perf: simplify HashJoinExec dynamic filter, drop CASE routing

[adriangb]

Which issue does this PR close?

• Refs Pushing down HashJoinExec build side dynamic filters makes tpch queries slower #19858 — "Pushing down HashJoinExec build side dynamic filters makes tpch queries slower". This PR addresses the structural side of the problem; the adaptive selectivity work ((Test) Advanced adaptive filter selectivity evaluation #20363) is complementary and would absorb the residual high-N + scan-pushdown costs called out below.

Rationale for this change

Today the Partitioned-mode HashJoinExec builds a dynamic filter that's structured around the repartition layout:

CASE hash_repartition % N
WHEN 0 THEN p0_bounds AND (p0_inlist | p0_hash_lookup)
WHEN 1 THEN p1_bounds AND (p1_inlist | p1_hash_lookup)
...
ELSE false
END

Two problems with this:

1. Per-row routing-hash cost. Every probe row pays a hash_repartition % N even though the partition's hash table will be probed anyway with a different seed (HASH_JOIN_SEED).
2. Coupling. The dynamic filter's shape depends on how the build side was repartitioned, even though semantically the filter is just "is this key somewhere in the build side?"

This PR replaces the routing CASE with a structure that depends only on the content of the build side, not its layout, and adds a cross-partition merged `IN (SET)` fast path so small joins can participate in parquet stats / bloom-filter pruning at the scan side.

What changes are included in this PR?

Five commits:

1. `perf: collapse all-Map dynamic filter into MultiMapLookupExpr` — new `MultiMapLookupExpr` hashes the join keys once with `HASH_JOIN_SEED` and ORs `contain_hashes()` across every reported partition's hash table.
2. `perf: collapse small all-InList dynamic filter into one cross-partition IN (SET)` — when every reported partition contributed an InList array and the cross-partition union is small enough, concatenate them into a single global `IN (SET)`.
3. `refactor: drop CASE routing from Partitioned dynamic filters` — `PushdownStrategy` now always carries the `Map` (the join's hash table is built unconditionally) plus an optional InList array; the routing CASE goes away.
4. `refactor: dedup cross-partition InList, reuse per-partition cap` — combine path deduplicates by `ScalarValue` and re-gates on distinct count. The existing `optimizer.hash_join_inlist_pushdown_max_distinct_values` knob caps both per-partition InList eligibility and the cross-partition merged set.
5. `docs: rewrite comments to describe the code, not the change` — comment cleanup.

Final filter-shape matrix

input	shape
`enable_dynamic_filter_pushdown=false`	no filter installed
`CollectLeft`, build empty	filter stays at `lit(true)`
`CollectLeft`, build small (under inlist caps)	`bounds AND IN (SET)`
`CollectLeft`, build large	`bounds AND hash_lookup`
`Partitioned`, all reported partitions empty	`lit(false)`
`Partitioned`, any partition canceled	`lit(true)`
`Partitioned`, every partition InList AND combined ≤ cap	`bounds AND IN (SET)` (parquet-prunable)
`Partitioned`, anything else	`bounds AND multi_hash_lookup`

No more `CASE`, no more `hash_repartition`, no more `REPARTITION_RANDOM_STATE` in the dynamic-filter path.

Performance

Per-row cost (Partitioned mode)

Let `N` = number of build-side partitions. For each probe row evaluated by the dynamic filter:

Shape	Hashes	Hash-table probes	Per-row cost
Legacy CASE-by-routing	2 (routing + lookup)	1 (one branch)	`O(1)`
`multi_hash_lookup` (all-Map fast path)	1 (lookup)	N (one per map)	`O(N)`
Merged `IN (SET)` (small-union fast path)	1	1 (one `static_filter`)	`O(1)` + scan-side row-group / bloom pruning

Two countervailing forces shape the result:

• Per-batch overhead: legacy CASE evaluates through `CaseExpr`, which can scale poorly with the number of `WHEN` branches at high N. `multi_hash_lookup` is a single straight-line OR loop and avoids this.
• Per-row overhead at high N: `multi_hash_lookup` does N probes per row. When the filter runs in the parquet scan hot loop (pushdown=true), this matters more than the CASE per-batch overhead. When the filter runs at join time (pushdown=false), it doesn't.

Benchmarks (TPC-H, runner = c4a-highmem-16, ARM Neoverse-V2, 16 cores)

Triggered four runs covering both N regimes and both pushdown configs.

Default partitioning (N ≈ ncores ≈ 16)

Config	`Total Time (HEAD)`	`Total Time (PR)`	Change
`pushdown_filters=false`	912.79 ms	916.39 ms	+0.4% (noise)
`pushdown_filters=true`	1259.10 ms	1166.39 ms	−7.4%

In the `pushdown=true` run the wins concentrate on queries where the dynamic filter feeds parquet stats / bloom-filter pruning at the scan:

Query	HEAD	PR	Change
Q17	165 / 171 ms	56 / 56 ms	+2.96× faster
Q18	77 / 77 ms	59 / 60 ms	+1.30× faster
Q20	47 / 47 ms	44 / 49 ms	+1.06× faster
Q3	49 / 49 ms	54 / 55 ms	1.10× slower
Q5	73 / 76 ms	77 / 82 ms	1.05× slower
Q9	112 / 115 ms	124 / 127 ms	1.10× slower
Q13	50 / 50 ms	64 / 64 ms	1.29× slower
Q14	41 / 42 ms	46 / 47 ms	1.11× slower

Q17 alone is 109 ms of the 93 ms total wall-clock improvement — that's the original issue's regression, fixed. The small-query regressions (Q3/Q5/Q9/Q13/Q14) are the all-Map shape paying `O(N)` probes per row for moderate-to-large build sides where the bounds prefix doesn't prune much.

High partition count (`target_partitions=128`)

A stress test of partition-count scaling on the same 16-core box.

Config	`Total Time (HEAD)`	`Total Time (PR)`	Change
`pushdown_filters=false`	2566.31 ms	1756.80 ms	−31.5% (1.46× faster)
`pushdown_filters=true`	3209.06 ms	3636.15 ms	+13.3%

Pushdown=false @ N=128: 11 queries faster, 0 slower, 11 unchanged. `multi_hash_lookup` cleanly beats the legacy 128-branch `CaseExpr` evaluation. Big wins on Q3 (1.68×), Q5 (1.82×), Q7 (1.75×), Q8 (2.15×), Q9 (1.64×), Q12 (1.81×), Q13 (1.76×), Q17 (1.59×), Q18 (1.29×), Q20 (2.05×), Q21 (1.43×).

Pushdown=true @ N=128: 4 faster, 9 slower. This is the case where the filter runs in the scan hot loop and `multi_hash_lookup`'s `O(N)` probes per row dominate. The wins (Q17 1.87×, Q18 1.39×, Q20 1.76×) survive because their merged `IN (SET)` prunes whole row groups before the per-row filter ever runs. The losses (Q5 1.88×, Q9 1.64×, Q21 1.62×, Q14 1.51×, Q8 1.37×) are the same all-Map shape paying 128 probes per row.

Summary of the regime grid

	pushdown=false	pushdown=true
N ≈ ncores (default)	noise (≈0%)	−7.4% (Q17 +2.96×)
N = 128	−31.5%	+13.3% (Q17 still +1.87×, but Q5/Q9 etc. regress)

Three of the four configs are wins (one big), one is a regression. The single regressing config is `high N + scan-side pushdown`, which is exactly the scenario `OptionalFilterPhysicalExpr` from #20363 is designed to absorb: the adaptive tracker would measure `multi_hash_lookup`'s low `bytes_pruned_per_second_of_eval_time` for queries like Q5/Q9/Q21 and drop the filter, while keeping Q17/Q18/Q20 (which prune scans aggressively).

A possible structural follow-up — re-introducing partition routing inside `MultiMapLookupExpr` (1 routing hash + 1 probe, so per-row cost matches legacy CASE) — would close the regression at any N, with or without #20363.

Are these changes tested?

• Existing `joins::hash_join` lib tests pass (380 tests).
• `physical_optimizer::filter_pushdown` integration tests pass (51 tests). Two snapshots updated:
  • `test_hashjoin_dynamic_filter_pushdown_partitioned` now produces `bounds AND struct(...) IN (SET) ([...])` directly (the size-gated path subsumes what `force_hash_collisions` previously fell into).
  • `test_hashjoin_hash_table_pushdown_partitioned` now positively asserts `multi_hash_lookup` and the absence of `hash_repartition`.
• `information_schema.slt` updated to reflect the new default for `hash_join_inlist_pushdown_max_distinct_values`.
• `push_down_filter_parquet.slt` passes.
• `cargo clippy -p datafusion-physical-plan --all-targets -- -D warnings` is clean.

Are there any user-facing changes?

• Default lowered: `optimizer.hash_join_inlist_pushdown_max_distinct_values` 150 → 20. Affects which build-side shapes choose InList vs. hash-table pushdown per partition, and now also gates the cross-partition merged InList. Users who depend on the old per-partition behavior can set it back to 150 in their config. Doc string in `config.rs` and `docs/source/user-guide/configs.md` are updated.
• Plan output: `EXPLAIN ANALYZE` for `Partitioned` hash joins no longer shows `CASE hash_repartition % N WHEN ...`. Instead it shows either `bounds AND struct(...) IN (SET) ([...])` (small joins) or `bounds AND multi_hash_lookup` (everything else).

🤖 Generated with Claude Code

[adriangb]

run benchmark tcph

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: falseDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: falsechanged:
ref: HEADenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: falseDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: false

[adriangb]

run benchmark tcph

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: trueDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: truechanged:
ref: HEADenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: trueDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: true

[adriangbot]

🤖 Criterion benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4345503296-1912-446v6 6.12.55+ #1 SMP Sun Feb 1 08:59:41 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing HEAD (7a8272f) to main diff
BENCH_NAME=tcph
BENCH_COMMAND=cargo bench --features=parquet --bench tcph
BENCH_FILTER=
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

Benchmark for this request failed.

Last 20 lines of output:

Click to expand

 Downloaded compression-codecs v0.4.38
Downloaded cfg_aliases v0.2.1
Downloaded aws-credential-types v1.2.14
Downloaded async-stream v0.3.6
Downloaded atoi v2.0.0
Downloaded arrow-string v58.1.0
Downloaded anstyle v1.0.14
Downloaded axum-core v0.5.6
Downloaded aws-smithy-xml v0.60.15
Downloaded aws-smithy-observability v0.2.6
Downloaded autocfg v1.5.0
Downloaded ahash v0.8.12
Downloaded anstream v1.0.0
Downloaded async-recursion v1.1.1
Downloaded async-ffi v0.5.0
Downloaded anes v0.1.6
Blocking waiting for file lock on package cache
error: no bench target named `tcph` in default-run packages
help: a target with a similar name exists: `chr`

---

File an issue against this benchmark runner

[adriangbot]

🤖 Criterion benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4345504799-1913-ftp4k 6.12.55+ #1 SMP Sun Feb 1 08:59:41 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing HEAD (7a8272f) to main diff
BENCH_NAME=tcph
BENCH_COMMAND=cargo bench --features=parquet --bench tcph
BENCH_FILTER=
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

Benchmark for this request failed.

Last 20 lines of output:

Click to expand

 Downloaded clap v4.6.1
Downloaded ciborium v0.2.2
Downloaded aws-smithy-xml v0.60.15
Downloaded async-stream-impl v0.3.6
Downloaded arrow-csv v58.1.0
Downloaded compression-core v0.4.32
Downloaded bytes-utils v0.1.4
Downloaded blake2 v0.10.6
Downloaded bitflags v2.11.1
Downloaded aws-types v1.3.15
Downloaded aws-smithy-async v1.2.14
Downloaded anstyle v1.0.14
Downloaded clap_lex v1.1.0
Downloaded aws-smithy-query v0.60.15
Downloaded aws-smithy-observability v0.2.6
Downloaded arrayref v0.3.9
Blocking waiting for file lock on package cache
error: no bench target named `tcph` in default-run packages
help: a target with a similar name exists: `chr`

---

File an issue against this benchmark runner

[adriangb]

run benchmark tpch

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: trueDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: truechanged:
ref: HEADenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: trueDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: true

[adriangb]

run benchmark tpch

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: falseDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: falsechanged:
ref: HEADenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: falseDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: false

