Plan a V2 scan as a chain of planners from one prepared scan by joseph-isaacs · Pull Request #10694 · vortex-data/vortex

joseph-isaacs · 2026-10-22T11:34:59Z

Draft, against develop. Replaces the stack #10641 → #10648 → #10664 → #10669 → #10678 → #10670 with one branch, none of whose commits adds a plan walk.

What

A V2 scan is planned once per file and run as a chain of planners, each emitting the next through the planning protocol. No stage derives anything a stage before it already knew. The engine composes the planners and owns every decision about where work runs: all CPU is a planner's compute, all IO is a spawn_io read.

PreparedScan                   per-file, IO-free plan set, made once per partition
  PrunePlanner      evaluates the filter's pruning plans over the partition as one batch of
                    zone-map reads and one compute; while it waits, the driver asks it to
                    speculate: the predicate prefetches and projection announcements of every
                    chunk, registered beside the zone-map reads so they coalesce with them;
                    emits one chunk planner for the rows no plan pruned
  ChunkPlanner      cuts those rows into chunks from the partition's own start, emits a filter
                    or projection planner per chunk, skips chunks nothing is selected or kept of
  FilterPlanner     prefetches its chunk's conjunct segments, announces the projection's,
                    runs the conjuncts in adaptive order, emits the projection
  ProjectionPlanner emits one morsel per 64K-row piece

Protocol additions in vortex-scan and vortex-io: Planner::speculate (default none), asked by the run of every parked planner when nothing can compute, before the run polls its sources; IoSource::speculate (default: submit), which a source may serve for every session of its service. A speculated fetch is refused.

The DataFusion driver (persistent/scan_v2.rs) runs the root planner on one task that awaits its reads in place, runs the partition's first chunk on that run, and hands the chunks spawned between two hand-offs to the consumer as one group, each on a task of its own up to the concurrency, or on the root run at a concurrency of one. The same driver serves the file API (vortex_file::v2::ScanDriver).

The segment IO service keeps speculated reads for whichever session fetches a segment, for as long as the service lives, since chunks straddling two partitions or two chunks are fetched by both.

Benchmarks

Local, hot page cache, 14 DataFusion partitions, interleaved A B B A per query, two rounds, 20 iterations at SF1 and 8 at SF10, pooled. Ratios are this head over the baseline; below 1.0 is faster.

All 22 queries, this head (`28e7298`) vs V1	median	min
SF1 (12 iterations, A B B A)	0.963	0.968
SF10 (6 iterations, A B B A)	0.958	0.969

SF1 wins: q12 0.77, q6 0.83, q19 0.82, q14 0.95, q11 0.95, q15 0.96; losses q13 1.03, q8/q9/q21 1.02. SF10 wins: q12 0.75, q6 0.78, q19 0.83, q1 0.88, q15 0.92, q14 0.93; losses q13 1.05, q22 1.03, q10 1.02. The head before this round (141570f) measured SF1 0.999 and SF10 0.973 against V1, and SF1 0.979 / SF10 1.000 against #10664.

Targeted, three rounds, q4/q6/q8/q9/q14/q15/q19/q22 at SF1 and q1/q6/q9/q12/q19 at SF10, median / min geomean against V1: SF1 0.984 / 0.963, SF10 0.884 / 0.894.

Planning once per file, across queries

Every query re-planned each table: the scan cache died with the last scan of a query, so the next query lowered the layout, optimized the projection and filter plans, proved the zones and decoded dictionary values again. q2 at SF1 ran 35 prepares per query. The cache now lives beside the parsed footer in DataFusion's session-level file metadata cache, invalidated with it when the file changes. Local, 14 partitions, against the commit before: SF1 0.965 (q19 0.90, q6 0.93, q2 0.96, q13 0.96), SF10 0.962. Against V1 on the targeted sets: SF1 0.955, SF10 0.81.

Object store: MinIO behind toxiproxy

Local MinIO holding the same SF1 and SF10 files, reached through toxiproxy with a latency toxic of 10 ms ± 4 ms jitter in each direction (about 20 to 28 ms per request) and, for the capped rows, a bandwidth toxic of 100 MB/s per connection. Ratios are this branch over V1; below 1.0 is faster. SF10 ran on 10 partitions so MinIO and the proxy had cores left; SF1 on 14.

MinIO	queries	this head	head before this round (`141570f`)	#10664
SF1, latency only	q1 q3 q6 q9 q12 q14 q15 q19		0.42 (q6 0.18, q19 0.25, q9 1.00)	0.83
SF1, + 100 MB/s cap	same	0.61	0.60
SF10, + 100 MB/s cap	q1 q6 q12 q15 q19	0.90 (q19 0.66, q1 0.79, q6 1.23)	1.27

q9 stays at parity because its three scan phases are serialized by the join tree and each is a couple of round trips in both scans. SF10 over MinIO is bandwidth bound (400 MB per q6 through one box) and V2's runs there are bimodal between about 250 and 800 ms; the median is what the table shows. The slow mode is state a previous V2 query leaves in the process: six single-iteration processes all ran q6 in 360 to 436 ms, while six-iteration processes alternate 230 to 260 ms and 530 to 785 ms, and V1 is stable at about 340 ms after a cold first run. Partition streams and chunk tasks do end within their query (checked with end events on q13: every stream of the orders scan ends 130 to 155 ms after it starts, before the next query begins), so the leftover is not a task or a stream; the scan stream timer's 2.1 s reading on a 200 ms query is a metrics artifact. What a finished V2 query leaves behind that slows the next one over the object store is still open; the HTTP connection pool and the proxy are the remaining suspects.

Two of this round's commits are for that path. Zone maps are read on their own (an urgent intent, coalesced only with other urgent reads) so pruning no longer waits for the data read beside them: over MinIO SF10 pruning finished at 25 to 50 ms instead of 320 to 390. And a file in an object store reads ahead only its first two chunks per partition; the read driver dispatches in registration order, so reading every chunk ahead made the last partitions' first chunks queue behind the first partitions' last (SF10 1.27 to 0.90). A local file keeps reading every chunk ahead, where reading two ahead cost SF10 6%.

The other two are local: a filtered partition is cut into chunks of a quarter of its rows (256K to 1M), so SF1 partitions have a second chunk to spawn and stragglers on the slow cores gate the query less (SF1 0.993, SF10 1.008 against the previous head); and interning probes under a read lock, since partitions opening together serialized on hashing the expression tree.

To rerun: brew install minio minio-mc toxiproxy, minio server <dir> on 127.0.0.1:9000, mc cp --recursive vortex-bench/data/tpch/<sf>/vortex-file-compressed local/tpch/<sf>/, toxiproxy-cli create -l 127.0.0.1:9100 -u 127.0.0.1:9000 minio, add the toxics, then the bench with AWS_ENDPOINT=http://127.0.0.1:9100 AWS_ALLOW_HTTP=true and --opt remote-data-dir=s3://tpch/<sf>/.

What was slower, and why

Found by instruction counts (/usr/bin/time -l) at 1 and 14 partitions and the read driver's trace, not by timing:

Partition rows were cut from the footer's row count on the 8192-row grid, but the writer buffers chunks to megabytes, so at 14 partitions nearly every boundary fell inside a chunk and both neighbours decoded it: q19 at SF1 retired 17% more instructions than V2 DataFusion scan stack: shared segment source + stage prefetch (benchmark run) #10664, none more at one partition. The opener now snaps planning-time rows to the file's natural splits.
Filtered chunks were cut on a file-wide 1M-row grid, giving straddling partitions a tail chunk read separately. Chunks start at the partition's first row.
Pruning plans ran one after another, each a wait and a hop. They run as one batch and one compute.
The zone-map fetch was polled at submit and dispatched alone before the chunk's speculated reads were registered. The run speculates before polling, and the session polls a fetch only when the run polls.
Speculated reads taken by one session were read again by the next session fetching a straddled segment; speculating one chunk per wait left SF10 partitions reading their last chunks late; projection-only partitions handed chunks to the consumer one hop at a time. Each fixed as described above.
The scan cache keys plans, pruned zones and values by the identity of a bound expression's tree, and every DataFusion partition binds the filter and projection on its own, so no two partitions ever shared an entry: every partition built every plan (q6: 448 µs of filter plans and 92 µs of projection, fourteen times over), proved its own zones and decoded every dictionary's values itself. A prepared scan now interns its expressions structurally in the file's cache and each plan is a per-key once-cell: one build per query, prepare 630 to 186 µs on q6, q8 and q9 instructions against V1 from 1.07 and 1.03 to 1.03 and 1.008.

Metrics and phase timing

A prepared scan carries vortex.scan.* metrics its stages update, registered under the partition's labels like the IO metrics, so --show-metrics prints them beside DataFusion's: pruning duration and rows, speculated requests, chunks planned and skipped, filter duration and conjunct rows, projection duration and the wait from its reads going out to its first piece, rows and batches. Timers also report _total. Debug events bracket every phase of a partition (RUST_LOG=vortex_layout::scan=debug,vortex_datafusion::persistent=debug): open prepared and stream returned, pruning started, reads out, speculated and finished, filter started and each conjunct, projection started, reads out, first piece and finished, each batch sent.

Median per lineitem partition at SF1, 14 partitions, microseconds from the partition's open: open prepared 500, stream 720, pruning reads out 1000, speculated 1260, pruning finished 1700 to 2000, filter done 4400 to 8300, projection done 300 later, first batch 400 later. The zone map is the first read dispatched, alone, 1.36 ms after it is polled: that is the read driver's dispatch latency, the same on V1's path.

What was tried and backed out

Branch ji/scan-v2-read-ahead-experiments holds a read-ahead planner after pruning (a window of chunks prefetched, the next announced, intervals listed while the zone maps are in flight), a file announcer that registers a scan's whole interest once per file, an urgent intent that reads zone maps on their own, a bounded drain in the read driver, and speculated reads shared per file rather than per scan. Each was measured against this head and V1 on q4/q6/q8/q9/q14/q15/q19/q22 at SF1 and q1/q6/q9/q12/q19 at SF10. The announcer halved reads on projection-only scans (q9 orders 11 to 2) but, for filtered scans, widened zone-map reads into neighbours' data; the rest left filtered scans issuing two to three times V1's reads, and every combination was 1 to 3% behind this head at SF1 and 4 to 7% at SF10. The IO-level pieces (urgent intent, read-alone segment requests, bounded driver drain) are correct and tested but did not pay on this workload.

Checks

cargo clippy --all-targets --all-features -- -D warnings on vortex-scan, vortex-io, vortex-layout, vortex-file, vortex-datafusion; cargo nextest run on the same (1812 tests on vortex-scan and vortex-io, 1569 on the others); cargo +nightly-2026-09-10 fmt. Not run: the workspace-wide checks CI runs.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE

Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> (cherry picked from commit 6e27e30)

The V2 opener gave every partition its own VortexFile, so each partition read and decoded chunks straddling its boundaries and the file's metadata segments on its own. V1 avoids this through the shared layout reader. Keep a weak per-path map of V2 segment sources beside `layout_readers` so concurrent partitions read through one source. TPC-H SF1 q4 went from 196 segment reads (52.4 MB) to 115 (34.8 MB) per query, matching V1. Hot SF10 geomean against V1 moved from 1.01 to about 0.96 on this machine. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> (cherry picked from commit bec59fe)

Prefetch needs the segments a stage will read before the stage runs. Compiling already registers every segment scan and does no IO, so `Scan::segments` compiles a plan over rows under a mask and lists them, and the per-plan `reach` walk that re-derived the same answer goes, with the `Reach` coordinates it carried. `Shares` used the walk to decode a segment once when several splits or readers of one scan read it, and goes with it. The V2 executor runs one split per scan and shares reads in its IO session, so a scan's splits now each read the segments they overlap, and a plan reading one segment through two paths reads it through each. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

A `PreparedScan` holds a scan's plans over one file, built once from the footer, and makes its planners; nothing in it touches a segment and no planner plans again. The planners form a chain, each emitting the next through the planning protocol: - `PrunePlanner` proves the filter's zone maps over the scan's rows, prefetching the zone tables together, and emits one chunk planner for the rows no proof excludes. Rows a proof excludes are never planned. - `ChunkPlanner` cuts those rows into chunks and emits a filter or a projection planner per chunk, skipping chunks nothing is selected or kept of. The run drives an emitted chunk ahead of the next, so chunks waiting on reads overlap. - `FilterPlanner` prefetches its chunk's conjunct segments and announces the projection's at its first compute, runs the conjuncts in the adaptive order over the surviving rows, and emits the projection. - `ProjectionPlanner` emits one morsel per piece of the rows. A stage with nothing to do is never made: a filter without proofs starts at the chunk planner, a scan without a filter at the projection. A scan runs as one task owning one planning run, instead of one task and run per split plus a private run for the proofs; `with_concurrency` sets nothing. The IO session keeps every read it starts, so a later stage or morsel of a chunk that reads the segment again shares the read. The segments a stage prefetches come from compiling its plans. DataFusion attaches each partition's rows from the cached footer when it repartitions a file, so the V2 opener reads exactly those rows instead of translating a byte range through the layout. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

A scan reads a zoned column through the zoned node's data child and never its zone table, yet the node and its zone table stayed in every optimized data plan. A new `ZonedDataRule` replaces a zoned data plan by its data, so an optimized plan carries a zone table only where a pruning proof reads it. Legacy stats get the same treatment through `LegacyStatsRule`; the rule that dropped legacy stats beneath a zoned plan is subsumed. A zone proof over legacy stats depended on the converted zoned node surviving child optimization. `ExpressionLegacyStatsRule` now converts the legacy child under a proof directly, through the shared `is_zone_proof` check, so the pruning rewrite still meets a zoned plan. With zone tables gone from data plans, the identity projection is optimized and cached like every other projection: chunked columns stay lazy through `Concat::map_chunks`, and a projection-only scan no longer lowers the zone tables it never reads. (cherry picked from commit 91be25d) Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

DataFusion opens a file once per partition, so each partition lowered the layout, planned the filter and proved the zone maps on its own. A scan cache now keeps each distinct filter's plans along with the zones their proofs excluded, and the opener keeps one cache per file that concurrent partitions find through a weak handle and that the last scan holding it frees. A prune planner skips the zone maps of a proof an earlier scan of the file already proved: they are not read again. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

codspeed · 2026-10-22T11:45:49Z

Merging this PR will regress 2 benchmarks

4 warnings, 1 note

⚠️ Walltime results may be unreliable

Some benchmarks used the Walltime instrument outside CodSpeed Macro Runners, which can make the results inconsistent.

Learn more about Macro Runners

See affected benchmarks

⚠️ Unable to measure the execution of 1 benchmark

This benchmark measured no execution time, so its result is not comparable and counts as unchanged. This usually happens when the compiler removes the code under test.

Prevent compiler optimizations

See affected benchmarks

⚠️ 19 benchmarks are dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

⚠️ Base and head ran on different runner types

Some benchmarks with significant performance changes were compared across different runner types, so the changes might not come from the code. Run base and head on the same runner type, or use CodSpeed Macro Runners for more stable results.

Learn more about Macro Runners

Open the report in CodSpeed to investigate

ℹ️ 40 benchmarks have no flamegraph

The flamegraph wasn't generated: the symbols of the profiling trace could not be resolved, because no information about the language frames was collected.

