +5.8%
Query 18 6.9 GiB 7.6 GiB +9.6%
Query 19 0 B 0 B 0.0%
Query 20 104 B 104 B +0.0%
Query 21 1.5 MiB 4.4 MiB +183.6%
Query 22 2.3 MiB 7.6 MiB +233.2%
Query 23 12.4 MiB 31.9 MiB +157.3%
Query 24 5.9 MiB 59.9 MiB +919.6%
Query 25 8.4 MiB 183.0 MiB +2088.8%
Query 26 6.2 MiB 62.6 MiB +903.3%
Query 27 2.8 MiB 3.1 MiB +10.8%
Query 28 1.8 GiB 1.7 GiB -5.1%
Query 29 624 B 624 B +0.0%
Query 30 684.9 MiB 676.7 MiB -1.2%
Query 31 1.5 GiB 1.5 GiB -3.1%
Query 32 7.1 GiB 7.1 GiB -0.0%
Query 33 9.6 GiB 7.6 GiB -21.0%
Query 34 10.1 GiB 8.0 GiB -20.3%
Query 35 670.4 MiB 661.0 MiB -1.4%
Query 36 99.0 MiB 86.9 MiB -12.2%
Query 37 7.9 MiB 7.9 MiB +0.4%
Query 38 6.0 MiB 5.2 MiB -12.3%
Query 39 212.9 MiB 308.8 MiB +45.0%
Query 40 2.5 MiB 2.3 MiB -8.7%
Query 41 3.3 MiB 3.1 MiB -7.8%
Query 42 1.8 MiB 1.7 MiB -6.3%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
clickbench_partitioned base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 10.1 GiB 12.8 GiB 2.7 GiB 1.3×
clickbench_partitioned changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 8.0 GiB 11.2 GiB 3.1 GiB 1.4×
Resource Usage

clickbench_partitioned — base (merge-base)

Metric Value
Wall time 345.1s
Peak memory 12.8 GiB
Avg memory 5.3 GiB
CPU user 3439.9s
CPU sys 288.0s
Peak spill 0 B

clickbench_partitioned — branch

Metric Value
Wall time 160.1s
Peak memory 11.2 GiB
Avg memory 4.3 GiB
CPU user 1598.1s
CPU sys 175.9s
Peak spill 0 B