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4345545637-1915-vxjfd 6.12.55+ #1 SMP Sun Feb 1 08:59:41 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing HEAD (7a8272f) to main diff using: tpch
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4345547022-1916-gt4f5 6.12.55+ #1 SMP Sun Feb 1 08:59:41 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing HEAD (7a8272f) to main diff using: tpch
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 43.41 / 44.20 ±0.95 / 46.07 ms │ 43.37 / 44.58 ±1.47 / 47.38 ms │ no change │
│ QQuery 2 │ 25.18 / 25.59 ±0.35 / 26.10 ms │ 25.09 / 25.23 ±0.24 / 25.72 ms │ no change │
│ QQuery 3 │ 48.02 / 48.76 ±0.41 / 49.09 ms │ 52.88 / 53.66 ±0.85 / 55.19 ms │ 1.10x slower │
│ QQuery 4 │ 22.73 / 22.86 ±0.11 / 23.02 ms │ 22.47 / 22.68 ±0.13 / 22.88 ms │ no change │
│ QQuery 5 │ 70.47 / 72.83 ±2.26 / 76.28 ms │ 73.07 / 76.76 ±3.21 / 81.56 ms │ 1.05x slower │
│ QQuery 6 │ 35.98 / 36.50 ±0.54 / 37.55 ms │ 36.12 / 37.55 ±1.15 / 38.84 ms │ no change │
│ QQuery 7 │ 57.86 / 58.04 ±0.21 / 58.45 ms │ 55.25 / 56.19 ±1.09 / 58.34 ms │ no change │
│ QQuery 8 │ 79.29 / 80.19 ±0.67 / 81.15 ms │ 76.28 / 76.74 ±0.51 / 77.67 ms │ no change │
│ QQuery 9 │ 110.20 / 112.39 ±2.16 / 115.28 ms │ 120.81 / 124.17 ±2.52 / 126.74 ms │ 1.10x slower │
│ QQuery 10 │ 76.60 / 77.36 ±0.60 / 78.04 ms │ 76.50 / 76.70 ±0.31 / 77.31 ms │ no change │
│ QQuery 11 │ 16.63 / 17.50 ±0.98 / 19.19 ms │ 16.15 / 16.94 ±0.76 / 18.03 ms │ no change │
│ QQuery 12 │ 46.65 / 47.21 ±0.60 / 48.34 ms │ 45.96 / 47.22 ±2.09 / 51.39 ms │ no change │
│ QQuery 13 │ 48.67 / 49.58 ±0.56 / 50.27 ms │ 63.30 / 63.75 ±0.36 / 64.23 ms │ 1.29x slower │
│ QQuery 14 │ 40.73 / 41.21 ±0.61 / 42.29 ms │ 44.72 / 45.58 ±0.69 / 46.50 ms │ 1.11x slower │
│ QQuery 15 │ 45.55 / 46.45 ±0.75 / 47.39 ms │ 45.01 / 46.74 ±1.97 / 50.52 ms │ no change │
│ QQuery 16 │ 23.72 / 24.27 ±0.57 / 25.18 ms │ 23.46 / 23.72 ±0.18 / 23.98 ms │ no change │
│ QQuery 17 │ 161.67 / 165.40 ±3.60 / 171.03 ms │ 55.41 / 55.87 ±0.49 / 56.52 ms │ +2.96x faster │
│ QQuery 18 │ 75.97 / 76.74 ±0.55 / 77.28 ms │ 57.68 / 58.81 ±0.99 / 59.97 ms │ +1.30x faster │
│ QQuery 19 │ 42.01 / 42.27 ±0.20 / 42.49 ms │ 42.38 / 42.80 ±0.29 / 43.14 ms │ no change │
│ QQuery 20 │ 45.92 / 46.56 ±0.53 / 47.48 ms │ 42.72 / 43.95 ±2.31 / 48.58 ms │ +1.06x faster │
│ QQuery 21 │ 84.97 / 86.79 ±2.22 / 91.01 ms │ 89.70 / 90.48 ±0.67 / 91.45 ms │ no change │
│ QQuery 22 │ 35.83 / 36.41 ±0.32 / 36.77 ms │ 36.07 / 36.24 ±0.13 / 36.46 ms │ no change │
└───────────┴───────────────────────────────────┴─────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 1259.10ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 1166.39ms │
│ Average Time (HEAD) │ 57.23ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 53.02ms │
│ Queries Faster │ 3 │
│ Queries Slower │ 5 │
│ Queries with No Change │ 14 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴───────────┘

Resource Usage

tpch — base (merge-base)

Metric	Value
Wall time	10.0s
Peak memory	5.3 GiB
Avg memory	4.6 GiB
CPU user	47.7s
CPU sys	2.6s
Peak spill	0 B

tpch — branch

Metric	Value
Wall time	10.0s
Peak memory	5.3 GiB
Avg memory	4.5 GiB
CPU user	42.2s
CPU sys	2.7s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 41.66 / 42.57 ±1.04 / 44.21 ms │ 40.50 / 41.42 ±1.12 / 43.61 ms │ no change │
│ QQuery 2 │ 21.92 / 22.31 ±0.47 / 23.20 ms │ 20.98 / 21.26 ±0.16 / 21.46 ms │ no change │
│ QQuery 3 │ 39.33 / 43.06 ±2.19 / 46.23 ms │ 38.00 / 39.19 ±1.17 / 40.70 ms │ +1.10x faster │
│ QQuery 4 │ 19.18 / 19.43 ±0.14 / 19.59 ms │ 18.65 / 19.08 ±0.62 / 20.30 ms │ no change │
│ QQuery 5 │ 48.30 / 50.23 ±1.65 / 52.60 ms │ 49.54 / 50.89 ±1.39 / 53.45 ms │ no change │
│ QQuery 6 │ 17.67 / 17.74 ±0.07 / 17.85 ms │ 17.64 / 17.83 ±0.15 / 18.08 ms │ no change │
│ QQuery 7 │ 55.53 / 56.45 ±0.75 / 57.57 ms │ 54.68 / 56.55 ±1.58 / 58.30 ms │ no change │
│ QQuery 8 │ 48.77 / 48.97 ±0.18 / 49.24 ms │ 48.75 / 48.86 ±0.08 / 49.01 ms │ no change │
│ QQuery 9 │ 54.11 / 55.41 ±0.91 / 56.61 ms │ 54.78 / 56.06 ±1.70 / 59.40 ms │ no change │
│ QQuery 10 │ 66.35 / 67.23 ±1.43 / 70.07 ms │ 67.59 / 69.20 ±1.44 / 71.78 ms │ no change │
│ QQuery 11 │ 14.35 / 14.72 ±0.47 / 15.60 ms │ 14.69 / 15.15 ±0.30 / 15.50 ms │ no change │
│ QQuery 12 │ 27.56 / 28.25 ±0.63 / 29.05 ms │ 28.14 / 28.76 ±0.80 / 30.30 ms │ no change │
│ QQuery 13 │ 38.35 / 38.83 ±0.29 / 39.13 ms │ 39.07 / 39.66 ±0.37 / 40.25 ms │ no change │
│ QQuery 14 │ 28.33 / 28.63 ±0.22 / 28.99 ms │ 28.93 / 29.22 ±0.23 / 29.56 ms │ no change │
│ QQuery 15 │ 34.09 / 34.42 ±0.31 / 35.01 ms │ 34.76 / 35.24 ±0.37 / 35.86 ms │ no change │
│ QQuery 16 │ 15.70 / 15.80 ±0.07 / 15.90 ms │ 16.00 / 16.27 ±0.38 / 17.03 ms │ no change │
│ QQuery 17 │ 79.29 / 80.63 ±0.82 / 81.84 ms │ 75.69 / 77.38 ±0.90 / 78.18 ms │ no change │
│ QQuery 18 │ 76.53 / 77.79 ±0.95 / 79.12 ms │ 78.73 / 79.49 ±0.46 / 80.05 ms │ no change │
│ QQuery 19 │ 38.12 / 38.39 ±0.16 / 38.58 ms │ 38.75 / 39.13 ±0.24 / 39.35 ms │ no change │
│ QQuery 20 │ 40.28 / 41.62 ±2.55 / 46.73 ms │ 40.75 / 42.65 ±3.22 / 49.05 ms │ no change │
│ QQuery 21 │ 64.08 / 65.63 ±1.07 / 66.75 ms │ 66.91 / 67.85 ±0.80 / 69.22 ms │ no change │
│ QQuery 22 │ 24.47 / 24.68 ±0.14 / 24.82 ms │ 24.97 / 25.24 ±0.19 / 25.47 ms │ no change │
└───────────┴────────────────────────────────┴─────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━┩
│ Total Time (HEAD) │ 912.79ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 916.39ms │
│ Average Time (HEAD) │ 41.49ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 41.65ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 21 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴──────────┘

Resource Usage

tpch — base (merge-base)

Metric	Value
Wall time	5.0s
Peak memory	5.5 GiB
Avg memory	4.9 GiB
CPU user	34.2s
CPU sys	2.5s
Peak spill	0 B

tpch — branch

Metric	Value
Wall time	5.0s
Peak memory	5.5 GiB
Avg memory	4.8 GiB
CPU user	34.3s
CPU sys	2.6s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangb]

run benchmark tpch

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_TARGET_PARTITIONS=128DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: falseDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: falsechanged:
ref: HEADenv:
DATAFUSION_EXECUTION_TARGET_PARTITIONS=128DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: falseDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: false

[adriangbot]

Hi @adriangb, your benchmark configuration could not be parsed (#21931 (comment)).

Error:invalid configuration: baseline.env: invalid type: string "DATAFUSION_EXECUTION_TARGET_PARTITIONS=128 DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS", expected a map at line 5 column 8

Supported benchmarks:

• Standard: clickbench_1, clickbench_extended, clickbench_partitioned, clickbench_pushdown, external_aggr, smj, sort_pushdown, sort_pushdown_inexact, sort_pushdown_inexact_overlap, sort_pushdown_inexact_unsorted, sort_pushdown_sorted, topk_tpch, tpcds, tpch, tpch10, tpch_mem, tpch_mem10
• Criterion: (any)

Usage:

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

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

env:
SHARED_SETTING: enabledbaseline:
ref: v45.0.0env:
DATAFUSION_RUNTIME_MEMORY_LIMIT: 1Gchanged:
ref: v46.0.0env:
DATAFUSION_RUNTIME_MEMORY_LIMIT: 2G

---

File an issue against this benchmark runner

[adriangb]

run benchmark tpch

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_TARGET_PARTITIONS=128DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: trueDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: truechanged:
ref: HEADenv:
DATAFUSION_EXECUTION_TARGET_PARTITIONS=128DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: trueDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: true

[adriangbot]

Hi @adriangb, your benchmark configuration could not be parsed (#21931 (comment)).

Error:invalid configuration: baseline.env: invalid type: string "DATAFUSION_EXECUTION_TARGET_PARTITIONS=128 DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS", expected a map at line 5 column 8

Supported benchmarks:

• Standard: clickbench_1, clickbench_extended, clickbench_partitioned, clickbench_pushdown, external_aggr, smj, sort_pushdown, sort_pushdown_inexact, sort_pushdown_inexact_overlap, sort_pushdown_inexact_unsorted, sort_pushdown_sorted, topk_tpch, tpcds, tpch, tpch10, tpch_mem, tpch_mem10
• Criterion: (any)

Usage:

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

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

env:
SHARED_SETTING: enabledbaseline:
ref: v45.0.0env:
DATAFUSION_RUNTIME_MEMORY_LIMIT: 1Gchanged:
ref: v46.0.0env:
DATAFUSION_RUNTIME_MEMORY_LIMIT: 2G

---

File an issue against this benchmark runner

[adriangb]

run benchmark tpch

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_TARGET_PARTITIONS: 128DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: falseDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: falsechanged:
ref: HEADenv:
DATAFUSION_EXECUTION_TARGET_PARTITIONS: 128DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: falseDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: false

[adriangb]

run benchmark tpch

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_TARGET_PARTITIONS: 128DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: trueDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: truechanged:
ref: HEADenv:
DATAFUSION_EXECUTION_TARGET_PARTITIONS: 128DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: trueDATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: true

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4345772857-1921-hxt5c 6.12.55+ #1 SMP Sun Feb 1 08:59:41 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing HEAD (18129fe) to main diff using: tpch
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4345770652-1920-8dwhv 6.12.55+ #1 SMP Sun Feb 1 08:59:41 UTC 2026 aarch64 GNU/Linux

CPU Details (lscpu)

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

Comparing HEAD (18129fe) to main diff using: tpch
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 50.39 / 51.37 ±0.72 / 52.18 ms │ 50.98 / 51.49 ±0.41 / 51.93 ms │ no change │
│ QQuery 2 │ 41.24 / 42.35 ±1.13 / 44.39 ms │ 42.22 / 43.18 ±1.52 / 46.19 ms │ no change │
│ QQuery 3 │ 121.16 / 137.14 ±10.30 / 149.26 ms │ 80.57 / 81.85 ±1.46 / 84.12 ms │ +1.68x faster │
│ QQuery 4 │ 35.18 / 37.11 ±1.11 / 38.15 ms │ 35.39 / 36.62 ±1.11 / 38.35 ms │ no change │
│ QQuery 5 │ 202.96 / 225.56 ±12.70 / 237.39 ms │ 121.17 / 124.14 ±2.37 / 127.42 ms │ +1.82x faster │
│ QQuery 6 │ 21.61 / 22.62 ±0.68 / 23.24 ms │ 22.16 / 22.79 ±0.61 / 23.75 ms │ no change │
│ QQuery 7 │ 209.79 / 218.28 ±4.51 / 222.71 ms │ 120.22 / 124.67 ±2.41 / 127.04 ms │ +1.75x faster │
│ QQuery 8 │ 207.20 / 212.46 ±3.23 / 216.63 ms │ 95.98 / 98.63 ±1.75 / 100.98 ms │ +2.15x faster │
│ QQuery 9 │ 196.07 / 198.42 ±2.40 / 202.39 ms │ 118.49 / 121.18 ±2.19 / 124.98 ms │ +1.64x faster │
│ QQuery 10 │ 83.06 / 84.09 ±0.83 / 85.33 ms │ 82.62 / 84.10 ±1.05 / 85.83 ms │ no change │
│ QQuery 11 │ 40.02 / 41.99 ±1.88 / 44.95 ms │ 40.26 / 41.90 ±1.06 / 43.37 ms │ no change │
│ QQuery 12 │ 113.38 / 118.16 ±4.21 / 123.64 ms │ 62.57 / 65.21 ±1.62 / 67.20 ms │ +1.81x faster │
│ QQuery 13 │ 105.37 / 111.78 ±4.20 / 118.07 ms │ 61.98 / 63.48 ±1.16 / 64.82 ms │ +1.76x faster │
│ QQuery 14 │ 44.86 / 45.82 ±0.69 / 46.79 ms │ 46.48 / 47.54 ±1.38 / 49.97 ms │ no change │
│ QQuery 15 │ 60.09 / 62.55 ±1.75 / 64.45 ms │ 62.15 / 63.99 ±1.88 / 67.34 ms │ no change │
│ QQuery 16 │ 41.66 / 43.44 ±1.00 / 44.29 ms │ 41.97 / 44.42 ±1.39 / 45.76 ms │ no change │
│ QQuery 17 │ 201.30 / 202.76 ±1.35 / 204.66 ms │ 125.61 / 127.81 ±2.07 / 131.26 ms │ +1.59x faster │
│ QQuery 18 │ 287.19 / 295.98 ±5.08 / 301.13 ms │ 226.72 / 230.02 ±2.75 / 234.29 ms │ +1.29x faster │
│ QQuery 19 │ 43.90 / 45.02 ±0.61 / 45.65 ms │ 44.09 / 44.77 ±0.51 / 45.66 ms │ no change │
│ QQuery 20 │ 127.40 / 139.73 ±6.29 / 145.14 ms │ 65.72 / 68.09 ±1.50 / 69.87 ms │ +2.05x faster │
│ QQuery 21 │ 181.94 / 191.00 ±8.25 / 205.46 ms │ 130.00 / 133.81 ±3.16 / 139.60 ms │ +1.43x faster │
│ QQuery 22 │ 37.94 / 38.71 ±0.62 / 39.46 ms │ 34.81 / 37.13 ±1.58 / 39.13 ms │ no change │
└───────────┴────────────────────────────────────┴─────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 2566.31ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 1756.80ms │
│ Average Time (HEAD) │ 116.65ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 79.85ms │
│ Queries Faster │ 11 │
│ Queries Slower │ 0 │
│ Queries with No Change │ 11 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴───────────┘