See affected benchmarks

⚡ 7 improved benchmarks
❌ 2 regressed benchmarks
✅ 2399 untouched benchmarks
⏩ 12 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	`BASE`	`HEAD`	Efficiency
❌	Walltime	`bitpack_blocked_compress_avx2`	6.7 µs	7.6 µs	-12.13%
❌	CPU Simulation	`test_scan_filter[int]`	3.1 ms	3.4 ms	-10.88%
⚡	CPU Simulation	`prepare[1000]`	430.3 µs	10.2 µs	×42
⚡	CPU Simulation	`prepare[100]`	74.8 µs	10.2 µs	×7.3
⚡	CPU Simulation	`prepare[1]`	16.5 µs	10.2 µs	+61.07%
⚡	CPU Simulation	`scan[1]`	95.9 µs	76.1 µs	+26.05%
⚡	CPU Simulation	`prepare_and_compile[100]`	322.9 µs	268.5 µs	+20.26%
⚡	CPU Simulation	`prepare_and_compile[1000]`	2.7 ms	2.3 ms	+16.75%
⚡	CPU Simulation	`prepare_and_compile[1]`	52.8 µs	47.7 µs	+10.63%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

_{Comparing ji/scan-v2-planner-chain (28e7298) with develop (d270200)}

12 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

github-actions · 2026-10-22T12:24:36Z

Polar Signals Profiling Results

Latest Run

Status	Commit	Job	Attempt	Link
🟢 Done	`da8a7c8`	tpch-nvme-10	1	Explore Profiling Data
🟢 Done	`da8a7c8`	tpch-s3	1	Explore Profiling Data
🟢 Done	`da8a7c8`	clickbench-nvme	1	Explore Profiling Data
🟢 Done	`da8a7c8`	polarsignals	1	Explore Profiling Data
🟢 Done	`da8a7c8`	tpch-nvme	1	Explore Profiling Data
🟢 Done	`da8a7c8`	tpcds-nvme	1	Explore Profiling Data
🟢 Done	`da8a7c8`	fineweb	1	Explore Profiling Data

Previous Runs (14)

Status	Commit	Job	Attempt	Link
🟢 Done	`af81c35`	clickbench-nvme	1	Explore Profiling Data
🟢 Done	`af81c35`	polarsignals	1	Explore Profiling Data
🟢 Done	`af81c35`	tpch-nvme-10	1	Explore Profiling Data
🟢 Done	`af81c35`	fineweb	1	Explore Profiling Data
🟢 Done	`af81c35`	tpcds-nvme	1	Explore Profiling Data
🟢 Done	`af81c35`	tpch-s3	1	Explore Profiling Data
🟢 Done	`af81c35`	tpch-nvme	1	Explore Profiling Data
🟢 Done	`40ce998`	clickbench-nvme	1	Explore Profiling Data
🟢 Done	`40ce998`	polarsignals	1	Explore Profiling Data
🟢 Done	`40ce998`	tpch-nvme-10	1	Explore Profiling Data
🟢 Done	`40ce998`	fineweb	1	Explore Profiling Data
🟢 Done	`40ce998`	tpch-s3	1	Explore Profiling Data
🟢 Done	`40ce998`	tpcds-nvme	1	Explore Profiling Data
🟢 Done	`40ce998`	tpch-nvme	1	Explore Profiling Data

Powered by Polar Signals Cloud

github-actions · 2026-10-22T12:26:08Z

Benchmarks: PolarSignals Profiling 📖

Commits: PR da8a7c8a vs base 31027fc2
Vortex (geomean): hot 1.010x ➖ · cold 1.060x ➖

datafusion / vortex-file-compressed / ns (1.010x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
polarsignals_q00/datafusion:vortex-file-compressed	45213553 / 44703721 / +1.1%	90066545 / 84854142 / +6.1%	0.50 / 0.53 / -4.7%
polarsignals_q01/datafusion:vortex-file-compressed	50248110 / 50117699 / +0.3%	113548357 / 107175165 / +5.9%	0.44 / 0.47 / -5.4%
polarsignals_q02/datafusion:vortex-file-compressed	43932958 / 43508170 / +1.0%	89395714 / 77807857 / +14.9% 🔴	0.49 / 0.56 / -12.1%
polarsignals_q03/datafusion:vortex-file-compressed	55772515 / 54672978 / +2.0%	117264591 / 111546045 / +5.1%	0.48 / 0.49 / -3.0%
polarsignals_q04/datafusion:vortex-file-compressed	9387902 / 9233656 / +1.7%	49247102 / 40393230 / +21.9% 🔴	0.19 / 0.23 / -16.6%
polarsignals_q05/datafusion:vortex-file-compressed	12140928 / 12216035 / -0.6%	52980785 / 53576648 / -1.1%	0.23 / 0.23 / +0.5%
polarsignals_q06/datafusion:vortex-file-compressed	23631802 / 23279492 / +1.5%	65805157 / 63559450 / +3.5%	0.36 / 0.37 / -2.0%
polarsignals_q07/datafusion:vortex-file-compressed	11120810 / 10987350 / +1.2%	52448145 / 51686778 / +1.5%	0.21 / 0.21 / -0.3%
polarsignals_q08/datafusion:vortex-file-compressed	98290235 / 98711504 / -0.4%	167124378 / 163762990 / +2.1%	0.59 / 0.60 / -2.4%
polarsignals_q09/datafusion:vortex-file-compressed	10139404 / 9896820 / +2.5%	49900450 / 49072206 / +1.7%	0.20 / 0.20 / +0.8%

No file size changes detected.

github-actions · 2026-10-22T12:26:25Z

Benchmarks: FineWeb NVMe 📖

Commits: PR da8a7c8a vs base 31027fc2
Verdict: No clear signal (low confidence)
Attributed Vortex impact: -1.4%
Engines: DataFusion No clear signal (-3.0%, low confidence) · DuckDB No clear signal (+0.1%, low confidence)
Vortex (geomean): hot 0.980x ➖ · cold 1.014x ➖
Parquet (geomean): hot 0.994x ➖ · cold 0.981x ➖
Shifts: Parquet (control) -0.6% · Median polish -0.4%

How to read Verdict and Engines

Verdict: Overall PR-level signal after subtracting baseline drift estimated from Parquet control rows. It can be Likely improvement, Likely regression, or No clear signal.
Engines: Per-engine attribution. DataFusion is compared against DataFusion/Parquet controls; DuckDB is compared against DuckDB/Parquet controls. This answers whether each engine improved or regressed independently.
Confidence: Based on directional consistency, share of rows above the noise floor, and control-run noise.
Hot vs cold: Every measurement is run several times. The first run is reported as the cold run, and the median of the runs after it is reported as the hot run. The verdict and significance use hot runs; each target's geomean reports hot and cold beside each other where the individual runs were recorded and hot alone where they were not, the cold column shows first-run cost per row, and hot/cold is how much of each run the warm path saves. Rows whose results predate per-run reporting show only one value, taken from the value the runner reported.
Table cells: Each cell reads PR / base / %diff. The hot and cold columns mark a change as 🔴 slower or 🟢 faster once it clears this suite's threshold; anything smaller is noise here and is left unmarked, as is hot/cold, because a shift in the warm-up ratio is not a win or a loss by itself.

datafusion / vortex-file-compressed / ns (0.957x ➖, 2↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
fineweb_q00/datafusion:vortex-file-compressed	6749274 / 6897139 / -2.1%	17879209 / 20738004 / -13.8% 🟢	0.38 / 0.33 / +13.5%
fineweb_q01/datafusion:vortex-file-compressed	7328603 / 8950928 / -18.1% 🟢	47024262 / 49376442 / -4.8%	0.16 / 0.18 / -14.0%
fineweb_q02/datafusion:vortex-file-compressed	10125066 / 9931704 / +1.9%	50825719 / 53640180 / -5.2%	0.20 / 0.19 / +7.6%
fineweb_q03/datafusion:vortex-file-compressed	19687004 / 20039514 / -1.8%	107916536 / 114286303 / -5.6%	0.18 / 0.18 / +4.0%
fineweb_q04/datafusion:vortex-file-compressed	67286096 / 66723486 / +0.8%	170001336 / 162291762 / +4.8%	0.40 / 0.41 / -3.7%
fineweb_q05/datafusion:vortex-file-compressed	66483020 / 66463894 / +0.0%	153547255 / 155104989 / -1.0%	0.43 / 0.43 / +1.0%
fineweb_q06/datafusion:vortex-file-compressed	18891536 / 19086454 / -1.0%	110496759 / 123752154 / -10.7% 🟢	0.17 / 0.15 / +10.9%
fineweb_q07/datafusion:vortex-file-compressed	18071722 / 16885802 / +7.0%	108441276 / 115415877 / -6.0%	0.17 / 0.15 / +13.9%
fineweb_q08/datafusion:vortex-file-compressed	7748459 / 9859151 / -21.4% 🟢	63470983 / 50386561 / +26.0% 🔴	0.12 / 0.20 / -37.6%

datafusion / parquet / ns (0.987x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
fineweb_q00/datafusion:parquet	9625359 / 9467992 / +1.7%	31340698 / 33397772 / -6.2%	0.31 / 0.28 / +8.3%
fineweb_q01/datafusion:parquet	74452221 / 74764121 / -0.4%	156200690 / 160503059 / -2.7%	0.48 / 0.47 / +2.3%
fineweb_q02/datafusion:parquet	73610650 / 75148070 / -2.0%	148949195 / 163942663 / -9.1%	0.49 / 0.46 / +7.8%
fineweb_q03/datafusion:parquet	74823972 / 75553874 / -1.0%	166005441 / 159658998 / +4.0%	0.45 / 0.47 / -4.8%
fineweb_q04/datafusion:parquet	81588981 / 84229642 / -3.1%	157043992 / 156251424 / +0.5%	0.52 / 0.54 / -3.6%
fineweb_q05/datafusion:parquet	80207407 / 81395592 / -1.5%	156000673 / 157499418 / -1.0%	0.51 / 0.52 / -0.5%
fineweb_q06/datafusion:parquet	76708418 / 77394579 / -0.9%	162962500 / 169259690 / -3.7%	0.47 / 0.46 / +2.9%
fineweb_q07/datafusion:parquet	73656350 / 75630914 / -2.6%	157318980 / 163432861 / -3.7%	0.47 / 0.46 / +1.2%
fineweb_q08/datafusion:parquet	72507166 / 73955137 / -2.0%	160823027 / 168547456 / -4.6%	0.45 / 0.44 / +2.8%

duckdb / vortex-file-compressed / ns (1.003x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
fineweb_q00/duckdb:vortex-file-compressed	2391974 / 2324310 / +2.9%	7427738 / 7436608 / -0.1%	0.32 / 0.31 / +3.0%
fineweb_q01/duckdb:vortex-file-compressed	9663244 / 9599964 / +0.7%	38728683 / 35034373 / +10.5% 🔴	0.25 / 0.27 / -8.9%
fineweb_q02/duckdb:vortex-file-compressed	10794432 / 10746668 / +0.4%	35609849 / 35746280 / -0.4%	0.30 / 0.30 / +0.8%
fineweb_q03/duckdb:vortex-file-compressed	19392815 / 19256648 / +0.7%	103239698 / 99547163 / +3.7%	0.19 / 0.19 / -2.9%
fineweb_q04/duckdb:vortex-file-compressed	68309682 / 68309000 / +0.0%	118928667 / 115019755 / +3.4%	0.57 / 0.59 / -3.3%
fineweb_q05/duckdb:vortex-file-compressed	67199069 / 66549545 / +1.0%	117367166 / 110435533 / +6.3%	0.57 / 0.60 / -5.0%
fineweb_q06/duckdb:vortex-file-compressed	19261115 / 19181495 / +0.4%	109936216 / 95116090 / +15.6% 🔴	0.18 / 0.20 / -13.1%
fineweb_q07/duckdb:vortex-file-compressed	18727370 / 18860556 / -0.7%	102034252 / 90064083 / +13.3% 🔴	0.18 / 0.21 / -12.4%
fineweb_q08/duckdb:vortex-file-compressed	5493542 / 5653240 / -2.8%	14352118 / 14636158 / -1.9%	0.38 / 0.39 / -0.9%

duckdb / parquet / ns (1.001x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
fineweb_q00/duckdb:parquet	27615378 / 27583605 / +0.1%	29761876 / 29899319 / -0.5%	0.93 / 0.92 / +0.6%
fineweb_q01/duckdb:parquet	58667148 / 60839176 / -3.6%	82977138 / 87811880 / -5.5%	0.71 / 0.69 / +2.0%
fineweb_q02/duckdb:parquet	65247696 / 64631548 / +1.0%	84797492 / 87520140 / -3.1%	0.77 / 0.74 / +4.2%
fineweb_q03/duckdb:parquet	128622593 / 128749726 / -0.1%	194507659 / 188767985 / +3.0%	0.66 / 0.68 / -3.0%
fineweb_q04/duckdb:parquet	178339767 / 181890972 / -2.0%	222456207 / 228431291 / -2.6%	0.80 / 0.80 / +0.7%
fineweb_q05/duckdb:parquet	179127992 / 173075328 / +3.5%	224113787 / 220798189 / +1.5%	0.80 / 0.78 / +2.0%
fineweb_q06/duckdb:parquet	96908414 / 96221180 / +0.7%	159272440 / 159338525 / -0.0%	0.61 / 0.60 / +0.8%
fineweb_q07/duckdb:parquet	98684986 / 97232478 / +1.5%	153660747 / 157247378 / -2.3%	0.64 / 0.62 / +3.9%
fineweb_q08/duckdb:parquet	33536240 / 33478700 / +0.2%	34689398 / 33890855 / +2.4%	0.97 / 0.99 / -2.1%

No file size changes detected.

github-actions · 2026-10-22T12:26:59Z

Benchmarks: TPC-H SF=1 on NVME 📖

Commits: PR da8a7c8a vs base 31027fc2
Verdict: No clear signal (low confidence)
Attributed Vortex impact: +5.8%
Engines: DataFusion Likely regression (+11.8%, medium confidence) · DuckDB No clear signal (+0.1%, low confidence)
Vortex (geomean): hot 1.060x ➖ · cold 0.984x ➖
Parquet (geomean): hot 1.002x ➖ · cold 1.004x ➖
Shifts: Parquet (control) +0.2% · Median polish +0.5%

How to read Verdict and Engines

Verdict: Overall PR-level signal after subtracting baseline drift estimated from Parquet control rows. It can be Likely improvement, Likely regression, or No clear signal.
Engines: Per-engine attribution. DataFusion is compared against DataFusion/Parquet controls; DuckDB is compared against DuckDB/Parquet controls. This answers whether each engine improved or regressed independently.
Confidence: Based on directional consistency, share of rows above the noise floor, and control-run noise.
Hot vs cold: Every measurement is run several times. The first run is reported as the cold run, and the median of the runs after it is reported as the hot run. The verdict and significance use hot runs; each target's geomean reports hot and cold beside each other where the individual runs were recorded and hot alone where they were not, the cold column shows first-run cost per row, and hot/cold is how much of each run the warm path saves. Rows whose results predate per-run reporting show only one value, taken from the value the runner reported.
Table cells: Each cell reads PR / base / %diff. The hot and cold columns mark a change as 🔴 slower or 🟢 faster once it clears this suite's threshold; anything smaller is noise here and is left unmarked, as is hot/cold, because a shift in the warm-up ratio is not a win or a loss by itself.