File an issue against this benchmark runner

@adriangbot

Copy link
Copy Markdown

🤖 Benchmark completed (GKE) | trigger

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

Comparing f323f02 (f323f02) to f323f02 diff

Run configuration
run benchmark clickbench_partitioned
env:
  DATAFUSION_EXECUTION_PARQUET_ENABLE_RLE_TO_DICTIONARY: "true"
  DATAFUSION_RUNTIME_MEMORY_LIMIT: "32G"
baseline:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
changed:
  ref: "f323f0265f72765c292397d9fed6f2d68f44eb9a"
  env:
    DATAFUSION_EXECUTION_PARQUET_DICTIONARY_READ_MAX_DISTINCT_VALUES: "40000"
resources:
  cpu: "12"
  memory: "65Gi"
  arch: "amd64"
CPU Details (lscpu)
Architecture:                            x86_64
CPU op-mode(s):                          32-bit, 64-bit
Address sizes:                           52 bits physical, 57 bits virtual
Byte Order:                              Little Endian
CPU(s):                                  16
On-line CPU(s) list:                     0-15
Vendor ID:                               GenuineIntel
Model name:                              INTEL(R) XEON(R) PLATINUM 8581C CPU @ 2.30GHz
CPU family:                              6
Model:                                   207
Thread(s) per core:                      2
Core(s) per socket:                      8
Socket(s):                               1
Stepping:                                2
BogoMIPS:                                4600.00
Flags:                                   fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht syscall nx pdpe1gb rdtscp lm constant_tsc rep_good nopl xtopology nonstop_tsc cpuid tsc_known_freq pni pclmulqdq monitor ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch ssbd ibrs ibpb stibp ibrs_enhanced fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm avx512f avx512dq rdseed adx smap avx512ifma clflushopt clwb avx512cd sha_ni avx512bw avx512vl xsaveopt xsavec xgetbv1 xsaves avx_vnni avx512_bf16 wbnoinvd arat avx512vbmi umip avx512_vbmi2 gfni vaes vpclmulqdq avx512_vnni avx512_bitalg avx512_vpopcntdq rdpid cldemote movdiri movdir64b fsrm md_clear serialize tsxldtrk amx_bf16 avx512_fp16 amx_tile amx_int8 arch_capabilities
Hypervisor vendor:                       KVM
Virtualization type:                     full
L1d cache:                               384 KiB (8 instances)
L1i cache:                               256 KiB (8 instances)
L2 cache:                                16 MiB (8 instances)
L3 cache:                                260 MiB (1 instance)
NUMA node(s):                            1
NUMA node0 CPU(s):                       0-15
Vulnerability Gather data sampling:      Not affected
Vulnerability Indirect target selection: Not affected
Vulnerability Itlb multihit:             Not affected
Vulnerability L1tf:                      Not affected
Vulnerability Mds:                       Not affected
Vulnerability Meltdown:                  Not affected
Vulnerability Mmio stale data:           Not affected
Vulnerability Reg file data sampling:    Not affected
Vulnerability Retbleed:                  Not affected
Vulnerability Spec rstack overflow:      Not affected
Vulnerability Spec store bypass:         Mitigation; Speculative Store Bypass disabled via prctl
Vulnerability Spectre v1:                Mitigation; usercopy/swapgs barriers and __user pointer sanitization
Vulnerability Spectre v2:                Mitigation; Enhanced / Automatic IBRS; IBPB conditional; PBRSB-eIBRS SW sequence; BHI BHI_DIS_S
Vulnerability Srbds:                     Not affected
Vulnerability Tsa:                       Not affected
Vulnerability Tsx async abort:           Not affected
Vulnerability Vmscape:                   Not affected
Details

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃        HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │     1.28 ms │                                     1.61 ms │  1.26x slower │
│ QQuery 1  │    20.28 ms │                                    19.95 ms │     no change │
│ QQuery 2  │    54.39 ms │                                    56.48 ms │     no change │
│ QQuery 3  │    57.76 ms │                                    58.72 ms │     no change │
│ QQuery 4  │   390.36 ms │                                   406.47 ms │     no change │
│ QQuery 5  │  1047.26 ms │                                   454.19 ms │ +2.31x faster │
│ QQuery 6  │     1.31 ms │                                     1.66 ms │  1.26x slower │
│ QQuery 7  │    20.70 ms │                                    21.37 ms │     no change │
│ QQuery 8  │   568.98 ms │                                   577.27 ms │     no change │
│ QQuery 9  │   695.14 ms │                                   672.34 ms │     no change │
│ QQuery 10 │   301.80 ms │                                   116.40 ms │ +2.59x faster │
│ QQuery 11 │   310.28 ms │                                   134.70 ms │ +2.30x faster │
│ QQuery 12 │  1026.10 ms │                                   427.55 ms │ +2.40x faster │
│ QQuery 13 │  1509.74 ms │                                   601.59 ms │ +2.51x faster │
│ QQuery 14 │  1183.30 ms │                                   433.43 ms │ +2.73x faster │
│ QQuery 15 │   446.77 ms │                                   459.58 ms │     no change │
│ QQuery 16 │  1892.48 ms │                                  1081.95 ms │ +1.75x faster │
│ QQuery 17 │  1893.92 ms │                                  1064.89 ms │ +1.78x faster │
│ QQuery 18 │  3084.64 ms │                                  2136.35 ms │ +1.44x faster │
│ QQuery 19 │    54.46 ms │                                    56.20 ms │     no change │
│ QQuery 20 │  1642.18 ms │                                   840.14 ms │ +1.95x faster │
│ QQuery 21 │  2125.30 ms │                                   868.63 ms │ +2.45x faster │
│ QQuery 22 │  5043.53 ms │                                  1674.75 ms │ +3.01x faster │
│ QQuery 23 │ 22765.51 ms │                                  5562.46 ms │ +4.09x faster │
│ QQuery 24 │   142.22 ms │                                    63.75 ms │ +2.23x faster │
│ QQuery 25 │   540.26 ms │                                   164.52 ms │ +3.28x faster │
│ QQuery 26 │   137.97 ms │                                    63.41 ms │ +2.18x faster │
│ QQuery 27 │  1738.37 ms │                                   853.48 ms │ +2.04x faster │
│ QQuery 28 │  3421.01 ms │                                  3715.85 ms │  1.09x slower │
│ QQuery 29 │    56.32 ms │                                    54.97 ms │     no change │
│ QQuery 30 │   845.84 ms │                                   492.84 ms │ +1.72x faster │
│ QQuery 31 │   897.02 ms │                                   518.94 ms │ +1.73x faster │
│ QQuery 32 │  1754.45 ms │                                  1658.04 ms │ +1.06x faster │
│ QQuery 33 │  4429.47 ms │                                  2273.09 ms │ +1.95x faster │
│ QQuery 34 │  4475.81 ms │                                  2312.78 ms │ +1.94x faster │
│ QQuery 35 │   398.05 ms │                                   390.55 ms │     no change │
│ QQuery 36 │   144.39 ms │                                    91.42 ms │ +1.58x faster │
│ QQuery 37 │   127.60 ms │                                    43.54 ms │ +2.93x faster │
│ QQuery 38 │   101.50 ms │                                    55.63 ms │ +1.82x faster │
│ QQuery 39 │   377.82 ms │                                   204.67 ms │ +1.85x faster │
│ QQuery 40 │    21.88 ms │                                    19.62 ms │ +1.12x faster │
│ QQuery 41 │    18.03 ms │                                    17.62 ms │     no change │
│ QQuery 42 │    13.75 ms │                                    13.85 ms │     no change │
└───────────┴─────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 65779.21ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 30737.27ms │
│ Average Time (HEAD)                                        │  1529.75ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   714.82ms │
│ Queries Faster                                             │         27 │
│ Queries Slower                                             │          3 │
│ Queries with No Change                                     │         13 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