Resource Usage

tpch — base (merge-base)

Metric	Value
Wall time	15.0s
Peak memory	6.4 GiB
Avg memory	5.5 GiB
CPU user	107.0s
CPU sys	3.7s
Peak spill	0 B

tpch — branch

Metric	Value
Wall time	10.0s
Peak memory	6.5 GiB
Avg memory	5.6 GiB
CPU user	58.1s
CPU sys	3.9s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 52.05 / 53.47 ±1.27 / 55.54 ms │ 52.46 / 53.78 ±1.30 / 55.95 ms │ no change │
│ QQuery 2 │ 43.53 / 45.21 ±1.02 / 46.38 ms │ 44.19 / 47.68 ±1.96 / 49.75 ms │ 1.05x slower │
│ QQuery 3 │ 150.12 / 158.75 ±5.93 / 167.47 ms │ 162.14 / 166.32 ±3.24 / 169.49 ms │ no change │
│ QQuery 4 │ 37.45 / 38.42 ±0.65 / 39.41 ms │ 37.47 / 39.01 ±1.57 / 41.51 ms │ no change │
│ QQuery 5 │ 250.10 / 260.92 ±8.26 / 270.85 ms │ 482.17 / 490.53 ±6.89 / 498.80 ms │ 1.88x slower │
│ QQuery 6 │ 37.75 / 38.66 ±1.18 / 40.91 ms │ 37.37 / 38.63 ±0.86 / 39.76 ms │ no change │
│ QQuery 7 │ 355.41 / 367.58 ±9.23 / 383.28 ms │ 299.28 / 302.02 ±2.32 / 306.23 ms │ +1.22x faster │
│ QQuery 8 │ 275.92 / 292.06 ±8.97 / 301.66 ms │ 397.50 / 400.91 ±2.51 / 404.69 ms │ 1.37x slower │
│ QQuery 9 │ 307.64 / 313.54 ±3.44 / 317.16 ms │ 503.08 / 513.16 ±8.07 / 526.26 ms │ 1.64x slower │
│ QQuery 10 │ 90.62 / 92.01 ±1.37 / 94.33 ms │ 92.26 / 93.08 ±0.94 / 94.78 ms │ no change │
│ QQuery 11 │ 38.05 / 39.69 ±1.26 / 41.91 ms │ 41.18 / 42.07 ±0.71 / 43.00 ms │ 1.06x slower │
│ QQuery 12 │ 136.20 / 148.87 ±7.23 / 156.97 ms │ 156.43 / 160.31 ±2.80 / 164.44 ms │ 1.08x slower │
│ QQuery 13 │ 118.32 / 125.06 ±8.53 / 140.29 ms │ 138.64 / 142.22 ±1.91 / 143.88 ms │ 1.14x slower │
│ QQuery 14 │ 79.33 / 80.29 ±1.00 / 82.11 ms │ 118.86 / 120.90 ±1.58 / 122.93 ms │ 1.51x slower │
│ QQuery 15 │ 70.79 / 71.76 ±0.78 / 72.92 ms │ 68.56 / 72.13 ±2.14 / 74.64 ms │ no change │
│ QQuery 16 │ 52.72 / 54.60 ±1.38 / 56.70 ms │ 51.29 / 54.30 ±1.66 / 56.01 ms │ no change │
│ QQuery 17 │ 289.38 / 294.78 ±5.76 / 302.41 ms │ 155.59 / 157.78 ±1.75 / 160.34 ms │ +1.87x faster │
│ QQuery 18 │ 228.93 / 243.10 ±8.59 / 255.82 ms │ 171.78 / 175.22 ±2.07 / 178.18 ms │ +1.39x faster │
│ QQuery 19 │ 44.33 / 45.59 ±0.80 / 46.55 ms │ 44.85 / 45.29 ±0.41 / 45.81 ms │ no change │
│ QQuery 20 │ 151.21 / 157.23 ±5.36 / 166.85 ms │ 87.82 / 89.43 ±1.11 / 90.61 ms │ +1.76x faster │
│ QQuery 21 │ 217.93 / 233.27 ±9.60 / 247.64 ms │ 375.11 / 377.42 ±1.62 / 379.55 ms │ 1.62x slower │
│ QQuery 22 │ 52.66 / 54.19 ±1.64 / 57.34 ms │ 51.94 / 53.94 ±1.32 / 56.07 ms │ no change │
└───────────┴───────────────────────────────────┴─────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 3209.06ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 3636.15ms │
│ Average Time (HEAD) │ 145.87ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 165.28ms │
│ Queries Faster │ 4 │
│ Queries Slower │ 9 │
│ Queries with No Change │ 9 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴───────────┘

Resource Usage

tpch — base (merge-base)

Metric	Value
Wall time	20.0s
Peak memory	6.2 GiB
Avg memory	5.4 GiB
CPU user	140.0s
CPU sys	3.4s
Peak spill	0 B

tpch — branch

Metric	Value
Wall time	20.0s
Peak memory	5.9 GiB
Avg memory	5.3 GiB
CPU user	149.8s
CPU sys	3.7s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangb]

run benchmarks clickbench_partitioned tpch tpch10 tpcds hj

env:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: "false"DATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: "false"baseline:
ref: mainchanged:
ref: HEAD