datafusion / vortex-file-compressed / ns (1.120x ❌, 0↑ 10↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/datafusion:vortex-file-compressed	17423728 / 14713851 / +18.4% 🔴	45736273 / 43934225 / +4.1%	0.38 / 0.33 / +13.8%
tpch_q02/datafusion:vortex-file-compressed	17725602 / 17054655 / +3.9%	35787769 / 34777973 / +2.9%	0.50 / 0.49 / +1.0%
tpch_q03/datafusion:vortex-file-compressed	22581307 / 18917295 / +19.4% 🔴	46942182 / 46703661 / +0.5%	0.48 / 0.41 / +18.8%
tpch_q04/datafusion:vortex-file-compressed	13185645 / 10058589 / +31.1% 🔴	35119882 / 34213054 / +2.7%	0.38 / 0.29 / +27.7%
tpch_q05/datafusion:vortex-file-compressed	39922199 / 36472139 / +9.5%	68999316 / 67806549 / +1.8%	0.58 / 0.54 / +7.6%
tpch_q06/datafusion:vortex-file-compressed	8777460 / 5884484 / +49.2% 🔴	31058940 / 30112638 / +3.1%	0.28 / 0.20 / +44.6%
tpch_q07/datafusion:vortex-file-compressed	55315529 / 51430597 / +7.6%	79763669 / 77890348 / +2.4%	0.69 / 0.66 / +5.0%
tpch_q08/datafusion:vortex-file-compressed	55449002 / 51800896 / +7.0%	77665602 / 82597471 / -6.0%	0.71 / 0.63 / +13.8%
tpch_q09/datafusion:vortex-file-compressed	48909965 / 49434260 / -1.1%	77734973 / 77925190 / -0.2%	0.63 / 0.63 / -0.8%
tpch_q10/datafusion:vortex-file-compressed	20472155 / 17012125 / +20.3% 🔴	52415286 / 52598266 / -0.3%	0.39 / 0.32 / +20.8%
tpch_q11/datafusion:vortex-file-compressed	15965039 / 14880533 / +7.3%	35561901 / 36611768 / -2.9%	0.45 / 0.41 / +10.5%
tpch_q12/datafusion:vortex-file-compressed	25359316 / 22683058 / +11.8% 🔴	44906183 / 44444826 / +1.0%	0.56 / 0.51 / +10.6%
tpch_q13/datafusion:vortex-file-compressed	20889160 / 19775237 / +5.6%	41760531 / 43251174 / -3.4%	0.50 / 0.46 / +9.4%
tpch_q14/datafusion:vortex-file-compressed	15720892 / 14825554 / +6.0%	41538966 / 41497107 / +0.1%	0.38 / 0.36 / +5.9%
tpch_q15/datafusion:vortex-file-compressed	21214764 / 18200006 / +16.6% 🔴	45791081 / 44292868 / +3.4%	0.46 / 0.41 / +12.8%
tpch_q16/datafusion:vortex-file-compressed	18552575 / 18259087 / +1.6%	40179403 / 41217467 / -2.5%	0.46 / 0.44 / +4.2%
tpch_q17/datafusion:vortex-file-compressed	29901879 / 28191278 / +6.1%	63588090 / 61015252 / +4.2%	0.47 / 0.46 / +1.8%
tpch_q18/datafusion:vortex-file-compressed	38333438 / 36862242 / +4.0%	60767468 / 58320267 / +4.2%	0.63 / 0.63 / -0.2%
tpch_q19/datafusion:vortex-file-compressed	18014061 / 14585925 / +23.5% 🔴	44748205 / 44922288 / -0.4%	0.40 / 0.32 / +24.0%
tpch_q20/datafusion:vortex-file-compressed	23368375 / 20334853 / +14.9% 🔴	46849103 / 47226541 / -0.8%	0.50 / 0.43 / +15.8%
tpch_q21/datafusion:vortex-file-compressed	45189419 / 38252573 / +18.1% 🔴	70001587 / 69625004 / +0.5%	0.65 / 0.55 / +17.5%
tpch_q22/datafusion:vortex-file-compressed	11428571 / 11949536 / -4.4%	32822904 / 34137394 / -3.9%	0.35 / 0.35 / -0.5%

datafusion / parquet / ns (1.002x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/datafusion:parquet	18499442 / 18370927 / +0.7%	46100539 / 46265614 / -0.4%	0.40 / 0.40 / +1.1%
tpch_q02/datafusion:parquet	48642152 / 49394968 / -1.5%	76032165 / 75717287 / +0.4%	0.64 / 0.65 / -1.9%
tpch_q03/datafusion:parquet	40525718 / 39990869 / +1.3%	72670247 / 72648407 / +0.0%	0.56 / 0.55 / +1.3%
tpch_q04/datafusion:parquet	17671116 / 17768672 / -0.5%	46566174 / 45759573 / +1.8%	0.38 / 0.39 / -2.3%
tpch_q05/datafusion:parquet	70050794 / 70645920 / -0.8%	102828249 / 104015549 / -1.1%	0.68 / 0.68 / +0.3%
tpch_q06/datafusion:parquet	10759852 / 10413573 / +3.3%	40027202 / 36831440 / +8.7%	0.27 / 0.28 / -4.9%
tpch_q07/datafusion:parquet	77544214 / 77956161 / -0.5%	116344140 / 111570258 / +4.3%	0.67 / 0.70 / -4.6%
tpch_q08/datafusion:parquet	88235016 / 88645454 / -0.5%	127096223 / 127106180 / -0.0%	0.69 / 0.70 / -0.5%
tpch_q09/datafusion:parquet	76565806 / 76597626 / -0.0%	115033706 / 110008271 / +4.6%	0.67 / 0.70 / -4.4%
tpch_q10/datafusion:parquet	31916954 / 31346841 / +1.8%	63767019 / 63455392 / +0.5%	0.50 / 0.49 / +1.3%
tpch_q11/datafusion:parquet	37799798 / 38416254 / -1.6%	65475739 / 66041101 / -0.9%	0.58 / 0.58 / -0.8%
tpch_q12/datafusion:parquet	28369100 / 27993240 / +1.3%	54411153 / 56362413 / -3.5%	0.52 / 0.50 / +5.0%
tpch_q13/datafusion:parquet	27927401 / 27988679 / -0.2%	50722789 / 48880660 / +3.8%	0.55 / 0.57 / -3.8%
tpch_q14/datafusion:parquet	23738555 / 23905606 / -0.7%	52823391 / 54651953 / -3.3%	0.45 / 0.44 / +2.7%
tpch_q15/datafusion:parquet	38417026 / 38523264 / -0.3%	69952084 / 70532073 / -0.8%	0.55 / 0.55 / +0.6%
tpch_q16/datafusion:parquet	28768553 / 28715719 / +0.2%	53041672 / 53060783 / -0.0%	0.54 / 0.54 / +0.2%
tpch_q17/datafusion:parquet	44429031 / 44233782 / +0.4%	82001808 / 82760337 / -0.9%	0.54 / 0.53 / +1.4%
tpch_q18/datafusion:parquet	44699406 / 44090906 / +1.4%	72974750 / 71405128 / +2.2%	0.61 / 0.62 / -0.8%
tpch_q19/datafusion:parquet	20719263 / 20917686 / -0.9%	56260758 / 52876133 / +6.4%	0.37 / 0.40 / -6.9%
tpch_q20/datafusion:parquet	54851549 / 54750943 / +0.2%	91044343 / 92655325 / -1.7%	0.60 / 0.59 / +2.0%
tpch_q21/datafusion:parquet	77544828 / 76831356 / +0.9%	112946232 / 107785703 / +4.8%	0.69 / 0.71 / -3.7%
tpch_q22/datafusion:parquet	21352820 / 21278347 / +0.3%	46335529 / 46144642 / +0.4%	0.46 / 0.46 / -0.1%

duckdb / vortex-file-compressed / ns (1.003x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/duckdb:vortex-file-compressed	10094150 / 10202574 / -1.1%	20953343 / 22770181 / -8.0%	0.48 / 0.45 / +7.5%
tpch_q02/duckdb:vortex-file-compressed	15728234 / 15517523 / +1.4%	25160498 / 24826836 / +1.3%	0.63 / 0.63 / +0.0%
tpch_q03/duckdb:vortex-file-compressed	19652103 / 19770140 / -0.6%	35899114 / 36586638 / -1.9%	0.55 / 0.54 / +1.3%
tpch_q04/duckdb:vortex-file-compressed	12815797 / 12664759 / +1.2%	25607422 / 26715559 / -4.1%	0.50 / 0.47 / +5.6%
tpch_q05/duckdb:vortex-file-compressed	20857351 / 20810652 / +0.2%	34044781 / 34729784 / -2.0%	0.61 / 0.60 / +2.2%
tpch_q06/duckdb:vortex-file-compressed	5871167 / 5769840 / +1.8%	15910350 / 16422571 / -3.1%	0.37 / 0.35 / +5.0%
tpch_q07/duckdb:vortex-file-compressed	25073842 / 24999895 / +0.3%	40082631 / 41904543 / -4.3%	0.63 / 0.60 / +4.9%
tpch_q08/duckdb:vortex-file-compressed	21990891 / 22007515 / -0.1%	39062478 / 39066846 / -0.0%	0.56 / 0.56 / -0.1%
tpch_q09/duckdb:vortex-file-compressed	22973256 / 22265113 / +3.2%	40243865 / 41886406 / -3.9%	0.57 / 0.53 / +7.4%
tpch_q10/duckdb:vortex-file-compressed	23623555 / 23644636 / -0.1%	39240778 / 41879278 / -6.3%	0.60 / 0.56 / +6.6%
tpch_q11/duckdb:vortex-file-compressed	7985668 / 7919924 / +0.8%	15898279 / 15870986 / +0.2%	0.50 / 0.50 / +0.7%
tpch_q12/duckdb:vortex-file-compressed	16760905 / 16677393 / +0.5%	31264885 / 34315077 / -8.9%	0.54 / 0.49 / +10.3%
tpch_q13/duckdb:vortex-file-compressed	21244183 / 21231747 / +0.1%	32222776 / 33667326 / -4.3%	0.66 / 0.63 / +4.5%
tpch_q14/duckdb:vortex-file-compressed	12243936 / 12465575 / -1.8%	24525863 / 24756963 / -0.9%	0.50 / 0.50 / -0.9%
tpch_q15/duckdb:vortex-file-compressed	9253340 / 9244595 / +0.1%	20110829 / 20087961 / +0.1%	0.46 / 0.46 / -0.0%
tpch_q16/duckdb:vortex-file-compressed	19074870 / 18978950 / +0.5%	25870245 / 25579185 / +1.1%	0.74 / 0.74 / -0.6%
tpch_q17/duckdb:vortex-file-compressed	14313482 / 13963565 / +2.5%	26915488 / 29001555 / -7.2%	0.53 / 0.48 / +10.5%
tpch_q18/duckdb:vortex-file-compressed	26801596 / 26642020 / +0.6%	35584232 / 37457586 / -5.0%	0.75 / 0.71 / +5.9%
tpch_q19/duckdb:vortex-file-compressed	12425061 / 12329480 / +0.8%	26045557 / 27996399 / -7.0%	0.48 / 0.44 / +8.3%
tpch_q20/duckdb:vortex-file-compressed	15284105 / 15457154 / -1.1%	31248105 / 33369989 / -6.4%	0.49 / 0.46 / +5.6%
tpch_q21/duckdb:vortex-file-compressed	39025684 / 39331323 / -0.8%	57523504 / 60670342 / -5.2%	0.68 / 0.65 / +4.7%
tpch_q22/duckdb:vortex-file-compressed	18657248 / 18843680 / -1.0%	28557078 / 29619121 / -3.6%	0.65 / 0.64 / +2.7%

duckdb / parquet / ns (1.002x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/duckdb:parquet	18182238 / 17674164 / +2.9%	22197005 / 22122981 / +0.3%	0.82 / 0.80 / +2.5%
tpch_q02/duckdb:parquet	31109474 / 30279321 / +2.7%	38962650 / 38931524 / +0.1%	0.80 / 0.78 / +2.7%
tpch_q03/duckdb:parquet	36234653 / 36416498 / -0.5%	45409854 / 47184665 / -3.8%	0.80 / 0.77 / +3.4%
tpch_q04/duckdb:parquet	26578857 / 26560817 / +0.1%	34523163 / 34543225 / -0.1%	0.77 / 0.77 / +0.1%
tpch_q05/duckdb:parquet	41160956 / 40857778 / +0.7%	51304688 / 50301955 / +2.0%	0.80 / 0.81 / -1.2%
tpch_q06/duckdb:parquet	10827658 / 11201439 / -3.3%	16800352 / 16016055 / +4.9%	0.64 / 0.70 / -7.8%
tpch_q07/duckdb:parquet	45341347 / 44370228 / +2.2%	54458087 / 56220641 / -3.1%	0.83 / 0.79 / +5.5%
tpch_q08/duckdb:parquet	48955780 / 48396529 / +1.2%	59301456 / 59989608 / -1.1%	0.83 / 0.81 / +2.3%
tpch_q09/duckdb:parquet	51957625 / 51772743 / +0.4%	60885273 / 60872937 / +0.0%	0.85 / 0.85 / +0.3%
tpch_q10/duckdb:parquet	42499983 / 42230028 / +0.6%	48803190 / 49912091 / -2.2%	0.87 / 0.85 / +2.9%
tpch_q11/duckdb:parquet	17164898 / 16954923 / +1.2%	24168839 / 24120127 / +0.2%	0.71 / 0.70 / +1.0%
tpch_q12/duckdb:parquet	22867094 / 22994759 / -0.6%	29716645 / 31003239 / -4.1%	0.77 / 0.74 / +3.8%
tpch_q13/duckdb:parquet	36066250 / 35734437 / +0.9%	40285277 / 41649814 / -3.3%	0.90 / 0.86 / +4.3%
tpch_q14/duckdb:parquet	26198903 / 26493370 / -1.1%	33994378 / 32997696 / +3.0%	0.77 / 0.80 / -4.0%
tpch_q15/duckdb:parquet	19287953 / 19104688 / +1.0%	24817741 / 25070351 / -1.0%	0.78 / 0.76 / +2.0%
tpch_q16/duckdb:parquet	31104418 / 30895398 / +0.7%	39078332 / 38855062 / +0.6%	0.80 / 0.80 / +0.1%
tpch_q17/duckdb:parquet	26864748 / 26787312 / +0.3%	35156431 / 33962520 / +3.5%	0.76 / 0.79 / -3.1%
tpch_q18/duckdb:parquet	38058890 / 38690518 / -1.6%	47411200 / 48086840 / -1.4%	0.80 / 0.80 / -0.2%
tpch_q19/duckdb:parquet	22542648 / 23552471 / -4.3%	31504742 / 31570739 / -0.2%	0.72 / 0.75 / -4.1%
tpch_q20/duckdb:parquet	32738791 / 32309714 / +1.3%	43118813 / 42977613 / +0.3%	0.76 / 0.75 / +1.0%
tpch_q21/duckdb:parquet	65167789 / 64649769 / +0.8%	77001240 / 76466838 / +0.7%	0.85 / 0.85 / +0.1%
tpch_q22/duckdb:parquet	24089658 / 24061977 / +0.1%	32020439 / 32226626 / -0.6%	0.75 / 0.75 / +0.8%

No file size changes detected.

github-actions · 2026-10-22T12:28:33Z

Benchmarks: TPC-H SF=10 on NVME 📖

Commits: PR da8a7c8a vs base 31027fc2
Verdict: No clear signal (low confidence)
Attributed Vortex impact: +4.6%
Engines: DataFusion No clear signal (+8.4%, low confidence) · DuckDB No clear signal (+0.9%, low confidence)
Vortex (geomean): hot 1.041x ➖ · cold 1.018x ➖
Parquet (geomean): hot 0.995x ➖ · cold 0.999x ➖
Shifts: Parquet (control) -0.5% · Median polish +0.0%

How to read Verdict and Engines

Verdict: Overall PR-level signal after subtracting baseline drift estimated from Parquet control rows. It can be Likely improvement, Likely regression, or No clear signal.
Engines: Per-engine attribution. DataFusion is compared against DataFusion/Parquet controls; DuckDB is compared against DuckDB/Parquet controls. This answers whether each engine improved or regressed independently.
Confidence: Based on directional consistency, share of rows above the noise floor, and control-run noise.
Hot vs cold: Every measurement is run several times. The first run is reported as the cold run, and the median of the runs after it is reported as the hot run. The verdict and significance use hot runs; each target's geomean reports hot and cold beside each other where the individual runs were recorded and hot alone where they were not, the cold column shows first-run cost per row, and hot/cold is how much of each run the warm path saves. Rows whose results predate per-run reporting show only one value, taken from the value the runner reported.
Table cells: Each cell reads PR / base / %diff. The hot and cold columns mark a change as 🔴 slower or 🟢 faster once it clears this suite's threshold; anything smaller is noise here and is left unmarked, as is hot/cold, because a shift in the warm-up ratio is not a win or a loss by itself.