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

Comparing HEAD and rich-T-kid_parquet-dictionary-ndv-threshold
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query     ┃                                      HEAD ┃ rich-T-kid_parquet-dictionary-ndv-threshold ┃        Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0  │               1.28 / 2.38 ±2.06 / 6.49 ms │                 1.61 / 2.71 ±2.09 / 6.89 ms │  1.14x slower │
│ QQuery 1  │            20.28 / 21.03 ±0.74 / 22.19 ms │              19.95 / 21.54 ±0.98 / 22.94 ms │     no change │
│ QQuery 2  │            54.39 / 58.55 ±3.27 / 63.95 ms │              56.48 / 59.20 ±1.87 / 61.90 ms │     no change │
│ QQuery 3  │            57.76 / 58.79 ±0.62 / 59.65 ms │              58.72 / 60.08 ±1.61 / 62.83 ms │     no change │
│ QQuery 4  │         390.36 / 405.66 ±8.71 / 416.79 ms │           406.47 / 414.94 ±7.59 / 426.17 ms │     no change │
│ QQuery 5  │     1047.26 / 1069.22 ±17.56 / 1092.54 ms │          454.19 / 469.31 ±10.52 / 483.33 ms │ +2.28x faster │
│ QQuery 6  │               1.31 / 1.57 ±0.36 / 2.23 ms │                 1.66 / 1.90 ±0.31 / 2.49 ms │  1.21x slower │
│ QQuery 7  │            20.70 / 21.12 ±0.29 / 21.42 ms │              21.37 / 25.03 ±3.62 / 31.13 ms │  1.19x slower │
│ QQuery 8  │         568.98 / 578.40 ±5.16 / 583.97 ms │           577.27 / 590.85 ±9.57 / 606.45 ms │     no change │
│ QQuery 9  │        695.14 / 712.56 ±12.38 / 732.02 ms │          672.34 / 729.16 ±29.43 / 750.57 ms │     no change │
│ QQuery 10 │         301.80 / 310.61 ±6.40 / 319.67 ms │           116.40 / 121.84 ±3.65 / 126.14 ms │ +2.55x faster │
│ QQuery 11 │         310.28 / 324.14 ±9.84 / 339.47 ms │           134.70 / 137.17 ±1.49 / 138.80 ms │ +2.36x faster │
│ QQuery 12 │     1026.10 / 1041.06 ±11.81 / 1055.73 ms │          427.55 / 450.69 ±15.09 / 468.29 ms │ +2.31x faster │
│ QQuery 13 │     1509.74 / 1565.61 ±63.03 / 1672.41 ms │          601.59 / 626.79 ±24.67 / 664.35 ms │ +2.50x faster │
│ QQuery 14 │     1183.30 / 1199.16 ±11.79 / 1213.27 ms │          433.43 / 450.66 ±12.92 / 466.02 ms │ +2.66x faster │
│ QQuery 15 │         446.77 / 455.09 ±5.88 / 462.83 ms │          459.58 / 480.96 ±11.13 / 490.99 ms │  1.06x slower │
│ QQuery 16 │     1892.48 / 1939.34 ±29.24 / 1974.57 ms │       1081.95 / 1097.50 ±14.44 / 1119.88 ms │ +1.77x faster │
│ QQuery 17 │     1893.92 / 1910.36 ±15.56 / 1938.73 ms │       1064.89 / 1090.19 ±19.77 / 1122.93 ms │ +1.75x faster │
│ QQuery 18 │     3084.64 / 3117.95 ±27.36 / 3155.49 ms │        2136.35 / 2148.36 ±9.70 / 2157.34 ms │ +1.45x faster │
│ QQuery 19 │            54.46 / 61.68 ±7.74 / 74.96 ms │              56.20 / 60.99 ±4.50 / 66.38 ms │     no change │
│ QQuery 20 │     1642.18 / 1685.86 ±23.56 / 1706.08 ms │          840.14 / 876.89 ±19.49 / 894.57 ms │ +1.92x faster │
│ QQuery 21 │     2125.30 / 2182.29 ±42.12 / 2243.64 ms │           868.63 / 883.75 ±8.78 / 892.75 ms │ +2.47x faster │
│ QQuery 22 │     5043.53 / 5124.47 ±55.09 / 5193.65 ms │       1674.75 / 1703.54 ±21.00 / 1734.42 ms │ +3.01x faster │
│ QQuery 23 │ 22765.51 / 23969.03 ±618.18 / 24486.10 ms │       5562.46 / 5631.33 ±38.97 / 5672.87 ms │ +4.26x faster │