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4471353144-175-5nqtd 6.12.68+ #1 SMP Wed Apr 1 02:23:28 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 HEAD (42adf29) to main diff using: clickbench_partitioned
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 6.30 / 6.79 ±0.90 / 8.60 ms │ 6.35 / 6.87 ±0.82 / 8.50 ms │ no change │
│ QQuery 2 │ 81.92 / 82.35 ±0.34 / 82.86 ms │ 80.19 / 81.09 ±0.59 / 81.74 ms │ no change │
│ QQuery 3 │ 28.86 / 29.30 ±0.41 / 29.82 ms │ 28.48 / 29.16 ±0.40 / 29.73 ms │ no change │
│ QQuery 4 │ 509.00 / 513.11 ±3.32 / 517.26 ms │ 516.51 / 521.28 ±3.39 / 526.97 ms │ no change │
│ QQuery 5 │ 52.71 / 53.27 ±0.51 / 54.11 ms │ 52.53 / 52.93 ±0.42 / 53.55 ms │ no change │
│ QQuery 6 │ 35.64 / 36.13 ±0.44 / 36.84 ms │ 35.27 / 35.52 ±0.18 / 35.79 ms │ no change │
│ QQuery 7 │ 105.71 / 106.79 ±1.35 / 109.41 ms │ 109.52 / 110.93 ±2.02 / 114.87 ms │ no change │
│ QQuery 8 │ 38.82 / 39.46 ±0.57 / 40.46 ms │ 38.52 / 39.02 ±0.40 / 39.53 ms │ no change │
│ QQuery 9 │ 52.60 / 55.38 ±2.80 / 59.89 ms │ 53.54 / 54.76 ±1.12 / 56.23 ms │ no change │
│ QQuery 10 │ 80.63 / 81.08 ±0.23 / 81.26 ms │ 79.98 / 81.17 ±1.73 / 84.61 ms │ no change │
│ QQuery 11 │ 310.38 / 315.15 ±4.13 / 322.16 ms │ 319.32 / 324.37 ±4.62 / 331.74 ms │ no change │
│ QQuery 12 │ 28.35 / 28.84 ±0.39 / 29.54 ms │ 28.36 / 29.07 ±0.49 / 29.77 ms │ no change │
│ QQuery 13 │ 128.18 / 129.17 ±1.14 / 131.33 ms │ 128.33 / 129.57 ±1.22 / 131.72 ms │ no change │
│ QQuery 14 │ 505.04 / 507.33 ±2.15 / 510.06 ms │ 509.09 / 512.61 ±3.29 / 517.01 ms │ no change │
│ QQuery 15 │ 60.13 / 61.36 ±0.87 / 62.60 ms │ 59.95 / 61.07 ±0.73 / 62.04 ms │ no change │
│ QQuery 16 │ 6.89 / 7.05 ±0.12 / 7.22 ms │ 6.98 / 7.12 ±0.23 / 7.59 ms │ no change │
│ QQuery 17 │ 80.63 / 82.41 ±1.50 / 84.69 ms │ 82.25 / 82.71 ±0.44 / 83.47 ms │ no change │
│ QQuery 18 │ 150.26 / 150.96 ±0.71 / 152.16 ms │ 155.00 / 155.78 ±0.96 / 157.62 ms │ no change │
│ QQuery 19 │ 41.27 / 41.62 ±0.30 / 42.02 ms │ 41.50 / 41.79 ±0.24 / 42.16 ms │ no change │
│ QQuery 20 │ 34.90 / 35.31 ±0.35 / 35.92 ms │ 34.91 / 35.58 ±0.60 / 36.52 ms │ no change │
│ QQuery 21 │ 17.77 / 17.99 ±0.15 / 18.23 ms │ 17.89 / 18.16 ±0.22 / 18.54 ms │ no change │
│ QQuery 22 │ 62.20 / 62.72 ±0.36 / 63.22 ms │ 61.91 / 62.25 ±0.23 / 62.55 ms │ no change │
│ QQuery 23 │ 473.40 / 476.52 ±2.83 / 480.64 ms │ 478.74 / 482.66 ±3.82 / 489.29 ms │ no change │
│ QQuery 24 │ 232.58 / 236.64 ±4.91 / 246.30 ms │ 239.44 / 242.55 ±2.55 / 246.69 ms │ no change │
│ QQuery 25 │ 112.58 / 113.96 ±0.94 / 115.41 ms │ 115.99 / 116.81 ±1.23 / 119.25 ms │ no change │
│ QQuery 26 │ 70.13 / 70.24 ±0.14 / 70.51 ms │ 70.76 / 71.20 ±0.45 / 72.02 ms │ no change │
│ QQuery 27 │ 7.00 / 7.12 ±0.15 / 7.42 ms │ 7.12 / 7.23 ±0.15 / 7.54 ms │ no change │
│ QQuery 28 │ 59.04 / 63.22 ±2.35 / 66.26 ms │ 58.46 / 61.76 ±1.81 / 63.50 ms │ no change │
│ QQuery 29 │ 99.20 / 102.49 ±2.22 / 106.08 ms │ 99.78 / 101.49 ±2.30 / 105.98 ms │ no change │
│ QQuery 30 │ 30.37 / 30.72 ±0.47 / 31.66 ms │ 30.39 / 30.74 ±0.32 / 31.19 ms │ no change │
│ QQuery 31 │ 112.20 / 113.88 ±2.46 / 118.73 ms │ 111.55 / 113.03 ±0.86 / 113.92 ms │ no change │
│ QQuery 32 │ 20.32 / 20.79 ±0.29 / 21.15 ms │ 20.83 / 21.12 ±0.23 / 21.52 ms │ no change │
│ QQuery 33 │ 39.38 / 39.57 ±0.22 / 39.97 ms │ 40.24 / 40.63 ±0.50 / 41.61 ms │ no change │
│ QQuery 34 │ 9.77 / 10.28 ±0.35 / 10.75 ms │ 10.41 / 10.78 ±0.33 / 11.18 ms │ no change │
│ QQuery 35 │ 80.74 / 81.35 ±0.60 / 82.17 ms │ 81.78 / 83.91 ±1.63 / 85.99 ms │ no change │
│ QQuery 36 │ 6.43 / 6.56 ±0.18 / 6.90 ms │ 6.57 / 6.71 ±0.19 / 7.09 ms │ no change │
│ QQuery 37 │ 7.05 / 7.28 ±0.16 / 7.54 ms │ 7.33 / 7.61 ±0.27 / 8.00 ms │ no change │
│ QQuery 38 │ 68.13 / 69.32 ±0.80 / 70.22 ms │ 69.09 / 69.57 ±0.50 / 70.54 ms │ no change │
│ QQuery 39 │ 97.12 / 98.38 ±0.80 / 99.32 ms │ 99.46 / 100.44 ±0.98 / 102.04 ms │ no change │
│ QQuery 40 │ 22.59 / 22.97 ±0.38 / 23.62 ms │ 24.03 / 24.16 ±0.16 / 24.48 ms │ 1.05x slower │
│ QQuery 41 │ 13.94 / 14.12 ±0.21 / 14.50 ms │ 14.05 / 14.27 ±0.18 / 14.60 ms │ no change │
│ QQuery 42 │ 24.27 / 24.60 ±0.40 / 25.36 ms │ 24.33 / 24.67 ±0.26 / 25.00 ms │ no change │
│ QQuery 43 │ 5.34 / 5.43 ±0.14 / 5.71 ms │ 5.47 / 5.57 ±0.12 / 5.81 ms │ no change │
│ QQuery 44 │ 10.91 / 11.07 ±0.11 / 11.19 ms │ 11.15 / 11.29 ±0.10 / 11.43 ms │ no change │
│ QQuery 45 │ 39.70 / 40.42 ±0.42 / 40.89 ms │ 40.64 / 41.21 ±0.31 / 41.46 ms │ no change │
│ QQuery 46 │ 13.16 / 13.40 ±0.21 / 13.78 ms │ 13.70 / 13.94 ±0.13 / 14.10 ms │ no change │
│ QQuery 47 │ 228.07 / 229.99 ±1.34 / 231.96 ms │ 233.53 / 237.12 ±3.70 / 244.19 ms │ no change │
│ QQuery 48 │ 103.78 / 104.31 ±0.59 / 105.33 ms │ 103.79 / 106.15 ±1.84 / 108.46 ms │ no change │
│ QQuery 49 │ 80.93 / 81.24 ±0.31 / 81.81 ms │ 82.00 / 82.78 ±0.42 / 83.30 ms │ no change │
│ QQuery 50 │ 60.03 / 60.64 ±0.56 / 61.50 ms │ 61.61 / 64.86 ±3.42 / 70.08 ms │ 1.07x slower │
│ QQuery 51 │ 91.91 / 94.46 ±2.31 / 98.05 ms │ 94.14 / 94.84 ±0.46 / 95.33 ms │ no change │
│ QQuery 52 │ 24.14 / 24.64 ±0.37 / 25.12 ms │ 24.84 / 25.20 ±0.49 / 26.17 ms │ no change │
│ QQuery 53 │ 30.44 / 30.87 ±0.38 / 31.55 ms │ 31.14 / 32.32 ±1.59 / 35.43 ms │ no change │
│ QQuery 54 │ 53.96 / 54.70 ±0.47 / 55.32 ms │ 54.03 / 55.80 ±1.87 / 59.38 ms │ no change │
│ QQuery 55 │ 23.87 / 25.00 ±1.89 / 28.75 ms │ 23.84 / 24.16 ±0.26 / 24.61 ms │ no change │
│ QQuery 56 │ 39.92 / 40.08 ±0.15 / 40.36 ms │ 39.84 / 40.16 ±0.23 / 40.47 ms │ no change │
│ QQuery 57 │ 175.91 / 177.86 ±1.53 / 180.07 ms │ 179.18 / 181.60 ±2.01 / 185.03 ms │ no change │
│ QQuery 58 │ 115.95 / 116.40 ±0.35 / 116.87 ms │ 115.92 / 116.32 ±0.23 / 116.53 ms │ no change │
│ QQuery 59 │ 118.44 / 119.57 ±1.00 / 121.43 ms │ 117.86 / 118.54 ±0.71 / 119.42 ms │ no change │
│ QQuery 60 │ 40.31 / 40.72 ±0.23 / 40.95 ms │ 40.42 / 40.56 ±0.25 / 41.06 ms │ no change │
│ QQuery 61 │ 14.08 / 14.18 ±0.11 / 14.37 ms │ 13.86 / 13.98 ±0.17 / 14.31 ms │ no change │
│ QQuery 62 │ 46.43 / 47.25 ±0.98 / 49.03 ms │ 47.21 / 48.23 ±0.91 / 49.54 ms │ no change │
│ QQuery 63 │ 30.50 / 31.06 ±0.71 / 32.41 ms │ 30.55 / 30.79 ±0.19 / 30.99 ms │ no change │
│ QQuery 64 │ 452.89 / 456.71 ±4.36 / 462.42 ms │ 467.57 / 473.83 ±4.94 / 481.12 ms │ no change │
│ QQuery 65 │ 143.68 / 145.94 ±2.73 / 151.29 ms │ 69.05 / 70.35 ±1.24 / 72.68 ms │ +2.07x faster │
│ QQuery 66 │ 83.39 / 85.64 ±3.67 / 92.96 ms │ 83.70 / 85.69 ±2.71 / 91.02 ms │ no change │
│ QQuery 67 │ 242.37 / 246.31 ±2.83 / 250.82 ms │ 241.24 / 246.82 ±3.64 / 250.64 ms │ no change │
│ QQuery 68 │ 13.94 / 14.04 ±0.13 / 14.30 ms │ 13.65 / 13.92 ±0.28 / 14.45 ms │ no change │
│ QQuery 69 │ 77.12 / 78.55 ±2.44 / 83.41 ms │ 76.18 / 76.40 ±0.34 / 77.08 ms │ no change │
│ QQuery 70 │ 104.34 / 111.07 ±8.99 / 128.31 ms │ 106.65 / 110.51 ±4.46 / 118.92 ms │ no change │
│ QQuery 71 │ 35.93 / 36.18 ±0.17 / 36.45 ms │ 35.86 / 36.39 ±0.56 / 37.08 ms │ no change │
│ QQuery 72 │ 2204.00 / 2278.47 ±69.05 / 2402.16 ms │ 2178.87 / 2244.60 ±55.46 / 2347.51 ms │ no change │
│ QQuery 73 │ 9.57 / 9.82 ±0.27 / 10.25 ms │ 9.49 / 9.71 ±0.28 / 10.25 ms │ no change │
│ QQuery 74 │ 177.67 / 180.81 ±2.35 / 184.72 ms │ 176.67 / 180.18 ±2.45 / 182.98 ms │ no change │
│ QQuery 75 │ 148.45 / 149.05 ±0.83 / 150.68 ms │ 149.21 / 149.92 ±0.56 / 150.83 ms │ no change │
│ QQuery 76 │ 35.70 / 36.95 ±2.08 / 41.11 ms │ 36.04 / 36.16 ±0.09 / 36.27 ms │ no change │
│ QQuery 77 │ 62.60 / 64.73 ±3.72 / 72.16 ms │ 61.51 / 62.91 ±1.90 / 66.66 ms │ no change │
│ QQuery 78 │ 189.09 / 189.56 ±0.42 / 190.16 ms │ 194.46 / 198.18 ±2.52 / 200.81 ms │ no change │
│ QQuery 79 │ 67.33 / 68.67 ±1.29 / 70.84 ms │ 67.17 / 67.88 ±0.59 / 68.97 ms │ no change │
│ QQuery 80 │ 101.78 / 105.34 ±3.99 / 112.45 ms │ 104.15 / 105.24 ±0.81 / 106.57 ms │ no change │
│ QQuery 81 │ 24.70 / 24.95 ±0.13 / 25.06 ms │ 24.25 / 25.77 ±2.62 / 31.00 ms │ no change │
│ QQuery 82 │ 16.85 / 16.89 ±0.05 / 16.97 ms │ 16.44 / 17.05 ±0.37 / 17.46 ms │ no change │
│ QQuery 83 │ 37.29 / 37.74 ±0.36 / 38.39 ms │ 36.21 / 36.75 ±0.32 / 37.15 ms │ no change │
│ QQuery 84 │ 43.27 / 43.77 ±0.31 / 44.14 ms │ 42.90 / 43.41 ±0.47 / 44.18 ms │ no change │
│ QQuery 85 │ 133.02 / 134.86 ±2.48 / 139.71 ms │ 134.18 / 136.85 ±2.35 / 140.62 ms │ no change │
│ QQuery 86 │ 25.03 / 25.36 ±0.30 / 25.86 ms │ 25.36 / 25.59 ±0.35 / 26.29 ms │ no change │
│ QQuery 87 │ 68.78 / 69.84 ±0.89 / 71.04 ms │ 68.61 / 69.30 ±0.36 / 69.60 ms │ no change │
│ QQuery 88 │ 64.78 / 65.18 ±0.41 / 65.92 ms │ 64.85 / 65.52 ±0.52 / 66.16 ms │ no change │
│ QQuery 89 │ 36.29 / 36.71 ±0.34 / 37.09 ms │ 36.32 / 37.01 ±0.51 / 37.61 ms │ no change │
│ QQuery 90 │ 17.40 / 17.71 ±0.18 / 17.92 ms │ 17.39 / 17.65 ±0.20 / 17.96 ms │ no change │
│ QQuery 91 │ 51.84 / 52.30 ±0.32 / 52.73 ms │ 51.89 / 52.15 ±0.19 / 52.42 ms │ no change │
│ QQuery 92 │ 29.82 / 30.63 ±0.96 / 32.50 ms │ 28.80 / 29.11 ±0.20 / 29.39 ms │ no change │
│ QQuery 93 │ 51.19 / 52.51 ±1.35 / 55.07 ms │ 52.01 / 52.71 ±0.52 / 53.50 ms │ no change │
│ QQuery 94 │ 38.03 / 38.61 ±0.70 / 39.87 ms │ 37.65 / 38.55 ±1.28 / 41.02 ms │ no change │
│ QQuery 95 │ 87.62 / 89.41 ±1.39 / 91.25 ms │ 88.36 / 88.93 ±0.43 / 89.67 ms │ no change │
│ QQuery 96 │ 24.54 / 24.67 ±0.17 / 25.01 ms │ 24.31 / 24.53 ±0.23 / 24.91 ms │ no change │
│ QQuery 97 │ 46.86 / 46.93 ±0.10 / 47.11 ms │ 46.79 / 47.36 ±0.44 / 48.10 ms │ no change │
│ QQuery 98 │ 42.06 / 42.52 ±0.26 / 42.75 ms │ 42.15 / 43.14 ±0.84 / 44.67 ms │ no change │
│ QQuery 99 │ 69.96 / 70.19 ±0.17 / 70.39 ms │ 70.78 / 71.10 ±0.28 / 71.57 ms │ no change │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 10595.50ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 10580.74ms │
│ Average Time (HEAD) │ 107.03ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 106.88ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 2 │
│ Queries with No Change │ 96 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴────────────┘

Resource Usage

tpcds — base (merge-base)

Metric	Value
Wall time	55.0s
Peak memory	6.9 GiB
Avg memory	6.2 GiB
CPU user	244.8s
CPU sys	6.1s
Peak spill	0 B

tpcds — branch

Metric	Value
Wall time	55.0s
Peak memory	6.8 GiB
Avg memory	6.1 GiB
CPU user	241.2s
CPU sys	6.5s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4471353144-176-lmx9b 6.12.68+ #1 SMP Wed Apr 1 02:23:28 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 HEAD (42adf29) to main diff using: tpch
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4471353144-177-5bflm 6.12.68+ #1 SMP Wed Apr 1 02:23:28 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 HEAD (42adf29) to main diff using: tpch10
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4471353144-178-62lgl 6.12.68+ #1 SMP Wed Apr 1 02:23:28 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 HEAD (42adf29) to main diff using: tpcds
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Criterion benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4471353144-179-kkrvn 6.12.68+ #1 SMP Wed Apr 1 02:23:28 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 HEAD (42adf29) to main diff
BENCH_NAME=hj
BENCH_COMMAND=cargo bench --features=parquet --bench hj
BENCH_FILTER=
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

Benchmark for this request failed.

Last 20 lines of output:

Click to expand

 * [new ref] refs/pull/21931/head -> worktree-dynamic-filter-restructure
* branch main -> FETCH_HEAD
Switched to branch 'worktree-dynamic-filter-restructure'
6a57b220842f342e3765912f5e64c299f9756529
From https://github.com/apache/datafusion
* branch refs/pull/21931/head -> FETCH_HEAD
Cloning into '/workspace/datafusion-base'...
From https://github.com/apache/datafusion
* branch refs/pull/21931/head -> FETCH_HEAD
Your branch is up to date with 'origin/main'.
Already on 'main'
rustc 1.95.0 (59807616e 2026-04-14)
42adf29b6bd78b10156cec3bc8838570fe666aed
47655fd6c9ef060d73497987e6ccb98e57196508
Blocking waiting for file lock on package cache
Blocking waiting for file lock on package cache
Blocking waiting for file lock on package cache
error: no bench target named `hj` in default-run packages
help: a target with a similar name exists: `chr`