datafusion / vortex-file-compressed / ns (1.084x ➖, 0↑ 7↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/datafusion:vortex-file-compressed	90847880 / 83819237 / +8.4%	140302397 / 153701322 / -8.7%	0.65 / 0.55 / +18.7%
tpch_q02/datafusion:vortex-file-compressed	70412568 / 64419306 / +9.3%	100600664 / 96263204 / +4.5%	0.70 / 0.67 / +4.6%
tpch_q03/datafusion:vortex-file-compressed	69591632 / 60645145 / +14.8% 🔴	125667285 / 118507951 / +6.0%	0.55 / 0.51 / +8.2%
tpch_q04/datafusion:vortex-file-compressed	37534559 / 33439041 / +12.2% 🔴	82474830 / 72923952 / +13.1% 🔴	0.46 / 0.46 / -0.8%
tpch_q05/datafusion:vortex-file-compressed	121231620 / 115614074 / +4.9%	181281987 / 186235243 / -2.7%	0.67 / 0.62 / +7.7%
tpch_q06/datafusion:vortex-file-compressed	16820231 / 13740406 / +22.4% 🔴	79720553 / 73911613 / +7.9%	0.21 / 0.19 / +13.5%
tpch_q07/datafusion:vortex-file-compressed	138377199 / 131864811 / +4.9%	201786205 / 187493110 / +7.6%	0.69 / 0.70 / -2.5%
tpch_q08/datafusion:vortex-file-compressed	147433821 / 141818703 / +4.0%	219684649 / 216220273 / +1.6%	0.67 / 0.66 / +2.3%
tpch_q09/datafusion:vortex-file-compressed	191011446 / 183402660 / +4.1%	268398622 / 261290511 / +2.7%	0.71 / 0.70 / +1.4%
tpch_q10/datafusion:vortex-file-compressed	77748751 / 70424327 / +10.4% 🔴	145462194 / 138512025 / +5.0%	0.53 / 0.51 / +5.1%
tpch_q11/datafusion:vortex-file-compressed	45801045 / 42350480 / +8.1%	68776288 / 68492390 / +0.4%	0.67 / 0.62 / +7.7%
tpch_q12/datafusion:vortex-file-compressed	42814561 / 43624111 / -1.9%	92813838 / 95767680 / -3.1%	0.46 / 0.46 / +1.3%
tpch_q13/datafusion:vortex-file-compressed	43925048 / 42791790 / +2.6%	86363273 / 75805000 / +13.9% 🔴	0.51 / 0.56 / -9.9%
tpch_q14/datafusion:vortex-file-compressed	25328235 / 24749109 / +2.3%	97353164 / 100787261 / -3.4%	0.26 / 0.25 / +5.9%
tpch_q15/datafusion:vortex-file-compressed	40336700 / 35026665 / +15.2% 🔴	101306293 / 98058902 / +3.3%	0.40 / 0.36 / +11.5%
tpch_q16/datafusion:vortex-file-compressed	39792612 / 36221533 / +9.9%	70600736 / 69120402 / +2.1%	0.56 / 0.52 / +7.6%
tpch_q17/datafusion:vortex-file-compressed	180495923 / 173749276 / +3.9%	231991216 / 225188365 / +3.0%	0.78 / 0.77 / +0.8%
tpch_q18/datafusion:vortex-file-compressed	203702129 / 198891015 / +2.4%	276908758 / 263320024 / +5.2%	0.74 / 0.76 / -2.6%
tpch_q19/datafusion:vortex-file-compressed	43125746 / 35539896 / +21.3% 🔴	119262665 / 97234865 / +22.7% 🔴	0.36 / 0.37 / -1.1%
tpch_q20/datafusion:vortex-file-compressed	66053763 / 60499980 / +9.2%	121790383 / 119484062 / +1.9%	0.54 / 0.51 / +7.1%
tpch_q21/datafusion:vortex-file-compressed	188730300 / 174984103 / +7.9%	249281643 / 235838248 / +5.7%	0.76 / 0.74 / +2.0%
tpch_q22/datafusion:vortex-file-compressed	28203708 / 25204477 / +11.9% 🔴	57348839 / 53986587 / +6.2%	0.49 / 0.47 / +5.3%

datafusion / parquet / ns (1.000x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/datafusion:parquet	105593350 / 106046246 / -0.4%	152768023 / 154339962 / -1.0%	0.69 / 0.69 / +0.6%
tpch_q02/datafusion:parquet	98489820 / 97901810 / +0.6%	128382765 / 132320645 / -3.0%	0.77 / 0.74 / +3.7%
tpch_q03/datafusion:parquet	123053874 / 123208356 / -0.1%	180934718 / 175918629 / +2.9%	0.68 / 0.70 / -2.9%
tpch_q04/datafusion:parquet	52159160 / 52045863 / +0.2%	103730727 / 91549124 / +13.3% 🔴	0.50 / 0.57 / -11.6%
tpch_q05/datafusion:parquet	163357479 / 165441557 / -1.3%	248015508 / 234294629 / +5.9%	0.66 / 0.71 / -6.7%
tpch_q06/datafusion:parquet	41554241 / 41307359 / +0.6%	90667770 / 95484066 / -5.0%	0.46 / 0.43 / +5.9%
tpch_q07/datafusion:parquet	190626813 / 188937757 / +0.9%	265326630 / 261265971 / +1.6%	0.72 / 0.72 / -0.7%
tpch_q08/datafusion:parquet	201434405 / 203101330 / -0.8%	276033599 / 275774551 / +0.1%	0.73 / 0.74 / -0.9%
tpch_q09/datafusion:parquet	223657028 / 226546002 / -1.3%	292859969 / 298158363 / -1.8%	0.76 / 0.76 / +0.5%
tpch_q10/datafusion:parquet	145085040 / 145612170 / -0.4%	214279571 / 210001129 / +2.0%	0.68 / 0.69 / -2.4%
tpch_q11/datafusion:parquet	59310900 / 59663136 / -0.6%	92846574 / 89876831 / +3.3%	0.64 / 0.66 / -3.8%
tpch_q12/datafusion:parquet	72912934 / 70495653 / +3.4%	133796535 / 134763205 / -0.7%	0.54 / 0.52 / +4.2%
tpch_q13/datafusion:parquet	83729071 / 82966721 / +0.9%	129046180 / 127743075 / +1.0%	0.65 / 0.65 / -0.1%
tpch_q14/datafusion:parquet	63698463 / 63772637 / -0.1%	144261841 / 150741556 / -4.3%	0.44 / 0.42 / +4.4%
tpch_q15/datafusion:parquet	108677053 / 108679558 / -0.0%	172715642 / 166185695 / +3.9%	0.63 / 0.65 / -3.8%
tpch_q16/datafusion:parquet	45617539 / 45565890 / +0.1%	75296457 / 78172943 / -3.7%	0.61 / 0.58 / +3.9%
tpch_q17/datafusion:parquet	220179082 / 221411362 / -0.6%	294839906 / 284304016 / +3.7%	0.75 / 0.78 / -4.1%
tpch_q18/datafusion:parquet	215660780 / 216882452 / -0.6%	270082270 / 275705190 / -2.0%	0.80 / 0.79 / +1.5%
tpch_q19/datafusion:parquet	96724312 / 97109592 / -0.4%	181820856 / 182169957 / -0.2%	0.53 / 0.53 / -0.2%
tpch_q20/datafusion:parquet	133827062 / 134012823 / -0.1%	192905461 / 196828447 / -2.0%	0.69 / 0.68 / +1.9%
tpch_q21/datafusion:parquet	301430007 / 298743526 / +0.9%	383864815 / 392867968 / -2.3%	0.79 / 0.76 / +3.3%
tpch_q22/datafusion:parquet	42136529 / 42619973 / -1.1%	72672574 / 72596600 / +0.1%	0.58 / 0.59 / -1.2%

duckdb / vortex-file-compressed / ns (0.999x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/duckdb:vortex-file-compressed	42387479 / 42902893 / -1.2%	81529061 / 81316156 / +0.3%	0.52 / 0.53 / -1.5%
tpch_q02/duckdb:vortex-file-compressed	29095035 / 29586319 / -1.7%	44078247 / 42937367 / +2.7%	0.66 / 0.69 / -4.2%
tpch_q03/duckdb:vortex-file-compressed	56930527 / 56715579 / +0.4%	113305801 / 112849072 / +0.4%	0.50 / 0.50 / -0.0%
tpch_q04/duckdb:vortex-file-compressed	52683165 / 52568778 / +0.2%	94115976 / 93624807 / +0.5%	0.56 / 0.56 / -0.3%
tpch_q05/duckdb:vortex-file-compressed	72863388 / 73335751 / -0.6%	122866333 / 122049176 / +0.7%	0.59 / 0.60 / -1.3%
tpch_q06/duckdb:vortex-file-compressed	19334357 / 19331339 / +0.0%	66836503 / 63483901 / +5.3%	0.29 / 0.30 / -5.0%
tpch_q07/duckdb:vortex-file-compressed	58576288 / 58457603 / +0.2%	128070779 / 127431986 / +0.5%	0.46 / 0.46 / -0.3%
tpch_q08/duckdb:vortex-file-compressed	88826290 / 88709278 / +0.1%	146188701 / 141430930 / +3.4%	0.61 / 0.63 / -3.1%
tpch_q09/duckdb:vortex-file-compressed	107678260 / 106326468 / +1.3%	175567906 / 175554087 / +0.0%	0.61 / 0.61 / +1.3%
tpch_q10/duckdb:vortex-file-compressed	100052865 / 101507511 / -1.4%	151778351 / 158297082 / -4.1%	0.66 / 0.64 / +2.8%
tpch_q11/duckdb:vortex-file-compressed	15212041 / 14853922 / +2.4%	27455719 / 28904087 / -5.0%	0.55 / 0.51 / +7.8%
tpch_q12/duckdb:vortex-file-compressed	57028138 / 57832003 / -1.4%	97766048 / 107336338 / -8.9%	0.58 / 0.54 / +8.3%
tpch_q13/duckdb:vortex-file-compressed	77593523 / 77638429 / -0.1%	104287262 / 103967248 / +0.3%	0.74 / 0.75 / -0.4%
tpch_q14/duckdb:vortex-file-compressed	31265296 / 31553049 / -0.9%	78188077 / 76871476 / +1.7%	0.40 / 0.41 / -2.6%
tpch_q15/duckdb:vortex-file-compressed	26330111 / 26746665 / -1.6%	70807215 / 74594159 / -5.1%	0.37 / 0.36 / +3.7%
tpch_q16/duckdb:vortex-file-compressed	51760224 / 51537955 / +0.4%	62410416 / 64321934 / -3.0%	0.83 / 0.80 / +3.5%
tpch_q17/duckdb:vortex-file-compressed	43600274 / 43986534 / -0.9%	83300560 / 83813312 / -0.6%	0.52 / 0.52 / -0.3%
tpch_q18/duckdb:vortex-file-compressed	101926567 / 99503222 / +2.4%	132706333 / 133874806 / -0.9%	0.77 / 0.74 / +3.3%
tpch_q19/duckdb:vortex-file-compressed	31793428 / 32071554 / -0.9%	90760416 / 90016871 / +0.8%	0.35 / 0.36 / -1.7%
tpch_q20/duckdb:vortex-file-compressed	48244946 / 47929906 / +0.7%	104249375 / 106171276 / -1.8%	0.46 / 0.45 / +2.5%
tpch_q21/duckdb:vortex-file-compressed	107034672 / 105528625 / +1.4%	159206287 / 156542185 / +1.7%	0.67 / 0.67 / -0.3%
tpch_q22/duckdb:vortex-file-compressed	35552894 / 35932351 / -1.1%	55074261 / 54963779 / +0.2%	0.65 / 0.65 / -1.3%

duckdb / parquet / ns (0.990x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/duckdb:parquet	74535487 / 72692709 / +2.5%	84933101 / 87726211 / -3.2%	0.88 / 0.83 / +5.9%
tpch_q02/duckdb:parquet	53660408 / 54189876 / -1.0%	62937247 / 63778463 / -1.3%	0.85 / 0.85 / +0.3%
tpch_q03/duckdb:parquet	96532054 / 98768238 / -2.3%	124222550 / 119419107 / +4.0%	0.78 / 0.83 / -6.0%
tpch_q04/duckdb:parquet	90437168 / 91612844 / -1.3%	103537578 / 105211640 / -1.6%	0.87 / 0.87 / +0.3%
tpch_q05/duckdb:parquet	126616691 / 126076797 / +0.4%	168075118 / 177419118 / -5.3%	0.75 / 0.71 / +6.0%
tpch_q06/duckdb:parquet	45516393 / 47330771 / -3.8%	55247360 / 53645913 / +3.0%	0.82 / 0.88 / -6.6%
tpch_q07/duckdb:parquet	109287668 / 109032834 / +0.2%	147290244 / 141998737 / +3.7%	0.74 / 0.77 / -3.4%
tpch_q08/duckdb:parquet	152555971 / 155607118 / -2.0%	180444623 / 180462460 / -0.0%	0.85 / 0.86 / -2.0%
tpch_q09/duckdb:parquet	167127993 / 165683350 / +0.9%	207719532 / 196977131 / +5.5%	0.80 / 0.84 / -4.3%
tpch_q10/duckdb:parquet	125662269 / 128580846 / -2.3%	138466367 / 140080792 / -1.2%	0.91 / 0.92 / -1.1%
tpch_q11/duckdb:parquet	33096471 / 33462542 / -1.1%	42121117 / 43922420 / -4.1%	0.79 / 0.76 / +3.1%
tpch_q12/duckdb:parquet	74982230 / 74495201 / +0.7%	84706020 / 84766684 / -0.1%	0.89 / 0.88 / +0.7%
tpch_q13/duckdb:parquet	128078950 / 127518974 / +0.4%	127315718 / 132925937 / -4.2%	1.01 / 0.96 / +4.9%
tpch_q14/duckdb:parquet	77242682 / 78206420 / -1.2%	115801664 / 113533328 / +2.0%	0.67 / 0.69 / -3.2%
tpch_q15/duckdb:parquet	70436668 / 68521170 / +2.8%	87888810 / 89576252 / -1.9%	0.80 / 0.76 / +4.8%
tpch_q16/duckdb:parquet	63678970 / 64843534 / -1.8%	67297615 / 66900748 / +0.6%	0.95 / 0.97 / -2.4%
tpch_q17/duckdb:parquet	86855646 / 87064658 / -0.2%	125046772 / 119704369 / +4.5%	0.69 / 0.73 / -4.5%
tpch_q18/duckdb:parquet	123058252 / 126330565 / -2.6%	148674912 / 153470734 / -3.1%	0.83 / 0.82 / +0.6%
tpch_q19/duckdb:parquet	76709621 / 77980781 / -1.6%	104578417 / 107039556 / -2.3%	0.73 / 0.73 / +0.7%
tpch_q20/duckdb:parquet	92196271 / 95394113 / -3.4%	118795418 / 126310209 / -5.9%	0.78 / 0.76 / +2.8%
tpch_q21/duckdb:parquet	178986580 / 184709016 / -3.1%	228160552 / 223904938 / +1.9%	0.78 / 0.82 / -4.9%
tpch_q22/duckdb:parquet	46254328 / 47216661 / -2.0%	54272944 / 56059568 / -3.2%	0.85 / 0.84 / +1.2%

No file size changes detected.

github-actions · 2026-10-22T12:30:13Z

Benchmarks: Clickbench on NVME 📖

Commits: PR da8a7c8a vs base 31027fc2
Verdict: No clear signal (low confidence)
Attributed Vortex impact: +0.4%
Engines: DataFusion No clear signal (+0.6%, low confidence) · DuckDB No clear signal (+0.1%, low confidence)
Vortex (geomean): hot 0.996x ➖ · cold 0.997x ➖
Parquet (geomean): hot 0.992x ➖ · cold 0.994x ➖
Shifts: Parquet (control) -0.8% · Median polish -0.6%