│ QQuery 24 │        142.22 / 162.06 ±14.85 / 187.28 ms │              63.75 / 64.69 ±0.63 / 65.43 ms │ +2.51x faster │
│ QQuery 25 │        540.26 / 553.56 ±10.96 / 567.30 ms │           164.52 / 167.60 ±2.78 / 172.75 ms │ +3.30x faster │
│ QQuery 26 │         137.97 / 147.56 ±6.76 / 157.14 ms │              63.41 / 64.68 ±0.92 / 65.76 ms │ +2.28x faster │
│ QQuery 27 │     1738.37 / 1779.66 ±47.50 / 1871.52 ms │           853.48 / 864.82 ±8.19 / 877.92 ms │ +2.06x faster │
│ QQuery 28 │     3421.01 / 3521.77 ±52.30 / 3565.70 ms │       3715.85 / 3779.50 ±54.43 / 3876.81 ms │  1.07x slower │
│ QQuery 29 │            56.32 / 61.85 ±7.94 / 77.49 ms │              54.97 / 62.07 ±6.99 / 71.51 ms │     no change │
│ QQuery 30 │        845.84 / 861.68 ±10.24 / 872.70 ms │          492.84 / 505.50 ±10.25 / 518.32 ms │ +1.70x faster │
│ QQuery 31 │        897.02 / 911.41 ±14.39 / 934.42 ms │          518.94 / 536.34 ±13.63 / 559.20 ms │ +1.70x faster │
│ QQuery 32 │     1754.45 / 1810.13 ±50.97 / 1876.76 ms │       1658.04 / 1692.75 ±40.68 / 1770.39 ms │ +1.07x faster │
│ QQuery 33 │     4429.47 / 4587.28 ±86.65 / 4686.85 ms │       2273.09 / 2319.63 ±33.26 / 2361.62 ms │ +1.98x faster │
│ QQuery 34 │     4475.81 / 4586.24 ±76.67 / 4715.04 ms │       2312.78 / 2353.16 ±22.65 / 2374.23 ms │ +1.95x faster │
│ QQuery 35 │        398.05 / 415.48 ±26.33 / 467.50 ms │          390.55 / 410.88 ±15.38 / 436.78 ms │     no change │
│ QQuery 36 │        144.39 / 159.16 ±12.90 / 182.85 ms │             91.42 / 98.55 ±3.73 / 101.44 ms │ +1.62x faster │
│ QQuery 37 │         127.60 / 132.51 ±4.77 / 140.12 ms │              43.54 / 45.88 ±1.84 / 49.19 ms │ +2.89x faster │
│ QQuery 38 │         101.50 / 102.68 ±1.06 / 104.62 ms │              55.63 / 62.00 ±5.66 / 70.81 ms │ +1.66x faster │
│ QQuery 39 │        377.82 / 396.79 ±18.24 / 428.41 ms │           204.67 / 214.07 ±6.33 / 223.66 ms │ +1.85x faster │
│ QQuery 40 │            21.88 / 23.79 ±1.78 / 27.06 ms │              19.62 / 23.14 ±5.57 / 34.14 ms │     no change │
│ QQuery 41 │            18.03 / 20.10 ±1.79 / 22.94 ms │              17.62 / 20.33 ±2.89 / 25.76 ms │     no change │
│ QQuery 42 │            13.75 / 15.68 ±1.45 / 17.50 ms │              13.85 / 15.21 ±1.44 / 17.16 ms │     no change │
└───────────┴───────────────────────────────────────────┴─────────────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary                                          ┃            ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD)                                          │ 68065.33ms │
│ Total Time (rich-T-kid_parquet-dictionary-ndv-threshold)   │ 31436.20ms │
│ Average Time (HEAD)                                        │  1582.91ms │
│ Average Time (rich-T-kid_parquet-dictionary-ndv-threshold) │   731.07ms │
│ Queries Faster                                             │         26 │
│ Queries Slower                                             │          5 │
│ Queries with No Change                                     │         12 │
│ Queries with Failure                                       │          0 │
└────────────────────────────────────────────────────────────┴────────────┘