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.23 / 4.75 ±6.89 / 18.53 ms │ 1.26 / 4.86 ±7.03 / 18.92 ms │ no change │
│ QQuery 1 │ 12.79 / 13.25 ±0.23 / 13.43 ms │ 12.63 / 13.15 ±0.34 / 13.54 ms │ no change │
│ QQuery 2 │ 36.14 / 36.45 ±0.30 / 36.91 ms │ 36.63 / 36.84 ±0.32 / 37.48 ms │ no change │
│ QQuery 3 │ 30.89 / 31.54 ±0.84 / 33.14 ms │ 31.34 / 31.60 ±0.29 / 32.15 ms │ no change │
│ QQuery 4 │ 232.90 / 236.75 ±2.79 / 240.43 ms │ 241.99 / 246.97 ±4.20 / 252.04 ms │ no change │
│ QQuery 5 │ 274.50 / 276.21 ±2.00 / 280.07 ms │ 286.37 / 287.71 ±1.41 / 290.13 ms │ no change │
│ QQuery 6 │ 7.00 / 7.22 ±0.22 / 7.64 ms │ 6.92 / 8.90 ±2.00 / 12.72 ms │ 1.23x slower │
│ QQuery 7 │ 14.13 / 14.44 ±0.48 / 15.39 ms │ 13.85 / 14.25 ±0.36 / 14.89 ms │ no change │
│ QQuery 8 │ 324.78 / 327.01 ±1.64 / 329.29 ms │ 321.14 / 324.63 ±2.44 / 327.76 ms │ no change │
│ QQuery 9 │ 446.08 / 459.48 ±7.90 / 469.34 ms │ 449.67 / 460.94 ±7.08 / 470.48 ms │ no change │
│ QQuery 10 │ 68.90 / 70.03 ±1.05 / 72.01 ms │ 68.15 / 69.17 ±1.10 / 71.31 ms │ no change │
│ QQuery 11 │ 79.58 / 81.48 ±1.92 / 84.97 ms │ 78.56 / 79.81 ±0.89 / 81.11 ms │ no change │
│ QQuery 12 │ 268.86 / 272.43 ±3.59 / 278.80 ms │ 272.40 / 275.86 ±2.97 / 280.04 ms │ no change │
│ QQuery 13 │ 396.43 / 413.16 ±12.72 / 429.28 ms │ 378.17 / 388.69 ±9.78 / 404.34 ms │ +1.06x faster │
│ QQuery 14 │ 292.56 / 296.43 ±3.81 / 301.53 ms │ 277.53 / 280.69 ±3.68 / 287.27 ms │ +1.06x faster │
│ QQuery 15 │ 279.93 / 285.62 ±4.06 / 290.15 ms │ 275.60 / 281.26 ±4.11 / 287.26 ms │ no change │
│ QQuery 16 │ 611.31 / 622.16 ±8.26 / 636.42 ms │ 593.04 / 615.11 ±15.20 / 633.47 ms │ no change │
│ QQuery 17 │ 618.95 / 623.27 ±3.91 / 629.77 ms │ 622.10 / 627.81 ±3.27 / 631.65 ms │ no change │
│ QQuery 18 │ 1226.25 / 1250.84 ±12.86 / 1262.67 ms │ 1206.24 / 1212.93 ±5.20 / 1219.87 ms │ no change │
│ QQuery 19 │ 28.17 / 36.78 ±15.63 / 68.01 ms │ 28.08 / 31.01 ±4.59 / 40.14 ms │ +1.19x faster │
│ QQuery 20 │ 513.19 / 524.03 ±9.06 / 540.29 ms │ 520.93 / 525.06 ±3.24 / 530.22 ms │ no change │
│ QQuery 21 │ 594.67 / 597.70 ±1.86 / 600.12 ms │ 605.07 / 609.98 ±2.60 / 612.03 ms │ no change │
│ QQuery 22 │ 1051.84 / 1066.70 ±16.17 / 1094.49 ms │ 1054.73 / 1078.13 ±13.20 / 1092.97 ms │ no change │
│ QQuery 23 │ 3166.28 / 3197.69 ±32.92 / 3259.13 ms │ 3158.53 / 3188.59 ±41.15 / 3267.56 ms │ no change │
│ QQuery 24 │ 42.18 / 44.07 ±1.49 / 46.57 ms │ 41.71 / 49.04 ±5.84 / 54.93 ms │ 1.11x slower │
│ QQuery 25 │ 111.21 / 112.40 ±1.00 / 113.99 ms │ 111.51 / 113.40 ±2.00 / 117.00 ms │ no change │
│ QQuery 26 │ 42.30 / 42.91 ±0.97 / 44.82 ms │ 42.25 / 43.77 ±2.09 / 47.76 ms │ no change │
│ QQuery 27 │ 673.87 / 679.26 ±8.39 / 695.92 ms │ 665.75 / 670.11 ±3.38 / 674.03 ms │ no change │
│ QQuery 28 │ 2973.99 / 3002.82 ±24.16 / 3041.98 ms │ 3003.19 / 3033.66 ±16.54 / 3048.59 ms │ no change │
│ QQuery 29 │ 41.95 / 50.71 ±7.47 / 60.40 ms │ 41.52 / 43.69 ±3.63 / 50.92 ms │ +1.16x faster │
│ QQuery 30 │ 298.57 / 305.82 ±6.20 / 315.06 ms │ 299.19 / 303.36 ±2.89 / 306.55 ms │ no change │
│ QQuery 31 │ 294.26 / 302.90 ±6.88 / 313.53 ms │ 291.06 / 295.83 ±3.94 / 302.92 ms │ no change │
│ QQuery 32 │ 951.59 / 962.13 ±8.88 / 973.32 ms │ 943.33 / 958.25 ±16.06 / 984.64 ms │ no change │
│ QQuery 33 │ 1404.57 / 1439.81 ±28.37 / 1489.19 ms │ 1416.36 / 1435.15 ±18.95 / 1467.04 ms │ no change │
│ QQuery 34 │ 1447.85 / 1495.67 ±28.99 / 1531.27 ms │ 1457.73 / 1496.01 ±22.49 / 1514.69 ms │ no change │
│ QQuery 35 │ 280.89 / 291.81 ±13.32 / 317.79 ms │ 285.40 / 312.38 ±33.21 / 375.95 ms │ 1.07x slower │
│ QQuery 36 │ 60.91 / 64.75 ±3.18 / 70.57 ms │ 60.53 / 64.91 ±6.94 / 78.74 ms │ no change │
│ QQuery 37 │ 34.91 / 39.99 ±6.12 / 51.05 ms │ 35.17 / 39.15 ±5.31 / 49.55 ms │ no change │
│ QQuery 38 │ 40.69 / 45.54 ±6.91 / 59.11 ms │ 43.72 / 46.61 ±3.42 / 51.93 ms │ no change │
│ QQuery 39 │ 119.57 / 130.34 ±6.66 / 140.20 ms │ 122.32 / 136.42 ±8.60 / 149.34 ms │ no change │
│ QQuery 40 │ 14.01 / 14.42 ±0.35 / 15.02 ms │ 14.03 / 15.64 ±1.66 / 18.75 ms │ 1.08x slower │
│ QQuery 41 │ 13.77 / 14.94 ±2.14 / 19.22 ms │ 13.60 / 14.01 ±0.26 / 14.28 ms │ +1.07x faster │
│ QQuery 42 │ 13.64 / 19.71 ±7.19 / 28.80 ms │ 12.87 / 14.86 ±3.29 / 21.43 ms │ +1.33x faster │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 19805.41ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 19780.21ms │
│ Average Time (HEAD) │ 460.59ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 460.00ms │
│ Queries Faster │ 6 │
│ Queries Slower │ 4 │
│ Queries with No Change │ 33 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴────────────┘

Resource Usage

clickbench_partitioned — base (merge-base)

Metric	Value
Wall time	100.0s
Peak memory	29.4 GiB
Avg memory	23.0 GiB
CPU user	1043.2s
CPU sys	63.2s
Peak spill	0 B

clickbench_partitioned — branch

Metric	Value
Wall time	100.0s
Peak memory	30.0 GiB
Avg memory	23.1 GiB
CPU user	1041.6s
CPU sys	63.4s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpch_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 38.53 / 40.46 ±2.41 / 44.62 ms │ 38.81 / 39.62 ±1.08 / 41.54 ms │ no change │
│ QQuery 2 │ 23.88 / 24.18 ±0.43 / 25.04 ms │ 24.44 / 24.87 ±0.48 / 25.56 ms │ no change │
│ QQuery 3 │ 46.61 / 47.18 ±0.59 / 48.28 ms │ 51.03 / 51.41 ±0.30 / 51.89 ms │ 1.09x slower │
│ QQuery 4 │ 21.58 / 21.73 ±0.09 / 21.83 ms │ 21.54 / 21.87 ±0.51 / 22.88 ms │ no change │
│ QQuery 5 │ 65.70 / 66.60 ±0.67 / 67.63 ms │ 69.10 / 69.54 ±0.41 / 70.30 ms │ no change │
│ QQuery 6 │ 34.91 / 35.28 ±0.32 / 35.67 ms │ 34.36 / 35.22 ±0.82 / 36.45 ms │ no change │
│ QQuery 7 │ 56.15 / 56.45 ±0.31 / 57.02 ms │ 52.78 / 52.99 ±0.24 / 53.43 ms │ +1.07x faster │
│ QQuery 8 │ 77.75 / 79.18 ±1.18 / 81.27 ms │ 73.40 / 74.85 ±0.92 / 76.26 ms │ +1.06x faster │
│ QQuery 9 │ 102.01 / 103.65 ±2.01 / 107.59 ms │ 113.77 / 118.63 ±4.58 / 124.44 ms │ 1.14x slower │
│ QQuery 10 │ 73.92 / 74.42 ±0.58 / 75.52 ms │ 73.12 / 73.39 ±0.37 / 74.10 ms │ no change │
│ QQuery 11 │ 14.94 / 15.14 ±0.13 / 15.27 ms │ 14.97 / 15.24 ±0.27 / 15.75 ms │ no change │
│ QQuery 12 │ 44.06 / 44.70 ±0.43 / 45.19 ms │ 43.30 / 43.70 ±0.41 / 44.48 ms │ no change │
│ QQuery 13 │ 45.46 / 45.70 ±0.14 / 45.86 ms │ 61.05 / 61.29 ±0.15 / 61.50 ms │ 1.34x slower │
│ QQuery 14 │ 38.48 / 38.71 ±0.13 / 38.85 ms │ 42.91 / 43.32 ±0.33 / 43.86 ms │ 1.12x slower │
│ QQuery 15 │ 40.52 / 41.20 ±0.63 / 42.38 ms │ 40.95 / 41.60 ±0.79 / 43.07 ms │ no change │
│ QQuery 16 │ 22.54 / 22.67 ±0.11 / 22.85 ms │ 22.58 / 22.73 ±0.17 / 23.04 ms │ no change │
│ QQuery 17 │ 158.58 / 161.33 ±3.25 / 167.29 ms │ 53.82 / 54.62 ±0.53 / 55.43 ms │ +2.95x faster │
│ QQuery 18 │ 70.35 / 71.03 ±0.43 / 71.63 ms │ 51.62 / 52.42 ±0.79 / 53.89 ms │ +1.36x faster │
│ QQuery 19 │ 39.85 / 40.51 ±0.52 / 41.09 ms │ 39.52 / 39.74 ±0.19 / 40.07 ms │ no change │
│ QQuery 20 │ 44.26 / 44.83 ±0.80 / 46.39 ms │ 40.67 / 41.47 ±1.24 / 43.94 ms │ +1.08x faster │
│ QQuery 21 │ 79.76 / 80.36 ±0.47 / 81.01 ms │ 82.28 / 83.64 ±0.95 / 84.81 ms │ no change │
│ QQuery 22 │ 34.57 / 34.85 ±0.30 / 35.39 ms │ 34.62 / 35.00 ±0.31 / 35.52 ms │ no change │
└───────────┴───────────────────────────────────┴─────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 1190.17ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 1097.17ms │
│ Average Time (HEAD) │ 54.10ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 49.87ms │
│ Queries Faster │ 5 │
│ Queries Slower │ 4 │
│ Queries with No Change │ 13 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴───────────┘

Resource Usage

tpch — base (merge-base)

Metric	Value
Wall time	10.0s
Peak memory	5.2 GiB
Avg memory	4.5 GiB
CPU user	45.2s
CPU sys	2.0s
Peak spill	0 B

tpch — branch

Metric	Value
Wall time	10.0s
Peak memory	5.2 GiB
Avg memory	4.4 GiB
CPU user	39.7s
CPU sys	2.0s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpch_sf10.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 315.15 / 317.51 ±1.61 / 319.58 ms │ 317.81 / 319.35 ±1.87 / 322.85 ms │ no change │
│ QQuery 2 │ 171.00 / 172.87 ±2.08 / 176.43 ms │ 230.07 / 235.31 ±4.66 / 241.48 ms │ 1.36x slower │
│ QQuery 3 │ 366.16 / 369.55 ±2.77 / 373.99 ms │ 425.61 / 432.73 ±5.36 / 442.22 ms │ 1.17x slower │
│ QQuery 4 │ 314.26 / 316.36 ±1.84 / 319.59 ms │ 393.79 / 396.80 ±2.66 / 399.98 ms │ 1.25x slower │
│ QQuery 5 │ 578.31 / 593.38 ±8.42 / 602.38 ms │ 781.28 / 795.52 ±9.20 / 804.71 ms │ 1.34x slower │
│ QQuery 6 │ 290.23 / 292.81 ±2.14 / 296.70 ms │ 289.70 / 291.97 ±1.89 / 295.31 ms │ no change │
│ QQuery 7 │ 423.99 / 425.85 ±1.76 / 428.45 ms │ 486.88 / 488.01 ±0.82 / 488.96 ms │ 1.15x slower │
│ QQuery 8 │ 610.72 / 614.01 ±3.09 / 618.69 ms │ 686.98 / 690.28 ±3.20 / 694.25 ms │ 1.12x slower │
│ QQuery 9 │ 1337.58 / 1347.25 ±5.52 / 1352.57 ms │ 1700.75 / 1718.82 ±12.85 / 1738.23 ms │ 1.28x slower │
│ QQuery 10 │ 396.58 / 400.48 ±3.06 / 404.15 ms │ 433.52 / 440.01 ±3.46 / 443.02 ms │ 1.10x slower │
│ QQuery 11 │ 136.45 / 139.24 ±4.42 / 147.97 ms │ 139.90 / 142.55 ±3.30 / 148.88 ms │ no change │
│ QQuery 12 │ 335.67 / 339.87 ±4.49 / 347.99 ms │ 388.59 / 393.28 ±5.14 / 399.65 ms │ 1.16x slower │
│ QQuery 13 │ 377.10 / 389.58 ±9.81 / 404.80 ms │ 428.66 / 434.28 ±6.76 / 446.01 ms │ 1.11x slower │
│ QQuery 14 │ 201.09 / 202.49 ±1.53 / 205.20 ms │ 203.44 / 204.70 ±1.12 / 206.76 ms │ no change │
│ QQuery 15 │ 396.60 / 400.91 ±3.03 / 404.03 ms │ 398.21 / 401.38 ±2.06 / 404.67 ms │ no change │
│ QQuery 16 │ 100.97 / 104.96 ±5.54 / 115.78 ms │ 111.94 / 113.32 ±1.33 / 115.61 ms │ 1.08x slower │
│ QQuery 17 │ 755.12 / 761.65 ±4.87 / 768.22 ms │ 662.50 / 667.97 ±5.03 / 677.44 ms │ +1.14x faster │
│ QQuery 18 │ 664.78 / 676.35 ±7.80 / 685.13 ms │ 620.18 / 630.24 ±7.41 / 641.90 ms │ +1.07x faster │
│ QQuery 19 │ 281.54 / 287.59 ±8.32 / 303.99 ms │ 286.33 / 293.19 ±8.12 / 305.21 ms │ no change │
│ QQuery 20 │ 318.05 / 321.19 ±1.64 / 322.82 ms │ 303.56 / 307.98 ±3.83 / 314.78 ms │ no change │
│ QQuery 21 │ 1153.91 / 1162.40 ±6.29 / 1171.19 ms │ 1341.00 / 1357.49 ±12.25 / 1373.93 ms │ 1.17x slower │
│ QQuery 22 │ 128.25 / 132.09 ±2.77 / 136.09 ms │ 121.66 / 127.85 ±4.16 / 132.25 ms │ no change │
└───────────┴──────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 9768.40ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 10883.03ms │
│ Average Time (HEAD) │ 444.02ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 494.68ms │
│ Queries Faster │ 2 │
│ Queries Slower │ 12 │
│ Queries with No Change │ 8 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴────────────┘