How to read Verdict and Engines

Verdict: Overall PR-level signal after subtracting baseline drift estimated from Parquet control rows. It can be Likely improvement, Likely regression, or No clear signal.
Engines: Per-engine attribution. DataFusion is compared against DataFusion/Parquet controls; DuckDB is compared against DuckDB/Parquet controls. This answers whether each engine improved or regressed independently.
Confidence: Based on directional consistency, share of rows above the noise floor, and control-run noise.
Hot vs cold: Every measurement is run several times. The first run is reported as the cold run, and the median of the runs after it is reported as the hot run. The verdict and significance use hot runs; each target's geomean reports hot and cold beside each other where the individual runs were recorded and hot alone where they were not, the cold column shows first-run cost per row, and hot/cold is how much of each run the warm path saves. Rows whose results predate per-run reporting show only one value, taken from the value the runner reported.
Table cells: Each cell reads PR / base / %diff. The hot and cold columns mark a change as 🔴 slower or 🟢 faster once it clears this suite's threshold; anything smaller is noise here and is left unmarked, as is hot/cold, because a shift in the warm-up ratio is not a win or a loss by itself.

datafusion / vortex-file-compressed / ns (1.000x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
clickbench_q00/datafusion:vortex-file-compressed	1265482 / 1261248 / +0.3%	22280559 / 22208176 / +0.3%	0.06 / 0.06 / +0.0%
clickbench_q01/datafusion:vortex-file-compressed	5912716 / 5741486 / +3.0%	28269088 / 32461466 / -12.9% 🟢	0.21 / 0.18 / +18.3%
clickbench_q02/datafusion:vortex-file-compressed	9553192 / 9506805 / +0.5%	36604223 / 36047214 / +1.5%	0.26 / 0.26 / -1.0%
clickbench_q03/datafusion:vortex-file-compressed	10845186 / 10987468 / -1.3%	53536981 / 54059496 / -1.0%	0.20 / 0.20 / -0.3%
clickbench_q04/datafusion:vortex-file-compressed	53828222 / 55368442 / -2.8%	87990060 / 90498730 / -2.8%	0.61 / 0.61 / -0.0%
clickbench_q05/datafusion:vortex-file-compressed	72950173 / 74096078 / -1.5%	115650254 / 114855147 / +0.7%	0.63 / 0.65 / -2.2%
clickbench_q06/datafusion:vortex-file-compressed	1287510 / 1282883 / +0.4%	22242326 / 22415841 / -0.8%	0.06 / 0.06 / +1.1%
clickbench_q07/datafusion:vortex-file-compressed	9374174 / 9278688 / +1.0%	31935195 / 33317213 / -4.1%	0.29 / 0.28 / +5.4%
clickbench_q08/datafusion:vortex-file-compressed	75361706 / 75295750 / +0.1%	115156686 / 119698431 / -3.8%	0.65 / 0.63 / +4.0%
clickbench_q09/datafusion:vortex-file-compressed	172906240 / 168945109 / +2.3%	205220873 / 208476106 / -1.6%	0.84 / 0.81 / +4.0%
clickbench_q10/datafusion:vortex-file-compressed	22294192 / 22039146 / +1.2%	68311565 / 66952078 / +2.0%	0.33 / 0.33 / -0.9%
clickbench_q11/datafusion:vortex-file-compressed	25538708 / 25674424 / -0.5%	67425779 / 70448132 / -4.3%	0.38 / 0.36 / +3.9%
clickbench_q12/datafusion:vortex-file-compressed	72054596 / 71714575 / +0.5%	116330376 / 122310525 / -4.9%	0.62 / 0.59 / +5.6%
clickbench_q13/datafusion:vortex-file-compressed	118587458 / 115821386 / +2.4%	184508362 / 177485708 / +4.0%	0.64 / 0.65 / -1.5%
clickbench_q14/datafusion:vortex-file-compressed	75620336 / 73953617 / +2.3%	117562189 / 121150935 / -3.0%	0.64 / 0.61 / +5.4%
clickbench_q15/datafusion:vortex-file-compressed	61766280 / 60744492 / +1.7%	96603702 / 98857521 / -2.3%	0.64 / 0.61 / +4.1%
clickbench_q16/datafusion:vortex-file-compressed	148928859 / 150124496 / -0.8%	208590641 / 207159238 / +0.7%	0.71 / 0.72 / -1.5%
clickbench_q17/datafusion:vortex-file-compressed	146225342 / 149583786 / -2.2%	205107510 / 209383454 / -2.0%	0.71 / 0.71 / -0.2%
clickbench_q18/datafusion:vortex-file-compressed	288267680 / 296181507 / -2.7%	369503645 / 373254936 / -1.0%	0.78 / 0.79 / -1.7%
clickbench_q19/datafusion:vortex-file-compressed	7565269 / 7476818 / +1.2%	48413390 / 48177529 / +0.5%	0.16 / 0.16 / +0.7%
clickbench_q20/datafusion:vortex-file-compressed	59810815 / 61254057 / -2.4%	233417334 / 224197416 / +4.1%	0.26 / 0.27 / -6.2%
clickbench_q21/datafusion:vortex-file-compressed	89153956 / 87192698 / +2.2%	271047661 / 265857519 / +2.0%	0.33 / 0.33 / +0.3%
clickbench_q22/datafusion:vortex-file-compressed	96132536 / 100680410 / -4.5%	342700069 / 345081792 / -0.7%	0.28 / 0.29 / -3.9%
clickbench_q23/datafusion:vortex-file-compressed	198190198 / 188356090 / +5.2%	472598535 / 384793861 / +22.8% 🔴	0.42 / 0.49 / -14.3%
clickbench_q24/datafusion:vortex-file-compressed	15519532 / 16014626 / -3.1%	66096262 / 65324025 / +1.2%	0.23 / 0.25 / -4.2%
clickbench_q25/datafusion:vortex-file-compressed	18025236 / 17481601 / +3.1%	59917251 / 59875824 / +0.1%	0.30 / 0.29 / +3.0%
clickbench_q26/datafusion:vortex-file-compressed	15456732 / 16242988 / -4.8%	67144606 / 66297846 / +1.3%	0.23 / 0.25 / -6.0%
clickbench_q27/datafusion:vortex-file-compressed	97332412 / 96676686 / +0.7%	240892013 / 239499020 / +0.6%	0.40 / 0.40 / +0.1%
clickbench_q28/datafusion:vortex-file-compressed	550231666 / 551104877 / -0.2%	609846215 / 631233857 / -3.4%	0.90 / 0.87 / +3.3%
clickbench_q29/datafusion:vortex-file-compressed	23265136 / 22995814 / +1.2%	50314326 / 51410757 / -2.1%	0.46 / 0.45 / +3.4%
clickbench_q30/datafusion:vortex-file-compressed	57604924 / 57650175 / -0.1%	104482918 / 108930612 / -4.1%	0.55 / 0.53 / +4.2%
clickbench_q31/datafusion:vortex-file-compressed	65780786 / 67793026 / -3.0%	151973790 / 153714180 / -1.1%	0.43 / 0.44 / -1.9%
clickbench_q32/datafusion:vortex-file-compressed	286239169 / 288644032 / -0.8%	379428075 / 391828553 / -3.2%	0.75 / 0.74 / +2.4%
clickbench_q33/datafusion:vortex-file-compressed	318725164 / 320609870 / -0.6%	413686843 / 427342547 / -3.2%	0.77 / 0.75 / +2.7%
clickbench_q34/datafusion:vortex-file-compressed	317525946 / 319377776 / -0.6%	417669717 / 418909519 / -0.3%	0.76 / 0.76 / -0.3%
clickbench_q35/datafusion:vortex-file-compressed	55743383 / 55197768 / +1.0%	91240167 / 90080396 / +1.3%	0.61 / 0.61 / -0.3%
clickbench_q36/datafusion:vortex-file-compressed	37520074 / 37367441 / +0.4%	64964825 / 65392519 / -0.7%	0.58 / 0.57 / +1.1%
clickbench_q37/datafusion:vortex-file-compressed	21924420 / 21896678 / +0.1%	48242052 / 48368288 / -0.3%	0.45 / 0.45 / +0.4%
clickbench_q38/datafusion:vortex-file-compressed	11891182 / 11694756 / +1.7%	39983273 / 40000035 / -0.0%	0.30 / 0.29 / +1.7%
clickbench_q39/datafusion:vortex-file-compressed	77904634 / 77638046 / +0.3%	111567580 / 110844802 / +0.7%	0.70 / 0.70 / -0.3%
clickbench_q40/datafusion:vortex-file-compressed	11349894 / 11104250 / +2.2%	38123804 / 37516031 / +1.6%	0.30 / 0.30 / +0.6%
clickbench_q41/datafusion:vortex-file-compressed	10746372 / 10684592 / +0.6%	36370716 / 35955665 / +1.2%	0.30 / 0.30 / -0.6%
clickbench_q42/datafusion:vortex-file-compressed	10620936 / 11052988 / -3.9%	36671702 / 36421968 / +0.7%	0.29 / 0.30 / -4.6%

datafusion / parquet / ns (0.994x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
clickbench_q00/datafusion:parquet	1268284 / 1253978 / +1.1%	50882526 / 51070083 / -0.4%	0.02 / 0.02 / +1.5%
clickbench_q01/datafusion:parquet	16287415 / 16001286 / +1.8%	72859266 / 73865020 / -1.4%	0.22 / 0.22 / +3.2%
clickbench_q02/datafusion:parquet	18679840 / 19313344 / -3.3%	82464107 / 80365040 / +2.6%	0.23 / 0.24 / -5.7%
clickbench_q03/datafusion:parquet	16836500 / 17204936 / -2.1%	83188817 / 80962081 / +2.8%	0.20 / 0.21 / -4.8%
clickbench_q04/datafusion:parquet	62252643 / 62535487 / -0.5%	127666630 / 124742286 / +2.3%	0.49 / 0.50 / -2.7%
clickbench_q05/datafusion:parquet	88325201 / 87995028 / +0.4%	159996413 / 160052326 / -0.0%	0.55 / 0.55 / +0.4%
clickbench_q06/datafusion:parquet	1300870 / 1285248 / +1.2%	50987041 / 50118233 / +1.7%	0.03 / 0.03 / -0.5%
clickbench_q07/datafusion:parquet	16652089 / 16962812 / -1.8%	67862931 / 68012807 / -0.2%	0.25 / 0.25 / -1.6%
clickbench_q08/datafusion:parquet	84922532 / 86004940 / -1.3%	166341808 / 162557590 / +2.3%	0.51 / 0.53 / -3.5%
clickbench_q09/datafusion:parquet	188825686 / 189220114 / -0.2%	276257412 / 265396654 / +4.1%	0.68 / 0.71 / -4.1%
clickbench_q10/datafusion:parquet	36723390 / 36977369 / -0.7%	102972333 / 101980767 / +1.0%	0.36 / 0.36 / -1.6%
clickbench_q11/datafusion:parquet	42614242 / 42919521 / -0.7%	113284403 / 111759799 / +1.4%	0.38 / 0.38 / -2.0%
clickbench_q12/datafusion:parquet	94634544 / 95408308 / -0.8%	163608896 / 164379422 / -0.5%	0.58 / 0.58 / -0.3%
clickbench_q13/datafusion:parquet	146819808 / 148886972 / -1.4%	229178392 / 227207886 / +0.9%	0.64 / 0.66 / -2.2%
clickbench_q14/datafusion:parquet	101250752 / 101473982 / -0.2%	174678287 / 173272347 / +0.8%	0.58 / 0.59 / -1.0%
clickbench_q15/datafusion:parquet	71557254 / 70604414 / +1.3%	138182510 / 131741062 / +4.9%	0.52 / 0.54 / -3.4%
clickbench_q16/datafusion:parquet	156451026 / 156500464 / -0.0%	235160976 / 237592083 / -1.0%	0.67 / 0.66 / +1.0%
clickbench_q17/datafusion:parquet	156694204 / 156087693 / +0.4%	235970437 / 237996379 / -0.9%	0.66 / 0.66 / +1.3%
clickbench_q18/datafusion:parquet	312232789 / 308928661 / +1.1%	416729347 / 422772746 / -1.4%	0.75 / 0.73 / +2.5%
clickbench_q19/datafusion:parquet	19151710 / 19176332 / -0.1%	76853719 / 77162837 / -0.4%	0.25 / 0.25 / +0.3%
clickbench_q20/datafusion:parquet	177828986 / 178482614 / -0.4%	285157685 / 280657834 / +1.6%	0.62 / 0.64 / -1.9%
clickbench_q21/datafusion:parquet	171666807 / 175119702 / -2.0%	300586750 / 300961775 / -0.1%	0.57 / 0.58 / -1.8%
clickbench_q22/datafusion:parquet	258684546 / 261557145 / -1.1%	425104079 / 429921012 / -1.1%	0.61 / 0.61 / +0.0%
clickbench_q23/datafusion:parquet	1008368063 / 1009086440 / -0.1%	1203904610 / 1218589573 / -1.2%	0.84 / 0.83 / +1.1%
clickbench_q24/datafusion:parquet	28990420 / 29748389 / -2.5%	89605947 / 89405958 / +0.2%	0.32 / 0.33 / -2.8%
clickbench_q25/datafusion:parquet	36241602 / 36886880 / -1.7%	102597381 / 102788329 / -0.2%	0.35 / 0.36 / -1.6%
clickbench_q26/datafusion:parquet	29821966 / 30452891 / -2.1%	88265906 / 90123702 / -2.1%	0.34 / 0.34 / -0.0%
clickbench_q27/datafusion:parquet	190511120 / 193338084 / -1.5%	290940762 / 297193685 / -2.1%	0.65 / 0.65 / +0.7%
clickbench_q28/datafusion:parquet	615605066 / 620037760 / -0.7%	704071467 / 707915517 / -0.5%	0.87 / 0.88 / -0.2%
clickbench_q29/datafusion:parquet	28704234 / 28969171 / -0.9%	90580745 / 90649284 / -0.1%	0.32 / 0.32 / -0.8%
clickbench_q30/datafusion:parquet	90615514 / 91789734 / -1.3%	174651228 / 173497366 / +0.7%	0.52 / 0.53 / -1.9%
clickbench_q31/datafusion:parquet	102535415 / 102789277 / -0.2%	204968003 / 197970932 / +3.5%	0.50 / 0.52 / -3.7%
clickbench_q32/datafusion:parquet	309244240 / 310018836 / -0.2%	420673060 / 432955253 / -2.8%	0.74 / 0.72 / +2.7%
clickbench_q33/datafusion:parquet	377467852 / 378390285 / -0.2%	474401100 / 478744919 / -0.9%	0.80 / 0.79 / +0.7%
clickbench_q34/datafusion:parquet	376406673 / 379447670 / -0.8%	477160260 / 484818964 / -1.6%	0.79 / 0.78 / +0.8%
clickbench_q35/datafusion:parquet	65139170 / 63527172 / +2.5%	122932097 / 126279210 / -2.7%	0.53 / 0.50 / +5.3%
clickbench_q36/datafusion:parquet	73339674 / 73459316 / -0.2%	135972239 / 135267962 / +0.5%	0.54 / 0.54 / -0.7%
clickbench_q37/datafusion:parquet	35281088 / 35504100 / -0.6%	91012874 / 92586204 / -1.7%	0.39 / 0.38 / +1.1%
clickbench_q38/datafusion:parquet	46888180 / 47666606 / -1.6%	109260567 / 110007885 / -0.7%	0.43 / 0.43 / -1.0%
clickbench_q39/datafusion:parquet	143543019 / 143325492 / +0.2%	212611585 / 215368185 / -1.3%	0.68 / 0.67 / +1.5%
clickbench_q40/datafusion:parquet	19781739 / 20218712 / -2.2%	77911215 / 79247199 / -1.7%	0.25 / 0.26 / -0.5%
clickbench_q41/datafusion:parquet	18437312 / 18665490 / -1.2%	74762397 / 75999552 / -1.6%	0.25 / 0.25 / +0.4%
clickbench_q42/datafusion:parquet	19194405 / 19457247 / -1.4%	75083269 / 75763974 / -0.9%	0.26 / 0.26 / -0.5%