Memory Pool Peaks

Peak MemoryPool reservation per query — what DataFusion's accounting believes it reserved. Recorded only for benchmarks that write a results JSON (the dfbench suites, and the suites bench.sh runs through the Criterion SQL harness since #25644), and only under a memory limit: DATAFUSION_RUNTIME_MEMORY_LIMIT, or one the suite sets itself.

Base: f323f0265f72765c292397d9fed6f2d68f44eb9a | Changed: f323f0265f72765c292397d9fed6f2d68f44eb9a

clickbench_partitioned

Query Base Changed Change
Query 0 0 B 0 B 0.0%
Query 1 104 B 104 B +0.0%
Query 2 936 B 936 B +0.0%
Query 3 312 B 312 B +0.0%
Query 4 786.3 MiB 831.2 MiB +5.7%
Query 5 1.3 GiB 1.2 GiB -8.3%
Query 6 0 B 0 B 0.0%
Query 7 40.4 MiB 60.2 MiB +48.9%
Query 8 899.6 MiB 924.7 MiB +2.8%
Query 9 635.9 MiB 678.0 MiB +6.6%
Query 10 70.8 MiB 112.4 MiB +58.8%
Query 11 80.2 MiB 112.8 MiB +40.7%
Query 12 1.5 GiB 1.4 GiB -6.5%
Query 13 1.7 GiB 1.6 GiB -10.4%
Query 14 1.6 GiB 1.3 GiB -20.2%
Query 15 1.1 GiB 1.2 GiB +8.1%
Query 16 3.2 GiB 3.4 GiB +6.4%
Query 17 3.1 GiB 3.3 GiB +7.0%
Query 18 7.0 GiB 8.2 GiB +15.8%
Query 19 0 B 0 B 0.0%
Query 20 104 B 104 B +0.0%
Query 21 1.3 MiB 4.4 MiB +234.3%
Query 22 2.5 MiB 7.7 MiB +209.3%
Query 23 12.6 MiB 33.0 MiB +161.1%
Query 24 6.7 MiB 61.1 MiB +818.5%
Query 25 7.8 MiB 179.6 MiB +2195.2%
Query 26 7.1 MiB 63.6 MiB +794.7%
Query 27 2.8 MiB 3.1 MiB +10.8%
Query 28 1.8 GiB 1.8 GiB -0.8%
Query 29 624 B 624 B +0.0%
Query 30 674.8 MiB 701.8 MiB +4.0%
Query 31 1.5 GiB 1.5 GiB +0.0%
Query 32 7.1 GiB 7.1 GiB -0.2%
Query 33 9.9 GiB 7.8 GiB -20.9%
Query 34 9.6 GiB 8.0 GiB -16.6%
Query 35 673.0 MiB 669.7 MiB -0.5%
Query 36 97.8 MiB 115.5 MiB +18.1%
Query 37 7.6 MiB 7.9 MiB +3.5%
Query 38 6.3 MiB 5.2 MiB -17.3%
Query 39 212.9 MiB 299.2 MiB +40.5%
Query 40 2.3 MiB 2.3 MiB +1.4%
Query 41 2.8 MiB 3.3 MiB +18.5%
Query 42 1.7 MiB 1.8 MiB +5.8%