Resource Usage

tpch10 — base (merge-base)

Metric	Value
Wall time	50.0s
Peak memory	9.9 GiB
Avg memory	8.1 GiB
CPU user	537.5s
CPU sys	20.2s
Peak spill	0 B

tpch10 — branch

Metric	Value
Wall time	55.0s
Peak memory	9.7 GiB
Avg memory	8.2 GiB
CPU user	598.3s
CPU sys	20.3s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark clickbench_partitioned.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 0 │ 1.19 / 4.51 ±6.59 / 17.70 ms │ 1.18 / 4.58 ±6.73 / 18.04 ms │ no change │
│ QQuery 1 │ 12.55 / 12.67 ±0.09 / 12.80 ms │ 12.50 / 12.75 ±0.15 / 12.89 ms │ no change │
│ QQuery 2 │ 35.28 / 35.48 ±0.12 / 35.66 ms │ 35.08 / 35.38 ±0.20 / 35.69 ms │ no change │
│ QQuery 3 │ 30.32 / 30.65 ±0.59 / 31.83 ms │ 29.99 / 30.50 ±0.48 / 31.31 ms │ no change │
│ QQuery 4 │ 225.06 / 229.66 ±2.83 / 232.87 ms │ 227.80 / 230.20 ±1.29 / 231.65 ms │ no change │
│ QQuery 5 │ 273.45 / 275.19 ±1.62 / 277.67 ms │ 273.62 / 274.68 ±1.27 / 277.15 ms │ no change │
│ QQuery 6 │ 5.46 / 5.88 ±0.26 / 6.15 ms │ 5.59 / 6.10 ±0.43 / 6.87 ms │ no change │
│ QQuery 7 │ 16.39 / 16.47 ±0.10 / 16.66 ms │ 16.11 / 16.16 ±0.03 / 16.21 ms │ no change │
│ QQuery 8 │ 319.86 / 321.42 ±1.27 / 323.71 ms │ 306.72 / 311.70 ±2.88 / 314.67 ms │ no change │
│ QQuery 9 │ 441.39 / 446.04 ±3.73 / 452.08 ms │ 440.78 / 446.44 ±4.91 / 452.09 ms │ no change │
│ QQuery 10 │ 92.31 / 93.13 ±0.44 / 93.51 ms │ 92.39 / 93.02 ±0.50 / 93.60 ms │ no change │
│ QQuery 11 │ 102.56 / 103.60 ±0.75 / 104.59 ms │ 102.65 / 106.15 ±4.23 / 114.31 ms │ no change │
│ QQuery 12 │ 299.84 / 303.23 ±2.59 / 307.11 ms │ 301.28 / 305.16 ±3.79 / 311.32 ms │ no change │
│ QQuery 13 │ 414.14 / 417.14 ±4.15 / 425.31 ms │ 412.82 / 416.38 ±2.38 / 419.17 ms │ no change │
│ QQuery 14 │ 314.00 / 320.66 ±5.73 / 329.17 ms │ 309.58 / 315.76 ±4.94 / 321.62 ms │ no change │
│ QQuery 15 │ 273.19 / 278.98 ±9.57 / 298.05 ms │ 272.92 / 275.24 ±2.35 / 278.99 ms │ no change │
│ QQuery 16 │ 612.59 / 618.84 ±8.45 / 635.41 ms │ 590.06 / 597.87 ±5.18 / 605.96 ms │ no change │
│ QQuery 17 │ 612.80 / 621.98 ±8.66 / 638.16 ms │ 598.17 / 601.22 ±2.15 / 604.45 ms │ no change │
│ QQuery 18 │ 1214.41 / 1244.18 ±17.84 / 1265.62 ms │ 1176.40 / 1198.50 ±14.51 / 1221.19 ms │ no change │
│ QQuery 19 │ 28.65 / 31.20 ±4.35 / 39.86 ms │ 28.95 / 31.23 ±3.42 / 38.00 ms │ no change │
│ QQuery 20 │ 507.20 / 515.71 ±7.22 / 528.23 ms │ 514.24 / 522.83 ±4.39 / 526.73 ms │ no change │
│ QQuery 21 │ 560.04 / 569.23 ±7.67 / 581.62 ms │ 563.55 / 569.78 ±4.56 / 576.29 ms │ no change │
│ QQuery 22 │ 911.34 / 915.05 ±2.80 / 919.06 ms │ 916.07 / 922.46 ±5.91 / 931.90 ms │ no change │
│ QQuery 23 │ 114.77 / 126.86 ±13.26 / 143.45 ms │ 113.41 / 119.51 ±4.20 / 124.63 ms │ +1.06x faster │
│ QQuery 24 │ 39.69 / 42.76 ±2.96 / 48.24 ms │ 39.84 / 41.09 ±1.26 / 43.40 ms │ no change │
│ QQuery 25 │ 144.20 / 148.17 ±5.62 / 159.34 ms │ 145.02 / 148.73 ±5.91 / 160.42 ms │ no change │
│ QQuery 26 │ 50.62 / 51.62 ±0.55 / 52.22 ms │ 61.78 / 65.74 ±3.80 / 72.52 ms │ 1.27x slower │
│ QQuery 27 │ 697.12 / 710.67 ±11.05 / 730.64 ms │ 715.07 / 727.18 ±6.84 / 735.26 ms │ no change │
│ QQuery 28 │ 3000.80 / 3015.23 ±8.42 / 3025.70 ms │ 3020.49 / 3043.37 ±17.54 / 3073.56 ms │ no change │
│ QQuery 29 │ 41.10 / 43.38 ±3.54 / 50.43 ms │ 41.05 / 45.25 ±5.45 / 55.17 ms │ no change │
│ QQuery 30 │ 300.00 / 309.58 ±5.82 / 318.01 ms │ 303.10 / 308.38 ±3.58 / 313.89 ms │ no change │
│ QQuery 31 │ 292.92 / 298.37 ±2.76 / 300.39 ms │ 290.63 / 296.15 ±3.16 / 299.73 ms │ no change │
│ QQuery 32 │ 947.59 / 969.28 ±32.67 / 1033.59 ms │ 884.35 / 904.34 ±16.63 / 932.10 ms │ +1.07x faster │
│ QQuery 33 │ 1373.07 / 1385.94 ±12.42 / 1402.79 ms │ 1380.30 / 1408.13 ±16.46 / 1428.18 ms │ no change │
│ QQuery 34 │ 1397.68 / 1411.47 ±14.11 / 1435.12 ms │ 1403.22 / 1423.08 ±16.46 / 1452.78 ms │ no change │
│ QQuery 35 │ 275.95 / 312.52 ±45.11 / 392.46 ms │ 277.77 / 293.82 ±14.66 / 317.26 ms │ +1.06x faster │
│ QQuery 36 │ 61.57 / 71.14 ±10.34 / 90.53 ms │ 62.52 / 70.25 ±6.57 / 81.33 ms │ no change │
│ QQuery 37 │ 35.97 / 36.67 ±0.51 / 37.47 ms │ 36.52 / 38.58 ±2.20 / 42.09 ms │ 1.05x slower │
│ QQuery 38 │ 34.56 / 38.28 ±4.58 / 45.63 ms │ 34.23 / 38.63 ±5.64 / 49.06 ms │ no change │
│ QQuery 39 │ 113.80 / 126.29 ±8.42 / 136.69 ms │ 115.82 / 126.37 ±5.41 / 131.03 ms │ no change │
│ QQuery 40 │ 18.37 / 18.79 ±0.31 / 19.12 ms │ 17.94 / 18.33 ±0.51 / 19.33 ms │ no change │
│ QQuery 41 │ 16.89 / 17.20 ±0.27 / 17.68 ms │ 16.75 / 16.84 ±0.07 / 16.92 ms │ no change │
│ QQuery 42 │ 14.18 / 14.40 ±0.20 / 14.77 ms │ 14.06 / 18.74 ±8.52 / 35.77 ms │ 1.30x slower │
└───────────┴───────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 16559.50ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 16476.63ms │
│ Average Time (HEAD) │ 385.10ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 383.18ms │
│ Queries Faster │ 3 │
│ Queries Slower │ 3 │
│ Queries with No Change │ 37 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴────────────┘

Resource Usage

clickbench_partitioned — base (merge-base)

Metric	Value
Wall time	85.0s
Peak memory	30.5 GiB
Avg memory	23.3 GiB
CPU user	875.8s
CPU sys	50.0s
Peak spill	0 B