duckdb / vortex-file-compressed / ns (0.992x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
clickbench_q00/duckdb:vortex-file-compressed	2431650 / 2473686 / -1.7%	25256696 / 25641112 / -1.5%	0.10 / 0.10 / -0.2%
clickbench_q01/duckdb:vortex-file-compressed	7601330 / 7611955 / -0.1%	29362444 / 30157460 / -2.6%	0.26 / 0.25 / +2.6%
clickbench_q02/duckdb:vortex-file-compressed	8382464 / 8320431 / +0.7%	40062040 / 36188808 / +10.7% 🔴	0.21 / 0.23 / -9.0%
clickbench_q03/duckdb:vortex-file-compressed	16216172 / 16123425 / +0.6%	61835256 / 60601006 / +2.0%	0.26 / 0.27 / -1.4%
clickbench_q04/duckdb:vortex-file-compressed	56895075 / 58327280 / -2.5%	99355695 / 102164949 / -2.7%	0.57 / 0.57 / +0.3%
clickbench_q05/duckdb:vortex-file-compressed	65759037 / 65073504 / +1.1%	101118633 / 95638431 / +5.7%	0.65 / 0.68 / -4.4%
clickbench_q06/duckdb:vortex-file-compressed	2744960 / 2745149 / -0.0%	29234001 / 33337882 / -12.3% 🟢	0.09 / 0.08 / +14.0%
clickbench_q07/duckdb:vortex-file-compressed	8793659 / 8903376 / -1.2%	29997128 / 32661695 / -8.2%	0.29 / 0.27 / +7.5%
clickbench_q08/duckdb:vortex-file-compressed	68829144 / 68774498 / +0.1%	115554928 / 116793085 / -1.1%	0.60 / 0.59 / +1.2%
clickbench_q09/duckdb:vortex-file-compressed	88915942 / 89691031 / -0.9%	136573079 / 138135714 / -1.1%	0.65 / 0.65 / +0.3%
clickbench_q10/duckdb:vortex-file-compressed	27433155 / 27774694 / -1.2%	71837361 / 73410011 / -2.1%	0.38 / 0.38 / +0.9%
clickbench_q11/duckdb:vortex-file-compressed	29718556 / 29951371 / -0.8%	74954839 / 75434535 / -0.6%	0.40 / 0.40 / -0.1%
clickbench_q12/duckdb:vortex-file-compressed	70727765 / 71791413 / -1.5%	102033935 / 101998086 / +0.0%	0.69 / 0.70 / -1.5%
clickbench_q13/duckdb:vortex-file-compressed	129268512 / 130243302 / -0.7%	169121329 / 171518396 / -1.4%	0.76 / 0.76 / +0.7%
clickbench_q14/duckdb:vortex-file-compressed	75801270 / 76517858 / -0.9%	106421608 / 105806229 / +0.6%	0.71 / 0.72 / -1.5%
clickbench_q15/duckdb:vortex-file-compressed	76712565 / 77428818 / -0.9%	118842129 / 116171970 / +2.3%	0.65 / 0.67 / -3.2%
clickbench_q16/duckdb:vortex-file-compressed	140182210 / 140817846 / -0.5%	184628084 / 187034386 / -1.3%	0.76 / 0.75 / +0.8%
clickbench_q17/duckdb:vortex-file-compressed	143548953 / 143503642 / +0.0%	187339728 / 180996517 / +3.5%	0.77 / 0.79 / -3.4%
clickbench_q18/duckdb:vortex-file-compressed	264890176 / 259813888 / +2.0%	325900310 / 327416375 / -0.5%	0.81 / 0.79 / +2.4%
clickbench_q19/duckdb:vortex-file-compressed	9183268 / 9322268 / -1.5%	44446540 / 46185175 / -3.8%	0.21 / 0.20 / +2.4%
clickbench_q20/duckdb:vortex-file-compressed	65633760 / 66135222 / -0.8%	173858353 / 176566155 / -1.5%	0.38 / 0.37 / +0.8%
clickbench_q21/duckdb:vortex-file-compressed	97577814 / 99385166 / -1.8%	279651925 / 283471200 / -1.3%	0.35 / 0.35 / -0.5%
clickbench_q22/duckdb:vortex-file-compressed	177322574 / 177169149 / +0.1%	436656327 / 442764855 / -1.4%	0.41 / 0.40 / +1.5%
clickbench_q23/duckdb:vortex-file-compressed	35334294 / 35823536 / -1.4%	94164697 / 91055304 / +3.4%	0.38 / 0.39 / -4.6%
clickbench_q24/duckdb:vortex-file-compressed	14715672 / 14829160 / -0.8%	40074480 / 41006563 / -2.3%	0.37 / 0.36 / +1.5%
clickbench_q25/duckdb:vortex-file-compressed	18956072 / 19130250 / -0.9%	55426813 / 54855838 / +1.0%	0.34 / 0.35 / -1.9%
clickbench_q26/duckdb:vortex-file-compressed	11440372 / 11709151 / -2.3%	39188643 / 38067292 / +2.9%	0.29 / 0.31 / -5.1%
clickbench_q27/duckdb:vortex-file-compressed	54678886 / 55375238 / -1.3%	223507209 / 228375116 / -2.1%	0.24 / 0.24 / +0.9%
clickbench_q28/duckdb:vortex-file-compressed	453700632 / 451705399 / +0.4%	527032266 / 524329649 / +0.5%	0.86 / 0.86 / -0.1%
clickbench_q29/duckdb:vortex-file-compressed	18162461 / 18569736 / -2.2%	49151786 / 45930727 / +7.0%	0.37 / 0.40 / -8.6%
clickbench_q30/duckdb:vortex-file-compressed	59114828 / 60649438 / -2.5%	113635360 / 114083074 / -0.4%	0.52 / 0.53 / -2.1%
clickbench_q31/duckdb:vortex-file-compressed	110570178 / 114332830 / -3.3%	200563454 / 179105668 / +12.0% 🔴	0.55 / 0.64 / -13.6%
clickbench_q32/duckdb:vortex-file-compressed	381331598 / 374520922 / +1.8%	418294883 / 411681052 / +1.6%	0.91 / 0.91 / +0.2%
clickbench_q33/duckdb:vortex-file-compressed	348182406 / 348384500 / -0.1%	452846864 / 455141387 / -0.5%	0.77 / 0.77 / +0.4%
clickbench_q34/duckdb:vortex-file-compressed	362974781 / 368452730 / -1.5%	468055704 / 472667423 / -1.0%	0.78 / 0.78 / -0.5%
clickbench_q35/duckdb:vortex-file-compressed	57644898 / 58682670 / -1.8%	94020227 / 96284674 / -2.4%	0.61 / 0.61 / +0.6%
clickbench_q36/duckdb:vortex-file-compressed	12563067 / 12338751 / +1.8%	34171586 / 34262539 / -0.3%	0.37 / 0.36 / +2.1%
clickbench_q37/duckdb:vortex-file-compressed	8611667 / 8753183 / -1.6%	30766483 / 30762892 / +0.0%	0.28 / 0.28 / -1.6%
clickbench_q38/duckdb:vortex-file-compressed	9273860 / 9137472 / +1.5%	33581209 / 31828746 / +5.5%	0.28 / 0.29 / -3.8%
clickbench_q39/duckdb:vortex-file-compressed	20600093 / 20703902 / -0.5%	42909098 / 43783996 / -2.0%	0.48 / 0.47 / +1.5%
clickbench_q40/duckdb:vortex-file-compressed	7533384 / 7791591 / -3.3%	30346857 / 31038524 / -2.2%	0.25 / 0.25 / -1.1%
clickbench_q41/duckdb:vortex-file-compressed	8305154 / 8373800 / -0.8%	31127906 / 30679660 / +1.5%	0.27 / 0.27 / -2.2%
clickbench_q42/duckdb:vortex-file-compressed	7454471 / 7588680 / -1.8%	29737109 / 30317986 / -1.9%	0.25 / 0.25 / +0.2%

duckdb / parquet / ns (0.991x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
clickbench_q00/duckdb:parquet	12516576 / 12650054 / -1.1%	19069447 / 20273137 / -5.9%	0.66 / 0.62 / +5.2%
clickbench_q01/duckdb:parquet	19299697 / 19809408 / -2.6%	26736147 / 26257577 / +1.8%	0.72 / 0.75 / -4.3%
clickbench_q02/duckdb:parquet	27594951 / 27390572 / +0.7%	38240937 / 37198607 / +2.8%	0.72 / 0.74 / -2.0%
clickbench_q03/duckdb:parquet	26806442 / 26730794 / +0.3%	37931313 / 38768087 / -2.2%	0.71 / 0.69 / +2.5%
clickbench_q04/duckdb:parquet	65784968 / 66184226 / -0.6%	77420527 / 80656303 / -4.0%	0.85 / 0.82 / +3.6%
clickbench_q05/duckdb:parquet	88029658 / 91007850 / -3.3%	107536113 / 109148979 / -1.5%	0.82 / 0.83 / -1.8%
clickbench_q06/duckdb:parquet	26234820 / 25562459 / +2.6%	36263895 / 34065390 / +6.5%	0.72 / 0.75 / -3.6%
clickbench_q07/duckdb:parquet	21196472 / 22205884 / -4.5%	28605173 / 29438201 / -2.8%	0.74 / 0.75 / -1.8%
clickbench_q08/duckdb:parquet	77711977 / 80949642 / -4.0%	97328617 / 95053383 / +2.4%	0.80 / 0.85 / -6.2%
clickbench_q09/duckdb:parquet	104277072 / 106525661 / -2.1%	124654964 / 125075821 / -0.3%	0.84 / 0.85 / -1.8%
clickbench_q10/duckdb:parquet	45701454 / 45409953 / +0.6%	60404030 / 60499553 / -0.2%	0.76 / 0.75 / +0.8%
clickbench_q11/duckdb:parquet	49987188 / 50835093 / -1.7%	70559357 / 66973884 / +5.4%	0.71 / 0.76 / -6.7%
clickbench_q12/duckdb:parquet	99255464 / 99540686 / -0.3%	119080513 / 119225411 / -0.1%	0.83 / 0.83 / -0.2%
clickbench_q13/duckdb:parquet	156190093 / 159564938 / -2.1%	181514708 / 184286331 / -1.5%	0.86 / 0.87 / -0.6%
clickbench_q14/duckdb:parquet	108370971 / 107740879 / +0.6%	131559694 / 130760619 / +0.6%	0.82 / 0.82 / -0.0%
clickbench_q15/duckdb:parquet	92352018 / 94285780 / -2.1%	104637085 / 107428276 / -2.6%	0.88 / 0.88 / +0.6%
clickbench_q16/duckdb:parquet	182501752 / 197607024 / -7.6%	190737575 / 191238549 / -0.3%	0.96 / 1.03 / -7.4%
clickbench_q17/duckdb:parquet	184999540 / 181182310 / +2.1%	192719395 / 203278137 / -5.2%	0.96 / 0.89 / +7.7%
clickbench_q18/duckdb:parquet	317572610 / 314302903 / +1.0%	341314353 / 353583023 / -3.5%	0.93 / 0.89 / +4.7%
clickbench_q19/duckdb:parquet	23283727 / 23914886 / -2.6%	31468311 / 32319404 / -2.6%	0.74 / 0.74 / -0.0%
clickbench_q20/duckdb:parquet	197712297 / 193695783 / +2.1%	245334397 / 242302698 / +1.3%	0.81 / 0.80 / +0.8%
clickbench_q21/duckdb:parquet	237326668 / 245648786 / -3.4%	282553029 / 279351140 / +1.1%	0.84 / 0.88 / -4.5%
clickbench_q22/duckdb:parquet	323416558 / 337148476 / -4.1%	371141898 / 372113452 / -0.3%	0.87 / 0.91 / -3.8%
clickbench_q23/duckdb:parquet	181442237 / 186685502 / -2.8%	249212512 / 256326831 / -2.8%	0.73 / 0.73 / -0.0%
clickbench_q24/duckdb:parquet	35883434 / 36682046 / -2.2%	47703255 / 54262723 / -12.1% 🟢	0.75 / 0.68 / +11.3%
clickbench_q25/duckdb:parquet	46835403 / 46988930 / -0.3%	67291658 / 71447665 / -5.8%	0.70 / 0.66 / +5.8%
clickbench_q26/duckdb:parquet	36597336 / 36955640 / -1.0%	53103674 / 51050191 / +4.0%	0.69 / 0.72 / -4.8%
clickbench_q27/duckdb:parquet	211609853 / 193576787 / +9.3%	243703020 / 242956218 / +0.3%	0.87 / 0.80 / +9.0%
clickbench_q28/duckdb:parquet	855714192 / 843733842 / +1.4%	841705742 / 840858308 / +0.1%	1.02 / 1.00 / +1.3%
clickbench_q29/duckdb:parquet	30881035 / 31183474 / -1.0%	39376511 / 39691605 / -0.8%	0.78 / 0.79 / -0.2%
clickbench_q30/duckdb:parquet	110605952 / 113654239 / -2.7%	141117112 / 141845146 / -0.5%	0.78 / 0.80 / -2.2%
clickbench_q31/duckdb:parquet	163571094 / 166963099 / -2.0%	185020234 / 212836273 / -13.1% 🟢	0.88 / 0.78 / +12.7%
clickbench_q32/duckdb:parquet	406488732 / 405030908 / +0.4%	431982113 / 417974618 / +3.4%	0.94 / 0.97 / -2.9%
clickbench_q33/duckdb:parquet	483882566 / 477225110 / +1.4%	464253488 / 488382159 / -4.9%	1.04 / 0.98 / +6.7%
clickbench_q34/duckdb:parquet	479874666 / 479203214 / +0.1%	469736779 / 467016979 / +0.6%	1.02 / 1.03 / -0.4%
clickbench_q35/duckdb:parquet	66398958 / 67239655 / -1.3%	80717217 / 80118372 / +0.7%	0.82 / 0.84 / -2.0%
clickbench_q36/duckdb:parquet	37148591 / 37914602 / -2.0%	45965642 / 45883790 / +0.2%	0.81 / 0.83 / -2.2%
clickbench_q37/duckdb:parquet	28206296 / 28606758 / -1.4%	37800437 / 37711912 / +0.2%	0.75 / 0.76 / -1.6%
clickbench_q38/duckdb:parquet	31011624 / 30529641 / +1.6%	39041096 / 40054896 / -2.5%	0.79 / 0.76 / +4.2%
clickbench_q39/duckdb:parquet	62624586 / 63057700 / -0.7%	70918791 / 72923579 / -2.7%	0.88 / 0.86 / +2.1%
clickbench_q40/duckdb:parquet	18237954 / 18709288 / -2.5%	25804026 / 27059683 / -4.6%	0.71 / 0.69 / +2.2%
clickbench_q41/duckdb:parquet	18659614 / 18954402 / -1.6%	26263083 / 27031776 / -2.8%	0.71 / 0.70 / +1.3%
clickbench_q42/duckdb:parquet	18229009 / 18036739 / +1.1%	28012880 / 27671709 / +1.2%	0.65 / 0.65 / -0.2%

No file size changes detected.

github-actions · 2026-10-22T12:30:14Z

Benchmarks: TPC-DS SF=1 on NVME 📖

Commits: PR da8a7c8a vs base 31027fc2
Verdict: No clear signal (low confidence)
Attributed Vortex impact: +0.5%
Engines: DataFusion No clear signal (+1.0%, low confidence) · DuckDB No clear signal (-0.1%, low confidence)
Vortex (geomean): hot 0.999x ➖ · cold 1.003x ➖
Parquet (geomean): hot 0.994x ➖ · cold 1.001x ➖
Shifts: Parquet (control) -0.6% · Median polish -0.5%