Pool accounting vs. process RSS

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

Benchmark Side Max pool peak Peak RSS Gap RSS / pool
clickbench_partitioned base (f323f0265f72765c292397d9fed6f2d68f44eb9a) 9.9 GiB 12.9 GiB 3.0 GiB 1.3×
clickbench_partitioned changed (f323f0265f72765c292397d9fed6f2d68f44eb9a) 8.2 GiB 10.5 GiB 2.3 GiB 1.3×
Resource Usage

clickbench_partitioned — base (merge-base)

Metric Value
Wall time 345.1s
Peak memory 12.9 GiB
Avg memory 5.3 GiB
CPU user 3440.3s
CPU sys 289.2s
Peak spill 0 B

clickbench_partitioned — branch

Metric Value
Wall time 160.0s
Peak memory 10.5 GiB
Avg memory 4.1 GiB
CPU user 1585.6s
CPU sys 175.2s
Peak spill 0 B

File an issue against this benchmark runner

@Rich-T-kid

Copy link
Copy Markdown
Contributor Author

thanks for taking a look @jayzhan211, I updated the PR in the latest commit to address your comments. I also provided for context in the pr description for different cases. thanks 🙇

@jayzhan211 jayzhan211 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@Rich-T-kid a few optional items on the description:

  • The benchmark table compares the cap with unbounded promotion, but a user deciding whether to turn the option on starts from the default (plain reads). ClickBench hits has no NDV statistics and its row groups hold 450k–549k rows, so cap = 20000 promotes no ClickBench column (I checked with information_schema.columns). The −51% therefore means "unbounded promotion is about 2x slower than plain", not that the cap speeds anything up. One run of cap vs default would show what the option buys.
  • The Rationale has no number. Something like: "Reading a low-cardinality Parquet string column as a dictionary makes GROUP BY on it about 2x faster (30k distinct 100-byte values: 0.09 s → 0.045 s), but enabling enable_rle_to_dictionary for every column makes high-cardinality columns several times slower (295k distinct values per row group: 4.4x). This PR adds a cap so only columns whose estimated cardinality is under it are read as dictionaries."

Comment on lines +74 to +76
if row_count > 0 && column.dictionary_page_offset().is_none() {
return None;
}

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@Rich-T-kid thanks for the updates! This check still promotes a column chunk whose dictionary outgrew the writer's page limit, so its later pages are PLAIN. With cap = 40000, a file of 100-byte strings with 30k distinct values per row group is promoted, and its queries run 1.2–2.5x slower than plain reads. Requiring every data page to be dictionary-encoded, as your reopen comment on #24113 describes, keeps that file plain. page_encoding_stats_mask is decoded by default, so the check needs no extra I/O:

Suggested change
if row_count > 0 && column.dictionary_page_offset().is_none() {
return None;
}
// A chunk whose dictionary outgrew the writer's page limit
// continues in PLAIN pages, and reads slower as a dictionary
// than as plain values.
let all_dictionary_pages = column
.dictionary_page_offset()
.is_some()
&& column.page_encoding_stats_mask().is_some_and(|mask| {
mask.is_only(parquet::basic::Encoding::RLE_DICTIONARY)
|| mask.is_only(
parquet::basic::Encoding::PLAIN_DICTIONARY,
)
});
if row_count > 0 && !all_dictionary_pages {
return None;
}

Files without page encoding stats then stay plain under the cap. The docs that say "every nonempty row group has a dictionary page" would change to "every data page is dictionary encoded".