clickbench_partitioned — branch

Metric	Value
Wall time	85.0s
Peak memory	29.3 GiB
Avg memory	22.9 GiB
CPU user	872.1s
CPU sys	49.3s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpcds_sf1.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 6.28 / 6.75 ±0.79 / 8.33 ms │ 6.09 / 6.53 ±0.83 / 8.18 ms │ no change │
│ QQuery 2 │ 46.88 / 47.51 ±0.43 / 48.15 ms │ 44.62 / 44.97 ±0.44 / 45.82 ms │ +1.06x faster │
│ QQuery 3 │ 27.98 / 28.39 ±0.43 / 29.02 ms │ 27.89 / 28.08 ±0.24 / 28.55 ms │ no change │
│ QQuery 4 │ 293.02 / 303.86 ±8.04 / 314.76 ms │ 292.43 / 295.25 ±4.33 / 303.87 ms │ no change │
│ QQuery 5 │ 78.27 / 79.24 ±1.34 / 81.87 ms │ 60.69 / 62.01 ±1.58 / 65.13 ms │ +1.28x faster │
│ QQuery 6 │ 29.94 / 30.38 ±0.55 / 31.44 ms │ 29.21 / 29.74 ±0.28 / 29.98 ms │ no change │
│ QQuery 7 │ 139.56 / 143.60 ±3.42 / 147.51 ms │ 142.06 / 144.87 ±3.32 / 151.40 ms │ no change │
│ QQuery 8 │ 20.25 / 20.57 ±0.30 / 20.94 ms │ 21.30 / 21.42 ±0.12 / 21.60 ms │ no change │
│ QQuery 9 │ 130.20 / 132.97 ±3.29 / 137.66 ms │ 136.95 / 142.55 ±4.33 / 149.24 ms │ 1.07x slower │
│ QQuery 10 │ 114.82 / 116.74 ±2.09 / 119.62 ms │ 113.09 / 118.08 ±5.39 / 127.70 ms │ no change │
│ QQuery 11 │ 198.16 / 201.82 ±2.92 / 207.05 ms │ 207.39 / 209.60 ±1.27 / 211.30 ms │ no change │
│ QQuery 12 │ 21.81 / 22.81 ±0.81 / 24.00 ms │ 22.72 / 22.85 ±0.11 / 23.03 ms │ no change │
│ QQuery 13 │ 144.47 / 147.95 ±5.27 / 158.34 ms │ 147.74 / 149.01 ±1.98 / 152.94 ms │ no change │
│ QQuery 14 │ 520.15 / 524.60 ±4.12 / 531.60 ms │ 511.82 / 521.57 ±5.04 / 525.67 ms │ no change │
│ QQuery 15 │ 20.26 / 20.94 ±1.03 / 22.97 ms │ 17.16 / 18.26 ±1.10 / 20.37 ms │ +1.15x faster │
│ QQuery 16 │ 6.95 / 7.10 ±0.16 / 7.40 ms │ 6.47 / 6.56 ±0.08 / 6.66 ms │ +1.08x faster │
│ QQuery 17 │ 76.71 / 77.33 ±0.53 / 78.17 ms │ 70.67 / 72.89 ±1.68 / 74.44 ms │ +1.06x faster │
│ QQuery 18 │ 279.13 / 287.41 ±4.47 / 292.16 ms │ 218.99 / 220.05 ±1.31 / 222.36 ms │ +1.31x faster │
│ QQuery 19 │ 47.78 / 48.25 ±0.50 / 49.02 ms │ 48.00 / 48.48 ±0.35 / 48.88 ms │ no change │
│ QQuery 20 │ 20.07 / 20.21 ±0.09 / 20.30 ms │ 20.19 / 20.72 ±0.75 / 22.20 ms │ no change │
│ QQuery 21 │ 16.84 / 17.10 ±0.19 / 17.38 ms │ 17.11 / 17.22 ±0.08 / 17.37 ms │ no change │
│ QQuery 22 │ 61.51 / 64.82 ±2.64 / 67.97 ms │ 63.95 / 66.95 ±3.36 / 73.52 ms │ no change │
│ QQuery 23 │ 378.78 / 381.22 ±1.90 / 384.55 ms │ 371.81 / 382.13 ±6.73 / 393.01 ms │ no change │
│ QQuery 24 │ 118.86 / 121.77 ±2.93 / 126.15 ms │ 116.49 / 118.99 ±2.47 / 122.18 ms │ no change │
│ QQuery 25 │ 76.84 / 78.27 ±1.10 / 79.98 ms │ 75.41 / 77.10 ±2.08 / 80.99 ms │ no change │
│ QQuery 26 │ 118.69 / 123.80 ±4.89 / 131.87 ms │ 117.71 / 120.26 ±3.18 / 126.14 ms │ no change │
│ QQuery 27 │ 7.47 / 7.62 ±0.12 / 7.82 ms │ 6.83 / 6.96 ±0.15 / 7.21 ms │ +1.10x faster │
│ QQuery 28 │ 81.20 / 82.35 ±1.92 / 86.18 ms │ 77.03 / 78.85 ±1.38 / 80.50 ms │ no change │
│ QQuery 29 │ 87.79 / 91.96 ±4.88 / 101.53 ms │ 83.23 / 85.08 ±1.74 / 87.86 ms │ +1.08x faster │
│ QQuery 30 │ 33.45 / 33.87 ±0.32 / 34.26 ms │ 33.04 / 33.37 ±0.27 / 33.66 ms │ no change │
│ QQuery 31 │ 134.38 / 137.28 ±2.44 / 141.52 ms │ 123.24 / 125.66 ±2.14 / 128.69 ms │ +1.09x faster │
│ QQuery 32 │ 22.00 / 22.55 ±0.57 / 23.60 ms │ 20.06 / 20.35 ±0.27 / 20.81 ms │ +1.11x faster │
│ QQuery 33 │ 41.56 / 42.09 ±0.44 / 42.85 ms │ 39.78 / 39.99 ±0.13 / 40.13 ms │ no change │
│ QQuery 34 │ 10.18 / 10.71 ±0.51 / 11.45 ms │ 10.21 / 10.35 ±0.24 / 10.82 ms │ no change │
│ QQuery 35 │ 116.74 / 119.14 ±4.00 / 127.09 ms │ 118.42 / 120.27 ±2.83 / 125.76 ms │ no change │
│ QQuery 36 │ 6.40 / 6.71 ±0.30 / 7.26 ms │ 6.33 / 6.47 ±0.16 / 6.76 ms │ no change │
│ QQuery 37 │ 7.53 / 7.67 ±0.09 / 7.80 ms │ 7.51 / 7.60 ±0.09 / 7.76 ms │ no change │
│ QQuery 38 │ 85.61 / 86.89 ±0.89 / 88.29 ms │ 84.71 / 85.14 ±0.33 / 85.58 ms │ no change │
│ QQuery 39 │ 91.15 / 94.11 ±4.05 / 101.97 ms │ 91.73 / 94.59 ±2.91 / 99.58 ms │ no change │
│ QQuery 40 │ 20.59 / 21.38 ±0.67 / 22.24 ms │ 20.01 / 20.34 ±0.20 / 20.64 ms │ no change │
│ QQuery 41 │ 13.36 / 13.52 ±0.13 / 13.69 ms │ 13.52 / 13.67 ±0.18 / 14.01 ms │ no change │
│ QQuery 42 │ 23.85 / 24.19 ±0.32 / 24.76 ms │ 22.47 / 22.68 ±0.15 / 22.84 ms │ +1.07x faster │
│ QQuery 43 │ 5.17 / 5.36 ±0.23 / 5.74 ms │ 5.04 / 5.13 ±0.11 / 5.35 ms │ no change │
│ QQuery 44 │ 10.57 / 10.74 ±0.13 / 10.94 ms │ 10.50 / 10.74 ±0.18 / 10.97 ms │ no change │
│ QQuery 45 │ 27.38 / 28.11 ±0.43 / 28.51 ms │ 26.19 / 26.50 ±0.27 / 26.94 ms │ +1.06x faster │
│ QQuery 46 │ 14.32 / 14.60 ±0.24 / 15.03 ms │ 14.07 / 14.21 ±0.16 / 14.48 ms │ no change │
│ QQuery 47 │ 245.69 / 253.25 ±10.77 / 274.62 ms │ 248.54 / 269.87 ±11.11 / 280.32 ms │ 1.07x slower │
│ QQuery 48 │ 134.80 / 139.15 ±3.34 / 144.34 ms │ 138.60 / 139.82 ±1.38 / 142.43 ms │ no change │
│ QQuery 49 │ 75.10 / 76.09 ±0.65 / 76.81 ms │ 73.68 / 76.56 ±2.18 / 80.06 ms │ no change │
│ QQuery 50 │ 146.15 / 149.71 ±2.46 / 153.76 ms │ 105.15 / 105.69 ±0.45 / 106.34 ms │ +1.42x faster │
│ QQuery 51 │ 103.49 / 106.69 ±2.50 / 110.23 ms │ 104.94 / 109.12 ±2.46 / 112.21 ms │ no change │
│ QQuery 52 │ 23.95 / 24.33 ±0.27 / 24.77 ms │ 23.13 / 23.48 ±0.33 / 23.94 ms │ no change │
│ QQuery 53 │ 39.19 / 39.88 ±0.71 / 41.18 ms │ 39.07 / 39.31 ±0.16 / 39.52 ms │ no change │
│ QQuery 54 │ 26.73 / 27.03 ±0.19 / 27.27 ms │ 25.85 / 26.14 ±0.18 / 26.36 ms │ no change │
│ QQuery 55 │ 23.17 / 23.50 ±0.32 / 23.96 ms │ 22.40 / 22.84 ±0.37 / 23.47 ms │ no change │
│ QQuery 56 │ 37.30 / 38.42 ±0.98 / 39.99 ms │ 36.12 / 37.52 ±2.04 / 41.57 ms │ no change │
│ QQuery 57 │ 135.26 / 137.82 ±2.92 / 143.51 ms │ 136.48 / 139.39 ±2.49 / 143.33 ms │ no change │
│ QQuery 58 │ 60.53 / 61.77 ±1.35 / 64.08 ms │ 59.18 / 60.65 ±1.89 / 64.30 ms │ no change │
│ QQuery 59 │ 93.26 / 95.26 ±2.42 / 99.51 ms │ 92.05 / 93.30 ±1.48 / 95.65 ms │ no change │
│ QQuery 60 │ 42.10 / 42.41 ±0.31 / 42.98 ms │ 41.01 / 41.38 ±0.37 / 42.06 ms │ no change │
│ QQuery 61 │ 13.53 / 13.75 ±0.20 / 14.08 ms │ 13.47 / 13.61 ±0.16 / 13.92 ms │ no change │
│ QQuery 62 │ 44.38 / 45.24 ±0.84 / 46.55 ms │ 43.89 / 45.41 ±1.82 / 48.81 ms │ no change │
│ QQuery 63 │ 40.09 / 41.43 ±1.98 / 45.31 ms │ 39.99 / 40.93 ±0.49 / 41.42 ms │ no change │
│ QQuery 64 │ 476.75 / 481.51 ±3.65 / 487.67 ms │ 522.82 / 531.01 ±6.71 / 541.03 ms │ 1.10x slower │
│ QQuery 65 │ 1060.50 / 1092.54 ±28.79 / 1130.15 ms │ 99.21 / 104.12 ±6.27 / 115.77 ms │ +10.49x faster │
│ QQuery 66 │ 70.12 / 71.64 ±1.53 / 74.58 ms │ 67.77 / 68.93 ±1.17 / 71.05 ms │ no change │
│ QQuery 67 │ 283.22 / 288.87 ±6.03 / 298.00 ms │ 269.30 / 276.45 ±6.85 / 288.73 ms │ no change │
│ QQuery 68 │ 15.22 / 15.38 ±0.10 / 15.53 ms │ 14.31 / 14.50 ±0.22 / 14.92 ms │ +1.06x faster │
│ QQuery 69 │ 112.03 / 117.69 ±5.40 / 127.88 ms │ 102.87 / 105.16 ±2.39 / 108.83 ms │ +1.12x faster │
│ QQuery 70 │ 131.43 / 137.39 ±3.95 / 141.48 ms │ 128.38 / 134.57 ±8.09 / 150.45 ms │ no change │
│ QQuery 71 │ 33.47 / 33.96 ±0.30 / 34.37 ms │ 31.48 / 32.56 ±0.82 / 33.56 ms │ no change │
│ QQuery 72 │ 199.22 / 201.13 ±1.98 / 204.92 ms │ 182.27 / 184.65 ±1.59 / 186.75 ms │ +1.09x faster │
│ QQuery 73 │ 10.75 / 10.98 ±0.14 / 11.17 ms │ 10.07 / 10.36 ±0.35 / 11.04 ms │ +1.06x faster │
│ QQuery 74 │ 139.86 / 144.03 ±5.80 / 155.51 ms │ 135.79 / 138.51 ±3.16 / 143.99 ms │ no change │
│ QQuery 75 │ 167.02 / 168.13 ±1.07 / 169.91 ms │ 167.26 / 170.64 ±2.30 / 173.05 ms │ no change │
│ QQuery 76 │ 46.90 / 50.06 ±4.03 / 57.62 ms │ 45.88 / 46.25 ±0.23 / 46.45 ms │ +1.08x faster │
│ QQuery 77 │ 72.75 / 73.14 ±0.40 / 73.81 ms │ 70.58 / 71.48 ±1.07 / 73.20 ms │ no change │
│ QQuery 78 │ 142.80 / 147.05 ±2.97 / 152.11 ms │ 143.92 / 145.22 ±1.91 / 148.97 ms │ no change │
│ QQuery 79 │ 75.93 / 76.14 ±0.16 / 76.37 ms │ 74.73 / 75.54 ±0.78 / 76.83 ms │ no change │
│ QQuery 80 │ 87.46 / 88.33 ±0.65 / 89.03 ms │ 84.91 / 86.44 ±1.39 / 88.95 ms │ no change │
│ QQuery 81 │ 28.46 / 28.89 ±0.31 / 29.37 ms │ 27.62 / 28.34 ±0.70 / 29.61 ms │ no change │
│ QQuery 82 │ 23.63 / 23.88 ±0.21 / 24.17 ms │ 23.28 / 23.60 ±0.25 / 23.92 ms │ no change │
│ QQuery 83 │ 36.95 / 37.25 ±0.22 / 37.48 ms │ 33.47 / 33.90 ±0.36 / 34.39 ms │ +1.10x faster │
│ QQuery 84 │ 56.70 / 56.93 ±0.16 / 57.13 ms │ 56.21 / 57.41 ±2.01 / 61.41 ms │ no change │
│ QQuery 85 │ 234.40 / 239.31 ±3.47 / 243.91 ms │ 168.96 / 172.71 ±2.85 / 175.91 ms │ +1.39x faster │
│ QQuery 86 │ 30.70 / 32.13 ±2.13 / 36.37 ms │ 30.02 / 30.69 ±1.08 / 32.86 ms │ no change │
│ QQuery 87 │ 89.02 / 89.56 ±0.39 / 89.96 ms │ 86.68 / 88.57 ±1.43 / 90.57 ms │ no change │
│ QQuery 88 │ 70.11 / 75.92 ±5.43 / 83.05 ms │ 72.56 / 76.75 ±3.86 / 83.30 ms │ no change │
│ QQuery 89 │ 47.00 / 49.47 ±3.60 / 56.55 ms │ 47.48 / 49.54 ±2.17 / 53.52 ms │ no change │
│ QQuery 90 │ 19.81 / 20.04 ±0.13 / 20.16 ms │ 20.11 / 20.25 ±0.12 / 20.42 ms │ no change │
│ QQuery 91 │ 71.91 / 72.67 ±0.47 / 73.31 ms │ 73.66 / 75.32 ±1.84 / 78.71 ms │ no change │
│ QQuery 92 │ 32.16 / 32.86 ±0.47 / 33.60 ms │ 29.42 / 29.99 ±0.40 / 30.52 ms │ +1.10x faster │
│ QQuery 93 │ 53.02 / 53.62 ±0.71 / 54.68 ms │ 53.43 / 55.20 ±1.56 / 57.13 ms │ no change │
│ QQuery 94 │ 39.28 / 39.81 ±0.54 / 40.86 ms │ 38.61 / 38.96 ±0.25 / 39.36 ms │ no change │
│ QQuery 95 │ 114.56 / 116.24 ±0.95 / 117.07 ms │ 91.02 / 91.90 ±0.55 / 92.72 ms │ +1.26x faster │
│ QQuery 96 │ 27.90 / 29.66 ±2.92 / 35.48 ms │ 27.93 / 28.25 ±0.26 / 28.57 ms │ no change │
│ QQuery 97 │ 50.41 / 51.58 ±0.78 / 52.82 ms │ 51.23 / 51.89 ±1.00 / 53.85 ms │ no change │
│ QQuery 98 │ 30.80 / 31.20 ±0.25 / 31.56 ms │ 29.63 / 29.97 ±0.26 / 30.42 ms │ no change │
│ QQuery 99 │ 68.79 / 69.49 ±0.74 / 70.90 ms │ 69.79 / 70.30 ±0.59 / 71.44 ms │ no change │
└───────────┴───────────────────────────────────────┴─────────────────────────────────────┴────────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 9580.49ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 8329.09ms │
│ Average Time (HEAD) │ 96.77ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 84.13ms │
│ Queries Faster │ 23 │
│ Queries Slower │ 3 │
│ Queries with No Change │ 73 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴───────────┘

Resource Usage

tpcds — base (merge-base)

Metric	Value
Wall time	50.0s
Peak memory	6.3 GiB
Avg memory	5.8 GiB
CPU user	133.4s
CPU sys	5.1s
Peak spill	0 B

tpcds — branch

Metric	Value
Wall time	45.0s
Peak memory	6.4 GiB
Avg memory	5.9 GiB
CPU user	115.0s
CPU sys	5.0s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangb]

run benchmarks tpch10

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: "false"DATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: "false"changed:
ref: HEADenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: "true"DATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: "true"

[adriangb]

The only regressions are tpch10 w/ filter pushdown on. I guess those are benefiting from the old CASE structure more than this new structure. I am re-running with pushdown on vs. off. I think that's the real comparison. My hypothesis is that the CASE structure is a bigger win when the hash join dynamic filters work but also a much bigger cost when they don't.