How to read Verdict and Engines

Verdict: Overall PR-level signal after subtracting baseline drift estimated from Parquet control rows. It can be Likely improvement, Likely regression, or No clear signal.
Engines: Per-engine attribution. DataFusion is compared against DataFusion/Parquet controls; DuckDB is compared against DuckDB/Parquet controls. This answers whether each engine improved or regressed independently.
Confidence: Based on directional consistency, share of rows above the noise floor, and control-run noise.
Hot vs cold: Every measurement is run several times. The first run is reported as the cold run, and the median of the runs after it is reported as the hot run. The verdict and significance use hot runs; each target's geomean reports hot and cold beside each other where the individual runs were recorded and hot alone where they were not, the cold column shows first-run cost per row, and hot/cold is how much of each run the warm path saves. Rows whose results predate per-run reporting show only one value, taken from the value the runner reported.
Table cells: Each cell reads PR / base / %diff. The hot and cold columns mark a change as 🔴 slower or 🟢 faster once it clears this suite's threshold; anything smaller is noise here and is left unmarked, as is hot/cold, because a shift in the warm-up ratio is not a win or a loss by itself.

datafusion / vortex-file-compressed / ns (1.005x ➖, 0↑ 1↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpcds_q01/datafusion:vortex-file-compressed	24716222 / 24422950 / +1.2%	42644690 / 41713721 / +2.2%	0.58 / 0.59 / -1.0%
tpcds_q02/datafusion:vortex-file-compressed	47359232 / 47249193 / +0.2%	65352114 / 65210228 / +0.2%	0.72 / 0.72 / +0.0%
tpcds_q03/datafusion:vortex-file-compressed	12250054 / 12259355 / -0.1%	26952047 / 27469091 / -1.9%	0.45 / 0.45 / +1.8%
tpcds_q04/datafusion:vortex-file-compressed	121709147 / 123073218 / -1.1%	155003686 / 152160996 / +1.9%	0.79 / 0.81 / -2.9%
tpcds_q05/datafusion:vortex-file-compressed	59728257 / 59858944 / -0.2%	80807882 / 80358024 / +0.6%	0.74 / 0.74 / -0.8%
tpcds_q06/datafusion:vortex-file-compressed	22871700 / 23056359 / -0.8%	44745799 / 41776705 / +7.1%	0.51 / 0.55 / -7.4%
tpcds_q07/datafusion:vortex-file-compressed	31691746 / 31123659 / +1.8%	52632193 / 52221538 / +0.8%	0.60 / 0.60 / +1.0%
tpcds_q08/datafusion:vortex-file-compressed	24737110 / 24761588 / -0.1%	41274175 / 40866146 / +1.0%	0.60 / 0.61 / -1.1%
tpcds_q09/datafusion:vortex-file-compressed	18266809 / 18211640 / +0.3%	35999576 / 33007998 / +9.1%	0.51 / 0.55 / -8.0%
tpcds_q10/datafusion:vortex-file-compressed	26927330 / 27274416 / -1.3%	47717668 / 46161892 / +3.4%	0.56 / 0.59 / -4.5%
tpcds_q11/datafusion:vortex-file-compressed	64831090 / 63905464 / +1.4%	90464656 / 91459183 / -1.1%	0.72 / 0.70 / +2.6%
tpcds_q12/datafusion:vortex-file-compressed	16689942 / 16902012 / -1.3%	31408803 / 31785413 / -1.2%	0.53 / 0.53 / -0.1%
tpcds_q13/datafusion:vortex-file-compressed	32447899 / 31875405 / +1.8%	54199388 / 54636491 / -0.8%	0.60 / 0.58 / +2.6%
tpcds_q14/datafusion:vortex-file-compressed	497982214 / 495254547 / +0.6%	555960113 / 546774120 / +1.7%	0.90 / 0.91 / -1.1%
tpcds_q15/datafusion:vortex-file-compressed	16062306 / 15420112 / +4.2%	33661749 / 33932883 / -0.8%	0.48 / 0.45 / +5.0%
tpcds_q16/datafusion:vortex-file-compressed	19754314 / 20087974 / -1.7%	38683695 / 37975694 / +1.9%	0.51 / 0.53 / -3.5%
tpcds_q17/datafusion:vortex-file-compressed	55317136 / 59171354 / -6.5%	74266786 / 74649401 / -0.5%	0.74 / 0.79 / -6.0%
tpcds_q18/datafusion:vortex-file-compressed	68258494 / 68902366 / -0.9%	90026172 / 90277737 / -0.3%	0.76 / 0.76 / -0.7%
tpcds_q19/datafusion:vortex-file-compressed	17315036 / 17092177 / +1.3%	38290714 / 39162658 / -2.2%	0.45 / 0.44 / +3.6%
tpcds_q20/datafusion:vortex-file-compressed	17879881 / 17636308 / +1.4%	33537439 / 34642134 / -3.2%	0.53 / 0.51 / +4.7%
tpcds_q21/datafusion:vortex-file-compressed	16195308 / 16451934 / -1.6%	38052801 / 38284480 / -0.6%	0.43 / 0.43 / -1.0%
tpcds_q22/datafusion:vortex-file-compressed	58135660 / 57363202 / +1.3%	79735432 / 80177293 / -0.6%	0.73 / 0.72 / +1.9%
tpcds_q23/datafusion:vortex-file-compressed	73518464 / 71131666 / +3.4%	100315287 / 105854192 / -5.2%	0.73 / 0.67 / +9.1%
tpcds_q24/datafusion:vortex-file-compressed	68065076 / 67904586 / +0.2%	89730861 / 90094037 / -0.4%	0.76 / 0.75 / +0.6%
tpcds_q25/datafusion:vortex-file-compressed	50228750 / 53333052 / -5.8%	77480020 / 76138784 / +1.8%	0.65 / 0.70 / -7.5%
tpcds_q26/datafusion:vortex-file-compressed	30625956 / 30640026 / -0.0%	48850559 / 48054573 / +1.7%	0.63 / 0.64 / -1.7%
tpcds_q27/datafusion:vortex-file-compressed	61860180 / 63534224 / -2.6%	87478282 / 80912224 / +8.1%	0.71 / 0.79 / -9.9%
tpcds_q28/datafusion:vortex-file-compressed	18176367 / 18333928 / -0.9%	36092030 / 34797932 / +3.7%	0.50 / 0.53 / -4.4%
tpcds_q29/datafusion:vortex-file-compressed	58977137 / 51117640 / +15.4% 🔴	79788691 / 71963254 / +10.9% 🔴	0.74 / 0.71 / +4.1%
tpcds_q30/datafusion:vortex-file-compressed	27115280 / 27094850 / +0.1%	44870013 / 43720666 / +2.6%	0.60 / 0.62 / -2.5%
tpcds_q31/datafusion:vortex-file-compressed	58498760 / 58192894 / +0.5%	82406807 / 79978380 / +3.0%	0.71 / 0.73 / -2.4%
tpcds_q32/datafusion:vortex-file-compressed	13499190 / 13494499 / +0.0%	30547960 / 29672834 / +2.9%	0.44 / 0.45 / -2.8%
tpcds_q33/datafusion:vortex-file-compressed	28475574 / 28132704 / +1.2%	48341097 / 47071656 / +2.7%	0.59 / 0.60 / -1.4%
tpcds_q34/datafusion:vortex-file-compressed	17672546 / 17543160 / +0.7%	33459986 / 34708690 / -3.6%	0.53 / 0.51 / +4.5%
tpcds_q35/datafusion:vortex-file-compressed	34424458 / 34127384 / +0.9%	57188915 / 55315562 / +3.4%	0.60 / 0.62 / -2.4%
tpcds_q36/datafusion:vortex-file-compressed	46802712 / 47413962 / -1.3%	72247259 / 71102774 / +1.6%	0.65 / 0.67 / -2.9%
tpcds_q37/datafusion:vortex-file-compressed	13297728 / 12941790 / +2.8%	32967869 / 32568213 / +1.2%	0.40 / 0.40 / +1.5%
tpcds_q38/datafusion:vortex-file-compressed	30709872 / 29790266 / +3.1%	51000244 / 49089690 / +3.9%	0.60 / 0.61 / -0.8%
tpcds_q39/datafusion:vortex-file-compressed	62712220 / 61958185 / +1.2%	87712544 / 88244056 / -0.6%	0.71 / 0.70 / +1.8%
tpcds_q40/datafusion:vortex-file-compressed	24920232 / 24117339 / +3.3%	43058527 / 42841029 / +0.5%	0.58 / 0.56 / +2.8%
tpcds_q41/datafusion:vortex-file-compressed	17371216 / 17009686 / +2.1%	30470298 / 30116323 / +1.2%	0.57 / 0.56 / +0.9%
tpcds_q42/datafusion:vortex-file-compressed	11561326 / 11194194 / +3.3%	25484435 / 25800524 / -1.2%	0.45 / 0.43 / +4.6%
tpcds_q43/datafusion:vortex-file-compressed	12942394 / 12851160 / +0.7%	30391256 / 29176165 / +4.2%	0.43 / 0.44 / -3.3%
tpcds_q44/datafusion:vortex-file-compressed	26269090 / 25861536 / +1.6%	47467299 / 48172347 / -1.5%	0.55 / 0.54 / +3.1%
tpcds_q45/datafusion:vortex-file-compressed	18850768 / 18524946 / +1.8%	36157537 / 37165720 / -2.7%	0.52 / 0.50 / +4.6%
tpcds_q46/datafusion:vortex-file-compressed	23286185 / 22879180 / +1.8%	44388183 / 45223748 / -1.8%	0.52 / 0.51 / +3.7%
tpcds_q47/datafusion:vortex-file-compressed	77479972 / 77956142 / -0.6%	106043505 / 105204058 / +0.8%	0.73 / 0.74 / -1.4%
tpcds_q48/datafusion:vortex-file-compressed	29558118 / 29227950 / +1.1%	51802771 / 50087705 / +3.4%	0.57 / 0.58 / -2.2%
tpcds_q49/datafusion:vortex-file-compressed	58078756 / 59353887 / -2.1%	79494955 / 78029772 / +1.9%	0.73 / 0.76 / -4.0%
tpcds_q50/datafusion:vortex-file-compressed	32542588 / 33840671 / -3.8%	53530188 / 51289595 / +4.4%	0.61 / 0.66 / -7.9%
tpcds_q51/datafusion:vortex-file-compressed	68009820 / 67295610 / +1.1%	90095041 / 88284880 / +2.1%	0.75 / 0.76 / -1.0%
tpcds_q52/datafusion:vortex-file-compressed	11770732 / 11874957 / -0.9%	26839760 / 26710267 / +0.5%	0.44 / 0.44 / -1.4%
tpcds_q53/datafusion:vortex-file-compressed	17488818 / 17981568 / -2.7%	36580133 / 37147095 / -1.5%	0.48 / 0.48 / -1.2%
tpcds_q54/datafusion:vortex-file-compressed	33122904 / 33014860 / +0.3%	59491296 / 57829652 / +2.9%	0.56 / 0.57 / -2.5%
tpcds_q55/datafusion:vortex-file-compressed	11522577 / 11171222 / +3.1%	26752620 / 25574530 / +4.6%	0.43 / 0.44 / -1.4%
tpcds_q56/datafusion:vortex-file-compressed	26968887 / 26417084 / +2.1%	47056889 / 47619460 / -1.2%	0.57 / 0.55 / +3.3%
tpcds_q57/datafusion:vortex-file-compressed	78199378 / 77382479 / +1.1%	103090365 / 100025981 / +3.1%	0.76 / 0.77 / -1.9%
tpcds_q58/datafusion:vortex-file-compressed	41558603 / 41043717 / +1.3%	63123615 / 62722170 / +0.6%	0.66 / 0.65 / +0.6%
tpcds_q59/datafusion:vortex-file-compressed	31369170 / 31111953 / +0.8%	50262953 / 50870624 / -1.2%	0.62 / 0.61 / +2.0%
tpcds_q60/datafusion:vortex-file-compressed	26978438 / 27227950 / -0.9%	47020306 / 45943148 / +2.3%	0.57 / 0.59 / -3.2%
tpcds_q61/datafusion:vortex-file-compressed	23615549 / 23187124 / +1.8%	43752698 / 42219484 / +3.6%	0.54 / 0.55 / -1.7%
tpcds_q62/datafusion:vortex-file-compressed	18123289 / 17865185 / +1.4%	33935541 / 33075175 / +2.6%	0.53 / 0.54 / -1.1%
tpcds_q63/datafusion:vortex-file-compressed	18004328 / 17696958 / +1.7%	37242129 / 37729044 / -1.3%	0.48 / 0.47 / +3.1%
tpcds_q64/datafusion:vortex-file-compressed	224740678 / 225325008 / -0.3%	259578589 / 257206818 / +0.9%	0.87 / 0.88 / -1.2%
tpcds_q65/datafusion:vortex-file-compressed	75942699 / 74155888 / +2.4%	93654315 / 93507785 / +0.2%	0.81 / 0.79 / +2.2%
tpcds_q66/datafusion:vortex-file-compressed	52674644 / 52179442 / +0.9%	71722585 / 70572058 / +1.6%	0.73 / 0.74 / -0.7%
tpcds_q67/datafusion:vortex-file-compressed	58078136 / 58509098 / -0.7%	87393425 / 86280391 / +1.3%	0.66 / 0.68 / -2.0%
tpcds_q68/datafusion:vortex-file-compressed	23157996 / 22531870 / +2.8%	43871603 / 43740568 / +0.3%	0.53 / 0.52 / +2.5%
tpcds_q69/datafusion:vortex-file-compressed	27929009 / 27859934 / +0.2%	48929284 / 47785128 / +2.4%	0.57 / 0.58 / -2.1%
tpcds_q70/datafusion:vortex-file-compressed	23083156 / 23744342 / -2.8%	45291790 / 43206892 / +4.8%	0.51 / 0.55 / -7.3%
tpcds_q71/datafusion:vortex-file-compressed	18860680 / 19730928 / -4.4%	39749935 / 38161372 / +4.2%	0.47 / 0.52 / -8.2%
tpcds_q72/datafusion:vortex-file-compressed	346323724 / 363326532 / -4.7%	400385919 / 377357859 / +6.1%	0.86 / 0.96 / -10.2%
tpcds_q73/datafusion:vortex-file-compressed	16297931 / 16227445 / +0.4%	34478109 / 36520074 / -5.6%	0.47 / 0.44 / +6.4%
tpcds_q74/datafusion:vortex-file-compressed	46059708 / 46262381 / -0.4%	71399366 / 70102268 / +1.9%	0.65 / 0.66 / -2.2%
tpcds_q75/datafusion:vortex-file-compressed	137896464 / 138025502 / -0.1%	157302228 / 160998509 / -2.3%	0.88 / 0.86 / +2.3%
tpcds_q76/datafusion:vortex-file-compressed	27522458 / 26231817 / +4.9%	43691864 / 43942565 / -0.6%	0.63 / 0.60 / +5.5%
tpcds_q77/datafusion:vortex-file-compressed	45670032 / 46538385 / -1.9%	67287867 / 63390677 / +6.1%	0.68 / 0.73 / -7.5%
tpcds_q78/datafusion:vortex-file-compressed	68014445 / 68747413 / -1.1%	95977661 / 94445487 / +1.6%	0.71 / 0.73 / -2.6%
tpcds_q79/datafusion:vortex-file-compressed	19865698 / 19483699 / +2.0%	39844608 / 41799781 / -4.7%	0.50 / 0.47 / +7.0%
tpcds_q80/datafusion:vortex-file-compressed	74843712 / 74639438 / +0.3%	102898428 / 102819886 / +0.1%	0.73 / 0.73 / +0.2%
tpcds_q81/datafusion:vortex-file-compressed	27988844 / 27068568 / +3.4%	43732309 / 43682226 / +0.1%	0.64 / 0.62 / +3.3%
tpcds_q82/datafusion:vortex-file-compressed	12521616 / 12997562 / -3.7%	33682811 / 34116796 / -1.3%	0.37 / 0.38 / -2.4%
tpcds_q83/datafusion:vortex-file-compressed	35692714 / 35278853 / +1.2%	54619662 / 51166456 / +6.7%	0.65 / 0.69 / -5.2%
tpcds_q84/datafusion:vortex-file-compressed	12064044 / 11794177 / +2.3%	26675820 / 25396598 / +5.0%	0.45 / 0.46 / -2.6%
tpcds_q85/datafusion:vortex-file-compressed	57503837 / 57925560 / -0.7%	82877138 / 80348889 / +3.1%	0.69 / 0.72 / -3.8%
tpcds_q86/datafusion:vortex-file-compressed	16610660 / 16457530 / +0.9%	31230201 / 32469222 / -3.8%	0.53 / 0.51 / +4.9%
tpcds_q87/datafusion:vortex-file-compressed	30119396 / 29567438 / +1.9%	49951548 / 50905596 / -1.9%	0.60 / 0.58 / +3.8%
tpcds_q88/datafusion:vortex-file-compressed	32250763 / 32104016 / +0.5%	69416222 / 63612905 / +9.1%	0.46 / 0.50 / -7.9%
tpcds_q89/datafusion:vortex-file-compressed	20982158 / 20663336 / +1.5%	39996198 / 39249159 / +1.9%	0.52 / 0.53 / -0.4%
tpcds_q90/datafusion:vortex-file-compressed	10452082 / 10668205 / -2.0%	26258269 / 26904779 / -2.4%	0.40 / 0.40 / +0.4%
tpcds_q91/datafusion:vortex-file-compressed	16101880 / 15822708 / +1.8%	35731458 / 35674979 / +0.2%	0.45 / 0.44 / +1.6%
tpcds_q92/datafusion:vortex-file-compressed	13698268 / 13465226 / +1.7%	30293833 / 29450460 / +2.9%	0.45 / 0.46 / -1.1%
tpcds_q93/datafusion:vortex-file-compressed	22254886 / 22288306 / -0.1%	41344369 / 39311503 / +5.2%	0.54 / 0.57 / -5.1%
tpcds_q94/datafusion:vortex-file-compressed	19541344 / 18989217 / +2.9%	37559687 / 35884910 / +4.7%	0.52 / 0.53 / -1.7%
tpcds_q95/datafusion:vortex-file-compressed	45319710 / 45085632 / +0.5%	65134074 / 65843865 / -1.1%	0.70 / 0.68 / +1.6%
tpcds_q96/datafusion:vortex-file-compressed	8701148 / 8687084 / +0.2%	24461655 / 24920610 / -1.8%	0.36 / 0.35 / +2.0%
tpcds_q97/datafusion:vortex-file-compressed	20951752 / 21076740 / -0.6%	41693563 / 40801915 / +2.2%	0.50 / 0.52 / -2.7%
tpcds_q98/datafusion:vortex-file-compressed	17246014 / 17153452 / +0.5%	35330948 / 35728656 / -1.1%	0.49 / 0.48 / +1.7%
tpcds_q99/datafusion:vortex-file-compressed	20888712 / 20814182 / +0.4%	35618957 / 35554835 / +0.2%	0.59 / 0.59 / +0.2%