Measurements and a regression test (fails on this head with `Dictionary(Int32, Utf8)`, passes with the change)

Data: 4 files × 2M rows, one row group each, cat = concat(repeat('x', 95), lpad(CAST(value % 30000 AS VARCHAR), 5, '0')), NDV written, default dictionary_page_size_limit (1 MiB < 30k × 100 B, so the writer falls back to PLAIN). Median of 5, three rounds, datafusion-cli release-nonlto, M-series laptop. Results are identical in both modes.

query plain cap = 40000 (promoted)
SELECT cat, count(*), sum(v) FROM t GROUP BY cat 0.096–0.133 s 0.142–0.248 s
SELECT count(DISTINCT cat) FROM t 0.074–0.113 s 0.121–0.190 s
SELECT count(*) FROM t WHERE cat LIKE '%12345' 0.037–0.045 s 0.076–0.112 s

With the change, the files stay Utf8View. The same data written with a 64 MiB page limit (no fallback) is still promoted and still wins on the group-by (0.090–0.126 s → 0.042–0.062 s).

Test, placed after DROP TABLE threshold_no_dict_page;:

# A column whose dictionary page outgrew its size limit continues in PLAIN
# pages, so it stays plain even when its NDV is within the cap.
query I
COPY (SELECT column1 AS fallback_page FROM (VALUES ('a'), ('b')))
TO 'test_files/scratch/parquet_rle_to_dictionary/threshold_fallback_page/p.parquet'
STORED AS PARQUET OPTIONS (
  'format.dictionary_page_size_limit' 1,
  'format.write_batch_size' 1,
  'format.data_page_row_count_limit' 1
);
----
2

statement ok
CREATE EXTERNAL TABLE threshold_fallback_page STORED AS PARQUET LOCATION 'test_files/scratch/parquet_rle_to_dictionary/threshold_fallback_page/';

query T
SELECT DISTINCT arrow_typeof(fallback_page) FROM threshold_fallback_page;
----
Utf8View

statement ok
DROP TABLE threshold_fallback_page;

Comment on lines +1470 to +1475
/// Large row-group dictionaries can be slower because each output batch can
/// retain the whole dictionary. The crossover is workload-dependent: in one
/// 20-million-row group-by benchmark with 40-byte strings, the case with about
/// 100,000 distinct values per row group was faster than plain reads, while the
/// case with about 295,000 was slower. Use a conservative cap and benchmark
/// representative queries.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Optional cleanup: the cap counts the per-file sum, but these numbers are per row group, so a reader who sets cap = 100000 from them won't promote the faster file. That benchmark wrote five row groups per file, so its per-file sums were 500,000 and about 1,460,000; with cap = 100000 the 100k file stays Utf8View, and it is promoted only at cap = 500000. Stating the numbers in the cap's unit:

Suggested change
/// Large row-group dictionaries can be slower because each output batch can
/// retain the whole dictionary. The crossover is workload-dependent: in one
/// 20-million-row group-by benchmark with 40-byte strings, the case with about
/// 100,000 distinct values per row group was faster than plain reads, while the
/// case with about 295,000 was slower. Use a conservative cap and benchmark
/// representative queries.
/// Large row-group dictionaries can be slower because each output batch can
/// retain the whole dictionary. The crossover is workload-dependent: in one
/// 20-million-row group-by benchmark with 40-byte strings and five row groups
/// per file, files with a per-file sum of about 500,000 (100,000 per row group)
/// were faster than plain reads, while files with about 1,460,000 (about
/// 295,000 per row group) were slower. Use a conservative cap and benchmark
/// representative queries.

configs.md and information_schema.slt would need regenerating.

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

Labels

auto detected api change Auto detected API change common Related to common crate datasource Changes to the datasource crate documentation Improvements or additions to documentation proto Related to proto crate sqllogictest SQL Logic Tests (.slt)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Allow users to define the cardinality threshold for dictionary reads

3 participants