[adriangb]

run benchmarks tpch10

baseline:
ref: mainenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: "false"DATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: "false"changed:
ref: mainenv:
DATAFUSION_EXECUTION_PARQUET_PUSHDOWN_FILTERS: "true"DATAFUSION_EXECUTION_PARQUET_REORDER_FILTERS: "true"

[adriangb]

control: #21931 (comment)
test: #21931 (comment)

the goal is to see if this change makes turning pushdown on closer to good or not

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4471617137-180-ctztd 6.12.68+ #1 SMP Wed Apr 1 02:23:28 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 HEAD (42adf29) to main diff using: tpch10
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark running (GKE) | trigger
Instance:c4a-highmem-16 (12 vCPU / 65 GiB) | Linux bench-c4471633434-181-nswg9 6.12.68+ #1 SMP Wed Apr 1 02:23:28 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 main (47655fd) to main diff using: tpch10
Results will be posted here when complete

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpch_sf10.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 317.41 / 318.52 ±0.92 / 319.98 ms │ 322.02 / 323.31 ±1.32 / 325.85 ms │ no change │
│ QQuery 2 │ 120.53 / 122.62 ±1.83 / 125.52 ms │ 231.77 / 234.15 ±2.01 / 237.76 ms │ 1.91x slower │
│ QQuery 3 │ 261.71 / 265.64 ±3.06 / 270.59 ms │ 430.75 / 435.88 ±5.94 / 444.17 ms │ 1.64x slower │
│ QQuery 4 │ 128.41 / 128.94 ±0.74 / 130.36 ms │ 413.86 / 418.21 ±4.00 / 423.52 ms │ 3.24x slower │
│ QQuery 5 │ 370.16 / 375.77 ±7.15 / 388.61 ms │ 812.65 / 828.15 ±8.75 / 838.55 ms │ 2.20x slower │
│ QQuery 6 │ 129.11 / 131.29 ±1.76 / 134.07 ms │ 297.76 / 298.98 ±1.50 / 301.92 ms │ 2.28x slower │
│ QQuery 7 │ 493.16 / 496.43 ±3.59 / 501.97 ms │ 501.97 / 507.40 ±3.51 / 511.83 ms │ no change │
│ QQuery 8 │ 406.39 / 414.95 ±6.23 / 425.13 ms │ 705.35 / 711.18 ±4.60 / 717.68 ms │ 1.71x slower │
│ QQuery 9 │ 575.78 / 578.45 ±3.38 / 585.04 ms │ 1804.40 / 1818.76 ±14.40 / 1844.20 ms │ 3.14x slower │
│ QQuery 10 │ 309.58 / 315.72 ±6.48 / 328.09 ms │ 442.99 / 451.15 ±5.92 / 458.21 ms │ 1.43x slower │
│ QQuery 11 │ 91.03 / 97.27 ±5.12 / 106.25 ms │ 142.45 / 144.00 ±1.08 / 145.79 ms │ 1.48x slower │
│ QQuery 12 │ 183.94 / 188.76 ±5.17 / 198.68 ms │ 392.15 / 397.05 ±6.68 / 410.08 ms │ 2.10x slower │
│ QQuery 13 │ 298.57 / 308.47 ±6.91 / 316.08 ms │ 423.92 / 447.37 ±16.72 / 468.93 ms │ 1.45x slower │
│ QQuery 14 │ 180.65 / 182.01 ±1.44 / 184.70 ms │ 205.84 / 210.40 ±7.28 / 224.88 ms │ 1.16x slower │
│ QQuery 15 │ 323.82 / 327.02 ±3.37 / 331.68 ms │ 402.97 / 410.96 ±5.14 / 417.49 ms │ 1.26x slower │
│ QQuery 16 │ 72.91 / 75.91 ±1.78 / 78.44 ms │ 115.93 / 117.82 ±1.36 / 119.31 ms │ 1.55x slower │
│ QQuery 17 │ 666.37 / 680.16 ±11.66 / 694.40 ms │ 668.92 / 674.00 ±5.82 / 685.15 ms │ no change │
│ QQuery 18 │ 691.55 / 704.23 ±9.34 / 720.54 ms │ 633.15 / 640.51 ±4.87 / 648.18 ms │ +1.10x faster │
│ QQuery 19 │ 256.09 / 276.49 ±15.91 / 296.16 ms │ 289.66 / 292.17 ±2.65 / 297.31 ms │ 1.06x slower │
│ QQuery 20 │ 292.17 / 299.60 ±6.33 / 309.52 ms │ 310.60 / 312.11 ±1.83 / 315.69 ms │ no change │
│ QQuery 21 │ 710.24 / 712.33 ±1.76 / 714.82 ms │ 1417.88 / 1420.43 ±2.42 / 1424.25 ms │ 1.99x slower │
│ QQuery 22 │ 64.13 / 67.51 ±3.45 / 73.68 ms │ 122.53 / 131.15 ±5.13 / 135.54 ms │ 1.94x slower │
└───────────┴────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━┩
│ Total Time (HEAD) │ 7068.09ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 11225.15ms │
│ Average Time (HEAD) │ 321.28ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 510.23ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 17 │
│ Queries with No Change │ 4 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴────────────┘

Resource Usage

tpch10 — base (merge-base)

Metric	Value
Wall time	40.0s
Peak memory	12.4 GiB
Avg memory	8.4 GiB
CPU user	375.6s
CPU sys	21.8s
Peak spill	0 B

tpch10 — branch

Metric	Value
Wall time	60.0s
Peak memory	10.0 GiB
Avg memory	8.0 GiB
CPU user	617.4s
CPU sys	20.4s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangbot]

🤖 Benchmark completed (GKE) | trigger

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

CPU Details (lscpu)

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

Details

Comparing HEAD and worktree-dynamic-filter-restructure
--------------------
Benchmark tpch_sf10.json
--------------------
┏━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Query ┃ HEAD ┃ worktree-dynamic-filter-restructure ┃ Change ┃
┡━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ QQuery 1 │ 313.69 / 316.71 ±3.99 / 324.52 ms │ 317.79 / 319.81 ±2.75 / 325.27 ms │ no change │
│ QQuery 2 │ 120.80 / 122.50 ±1.66 / 125.39 ms │ 172.62 / 174.62 ±1.30 / 175.99 ms │ 1.43x slower │
│ QQuery 3 │ 252.94 / 255.75 ±2.18 / 258.85 ms │ 364.39 / 366.89 ±2.02 / 370.38 ms │ 1.43x slower │
│ QQuery 4 │ 127.66 / 129.75 ±2.31 / 134.19 ms │ 315.90 / 320.12 ±3.76 / 326.84 ms │ 2.47x slower │
│ QQuery 5 │ 383.09 / 391.29 ±6.62 / 399.51 ms │ 589.24 / 598.78 ±8.76 / 613.68 ms │ 1.53x slower │
│ QQuery 6 │ 129.43 / 130.34 ±0.82 / 131.53 ms │ 290.07 / 291.85 ±1.92 / 295.58 ms │ 2.24x slower │
│ QQuery 7 │ 508.07 / 514.76 ±6.35 / 526.52 ms │ 435.58 / 438.26 ±2.34 / 441.66 ms │ +1.17x faster │
│ QQuery 8 │ 408.40 / 416.09 ±6.18 / 425.45 ms │ 621.61 / 624.81 ±2.03 / 627.92 ms │ 1.50x slower │
│ QQuery 9 │ 564.97 / 580.60 ±8.16 / 588.51 ms │ 1347.46 / 1361.81 ±11.04 / 1373.81 ms │ 2.35x slower │
│ QQuery 10 │ 310.11 / 321.14 ±6.81 / 328.39 ms │ 397.12 / 403.50 ±3.90 / 408.62 ms │ 1.26x slower │
│ QQuery 11 │ 88.55 / 94.57 ±5.03 / 103.50 ms │ 141.29 / 148.45 ±10.12 / 167.29 ms │ 1.57x slower │
│ QQuery 12 │ 178.26 / 184.80 ±7.06 / 194.59 ms │ 339.41 / 342.50 ±2.04 / 344.52 ms │ 1.85x slower │
│ QQuery 13 │ 297.96 / 306.97 ±7.40 / 320.27 ms │ 400.72 / 413.28 ±11.97 / 434.98 ms │ 1.35x slower │
│ QQuery 14 │ 175.27 / 177.23 ±2.25 / 181.63 ms │ 199.93 / 203.72 ±2.53 / 206.61 ms │ 1.15x slower │
│ QQuery 15 │ 316.41 / 324.34 ±4.93 / 331.62 ms │ 395.26 / 401.50 ±3.68 / 405.32 ms │ 1.24x slower │
│ QQuery 16 │ 74.74 / 77.56 ±1.91 / 79.93 ms │ 102.27 / 103.73 ±1.12 / 105.41 ms │ 1.34x slower │
│ QQuery 17 │ 683.03 / 696.23 ±7.97 / 704.35 ms │ 757.66 / 771.08 ±17.04 / 804.72 ms │ 1.11x slower │
│ QQuery 18 │ 690.26 / 707.63 ±10.44 / 718.66 ms │ 679.79 / 693.25 ±11.13 / 710.98 ms │ no change │
│ QQuery 19 │ 253.93 / 267.66 ±10.27 / 281.32 ms │ 283.10 / 291.69 ±10.03 / 310.93 ms │ 1.09x slower │
│ QQuery 20 │ 287.04 / 291.81 ±3.65 / 297.45 ms │ 325.34 / 331.13 ±5.97 / 341.37 ms │ 1.13x slower │
│ QQuery 21 │ 702.28 / 724.77 ±19.04 / 747.34 ms │ 1179.06 / 1183.21 ±3.51 / 1189.51 ms │ 1.63x slower │
│ QQuery 22 │ 65.44 / 67.93 ±1.55 / 70.08 ms │ 128.58 / 134.66 ±3.24 / 137.90 ms │ 1.98x slower │
└───────────┴────────────────────────────────────┴───────────────────────────────────────┴───────────────┘
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━┓
┃ Benchmark Summary ┃ ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━┩
│ Total Time (HEAD) │ 7100.43ms │
│ Total Time (worktree-dynamic-filter-restructure) │ 9918.66ms │
│ Average Time (HEAD) │ 322.75ms │
│ Average Time (worktree-dynamic-filter-restructure) │ 450.85ms │
│ Queries Faster │ 1 │
│ Queries Slower │ 19 │
│ Queries with No Change │ 2 │
│ Queries with Failure │ 0 │
└────────────────────────────────────────────────────┴───────────┘

Resource Usage

tpch10 — base (merge-base)

Metric	Value
Wall time	40.0s
Peak memory	12.0 GiB
Avg memory	8.5 GiB
CPU user	378.0s
CPU sys	20.6s
Peak spill	0 B

tpch10 — branch

Metric	Value
Wall time	50.0s
Peak memory	9.8 GiB
Avg memory	8.1 GiB
CPU user	546.5s
CPU sys	18.5s
Peak spill	0 B

---

File an issue against this benchmark runner

[adriangb]

Hmm those SF10 results aren’t great…

[gene-bordegaray]

The CPU utilization went up a good amount.

Could it be better to have some threshold to only use the global_minmax and drop the multi_hash_lookup so we aren't probing every partition map when expensive?

Or we could split and tune datafusion.optimizer.hash_join_inlist_pushdown_max_distinct_values. It is set to 20 right now which caps how large the arrays can be per partition and the union arrays across partitions. This seems pretty low especially for the global. Maybe split this into 2 configs and we can tune them a bit more?

[gene-bordegaray]

hey @adriangb are you still interested in this work. I would be willing to pick up and investigate this

[adriangb]

i am but i’m just strapped way too thin atm, if you can take a look would really appreciate it!

[gene-bordegaray]

i am but i’m just strapped way too thin atm, if you can take a look would really appreciate it!

@adriangb cool I will play around with it a bit and post updates here for you. No rush but just had this on queue of PRs 😄

[github-actions]

Thank you for your contribution. Unfortunately, this pull request is stale because it has been open 60 days with no activity. Please remove the stale label or comment or this will be closed in 7 days.