datafusion / parquet / ns (0.995x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpcds_q01/datafusion:parquet	43122274 / 43846422 / -1.7%	60709965 / 60005347 / +1.2%	0.71 / 0.73 / -2.8%
tpcds_q02/datafusion:parquet	54731755 / 55757277 / -1.8%	72901520 / 71357306 / +2.2%	0.75 / 0.78 / -3.9%
tpcds_q03/datafusion:parquet	18765100 / 19050948 / -1.5%	30330028 / 32322113 / -6.2%	0.62 / 0.59 / +5.0%
tpcds_q04/datafusion:parquet	181538452 / 180009476 / +0.8%	206122303 / 202731080 / +1.7%	0.88 / 0.89 / -0.8%
tpcds_q05/datafusion:parquet	67359654 / 67452735 / -0.1%	90033825 / 91252454 / -1.3%	0.75 / 0.74 / +1.2%
tpcds_q06/datafusion:parquet	33085440 / 33397300 / -0.9%	48944273 / 48649176 / +0.6%	0.68 / 0.69 / -1.5%
tpcds_q07/datafusion:parquet	46942410 / 46694988 / +0.5%	63351948 / 67932655 / -6.7%	0.74 / 0.69 / +7.8%
tpcds_q08/datafusion:parquet	30794636 / 31199608 / -1.3%	47132560 / 45345275 / +3.9%	0.65 / 0.69 / -5.0%
tpcds_q09/datafusion:parquet	31854138 / 32116022 / -0.8%	37512188 / 44213601 / -15.2% 🟢	0.85 / 0.73 / +16.9%
tpcds_q10/datafusion:parquet	43046291 / 44087352 / -2.4%	60643992 / 55032584 / +10.2% 🔴	0.71 / 0.80 / -11.4%
tpcds_q11/datafusion:parquet	97099779 / 99118944 / -2.0%	120445041 / 115581825 / +4.2%	0.81 / 0.86 / -6.0%
tpcds_q12/datafusion:parquet	23769012 / 23973628 / -0.9%	36189479 / 35973293 / +0.6%	0.66 / 0.67 / -1.4%
tpcds_q13/datafusion:parquet	46967030 / 46483706 / +1.0%	69059319 / 66333368 / +4.1%	0.68 / 0.70 / -2.9%
tpcds_q14/datafusion:parquet	536537559 / 538154738 / -0.3%	582731402 / 588539299 / -1.0%	0.92 / 0.91 / +0.7%
tpcds_q15/datafusion:parquet	23839817 / 23329618 / +2.2%	38328547 / 37781490 / +1.4%	0.62 / 0.62 / +0.7%
tpcds_q16/datafusion:parquet	36897586 / 36544121 / +1.0%	50640818 / 50619896 / +0.0%	0.73 / 0.72 / +0.9%
tpcds_q17/datafusion:parquet	73692464 / 75119042 / -1.9%	96385034 / 96474387 / -0.1%	0.76 / 0.78 / -1.8%
tpcds_q18/datafusion:parquet	69389275 / 66836378 / +3.8%	93468550 / 88630199 / +5.5%	0.74 / 0.75 / -1.6%
tpcds_q19/datafusion:parquet	29144616 / 29729388 / -2.0%	43657411 / 42819515 / +2.0%	0.67 / 0.69 / -3.8%
tpcds_q20/datafusion:parquet	25627355 / 24969518 / +2.6%	37690319 / 38866462 / -3.0%	0.68 / 0.64 / +5.8%
tpcds_q21/datafusion:parquet	20114950 / 20161090 / -0.2%	31776469 / 31800546 / -0.1%	0.63 / 0.63 / -0.2%
tpcds_q22/datafusion:parquet	77650293 / 77142012 / +0.7%	89131932 / 87858154 / +1.4%	0.87 / 0.88 / -0.8%
tpcds_q23/datafusion:parquet	92052668 / 91349462 / +0.8%	111421485 / 113181120 / -1.6%	0.83 / 0.81 / +2.4%
tpcds_q24/datafusion:parquet	90608282 / 92770976 / -2.3%	115712225 / 113350744 / +2.1%	0.78 / 0.82 / -4.3%
tpcds_q25/datafusion:parquet	81672518 / 80229011 / +1.8%	101450449 / 100842256 / +0.6%	0.81 / 0.80 / +1.2%
tpcds_q26/datafusion:parquet	42999488 / 43787389 / -1.8%	60762223 / 57659742 / +5.4%	0.71 / 0.76 / -6.8%
tpcds_q27/datafusion:parquet	75514446 / 75409834 / +0.1%	94205167 / 96101182 / -2.0%	0.80 / 0.78 / +2.2%
tpcds_q28/datafusion:parquet	28417208 / 28846260 / -1.5%	43553068 / 43554917 / -0.0%	0.65 / 0.66 / -1.5%
tpcds_q29/datafusion:parquet	74536804 / 74650711 / -0.2%	97750473 / 95427162 / +2.4%	0.76 / 0.78 / -2.5%
tpcds_q30/datafusion:parquet	46339880 / 47178713 / -1.8%	58928522 / 58551559 / +0.6%	0.79 / 0.81 / -2.4%
tpcds_q31/datafusion:parquet	79937916 / 80818462 / -1.1%	98005770 / 98873811 / -0.9%	0.82 / 0.82 / -0.2%
tpcds_q32/datafusion:parquet	26450072 / 26086708 / +1.4%	41634529 / 39508030 / +5.4%	0.64 / 0.66 / -3.8%
tpcds_q33/datafusion:parquet	36784062 / 38558912 / -4.6%	52610704 / 53023663 / -0.8%	0.70 / 0.73 / -3.9%
tpcds_q34/datafusion:parquet	24522102 / 24739372 / -0.9%	35480188 / 35909965 / -1.2%	0.69 / 0.69 / +0.3%
tpcds_q35/datafusion:parquet	49969137 / 51716168 / -3.4%	66001829 / 64607301 / +2.2%	0.76 / 0.80 / -5.4%
tpcds_q36/datafusion:parquet	50805662 / 53633372 / -5.3%	72959463 / 71365370 / +2.2%	0.70 / 0.75 / -7.3%
tpcds_q37/datafusion:parquet	21333390 / 21059308 / +1.3%	34803289 / 32948046 / +5.6%	0.61 / 0.64 / -4.1%
tpcds_q38/datafusion:parquet	44888200 / 44653446 / +0.5%	63865606 / 60746799 / +5.1%	0.70 / 0.74 / -4.4%
tpcds_q39/datafusion:parquet	67455412 / 67618510 / -0.2%	84404741 / 86536931 / -2.5%	0.80 / 0.78 / +2.3%
tpcds_q40/datafusion:parquet	29811554 / 30076230 / -0.9%	50319519 / 48872359 / +3.0%	0.59 / 0.62 / -3.7%
tpcds_q41/datafusion:parquet	20011630 / 19795827 / +1.1%	31714575 / 30811184 / +2.9%	0.63 / 0.64 / -1.8%
tpcds_q42/datafusion:parquet	18068536 / 18211920 / -0.8%	30298455 / 32111581 / -5.6%	0.60 / 0.57 / +5.1%
tpcds_q43/datafusion:parquet	19488348 / 19268661 / +1.1%	31562175 / 30985079 / +1.9%	0.62 / 0.62 / -0.7%
tpcds_q44/datafusion:parquet	35912020 / 36438808 / -1.4%	49704416 / 51954713 / -4.3%	0.72 / 0.70 / +3.0%
tpcds_q45/datafusion:parquet	28007480 / 27398212 / +2.2%	41115827 / 39109204 / +5.1%	0.68 / 0.70 / -2.8%
tpcds_q46/datafusion:parquet	28895813 / 29286111 / -1.3%	46185459 / 44982114 / +2.7%	0.63 / 0.65 / -3.9%
tpcds_q47/datafusion:parquet	97910254 / 98429680 / -0.5%	119202856 / 123516112 / -3.5%	0.82 / 0.80 / +3.1%
tpcds_q48/datafusion:parquet	42398953 / 41857703 / +1.3%	59890550 / 59999377 / -0.2%	0.71 / 0.70 / +1.5%
tpcds_q49/datafusion:parquet	72253965 / 72155906 / +0.1%	96096368 / 92353355 / +4.1%	0.75 / 0.78 / -3.8%
tpcds_q50/datafusion:parquet	47326836 / 47630214 / -0.6%	70411162 / 66493960 / +5.9%	0.67 / 0.72 / -6.2%
tpcds_q51/datafusion:parquet	75047358 / 75748952 / -0.9%	95827991 / 93780446 / +2.2%	0.78 / 0.81 / -3.0%
tpcds_q52/datafusion:parquet	18400886 / 18440540 / -0.2%	30729799 / 31499026 / -2.4%	0.60 / 0.59 / +2.3%
tpcds_q53/datafusion:parquet	22604204 / 23311745 / -3.0%	39101204 / 38476973 / +1.6%	0.58 / 0.61 / -4.6%
tpcds_q54/datafusion:parquet	53130748 / 52869332 / +0.5%	66883370 / 70753202 / -5.5%	0.79 / 0.75 / +6.3%
tpcds_q55/datafusion:parquet	18255374 / 18285992 / -0.2%	29672866 / 30486470 / -2.7%	0.62 / 0.60 / +2.6%
tpcds_q56/datafusion:parquet	41288987 / 40571934 / +1.8%	55175291 / 54021515 / +2.1%	0.75 / 0.75 / -0.4%
tpcds_q57/datafusion:parquet	101628016 / 103228732 / -1.6%	120959043 / 120073909 / +0.7%	0.84 / 0.86 / -2.3%
tpcds_q58/datafusion:parquet	65596535 / 65899958 / -0.5%	83391260 / 83882670 / -0.6%	0.79 / 0.79 / +0.1%
tpcds_q59/datafusion:parquet	48248306 / 48264130 / -0.0%	66090610 / 64543094 / +2.4%	0.73 / 0.75 / -2.4%
tpcds_q60/datafusion:parquet	40251446 / 39449744 / +2.0%	55313027 / 53105060 / +4.2%	0.73 / 0.74 / -2.0%
tpcds_q61/datafusion:parquet	36155258 / 37940298 / -4.7%	52835194 / 48420637 / +9.1%	0.68 / 0.78 / -12.7%
tpcds_q62/datafusion:parquet	25100457 / 24913400 / +0.8%	37578641 / 37514164 / +0.2%	0.67 / 0.66 / +0.6%
tpcds_q63/datafusion:parquet	22450176 / 22438395 / +0.1%	39179529 / 37774202 / +3.7%	0.57 / 0.59 / -3.5%
tpcds_q64/datafusion:parquet	289152880 / 292243279 / -1.1%	328094674 / 327971890 / +0.0%	0.88 / 0.89 / -1.1%
tpcds_q65/datafusion:parquet	121427228 / 122812174 / -1.1%	132407780 / 140616984 / -5.8%	0.92 / 0.87 / +5.0%
tpcds_q66/datafusion:parquet	59601298 / 60823197 / -2.0%	79220195 / 78250959 / +1.2%	0.75 / 0.78 / -3.2%
tpcds_q67/datafusion:parquet	68191144 / 69017602 / -1.2%	93295427 / 92414069 / +1.0%	0.73 / 0.75 / -2.1%
tpcds_q68/datafusion:parquet	29255714 / 29713764 / -1.5%	48796309 / 47088362 / +3.6%	0.60 / 0.63 / -5.0%
tpcds_q69/datafusion:parquet	43524280 / 43409056 / +0.3%	55392598 / 58228856 / -4.9%	0.79 / 0.75 / +5.4%
tpcds_q70/datafusion:parquet	80055928 / 80976967 / -1.1%	98352161 / 97646049 / +0.7%	0.81 / 0.83 / -1.8%
tpcds_q71/datafusion:parquet	25453088 / 26198760 / -2.8%	38546848 / 38817389 / -0.7%	0.66 / 0.67 / -2.2%
tpcds_q72/datafusion:parquet	524075677 / 520424787 / +0.7%	555217179 / 548824251 / +1.2%	0.94 / 0.95 / -0.5%
tpcds_q73/datafusion:parquet	23479464 / 23774588 / -1.2%	36106770 / 35994926 / +0.3%	0.65 / 0.66 / -1.5%
tpcds_q74/datafusion:parquet	69958746 / 71964559 / -2.8%	91905728 / 88593375 / +3.7%	0.76 / 0.81 / -6.3%
tpcds_q75/datafusion:parquet	152111575 / 155009072 / -1.9%	177993219 / 177967028 / +0.0%	0.85 / 0.87 / -1.9%
tpcds_q76/datafusion:parquet	34213314 / 33216860 / +3.0%	46541639 / 47889564 / -2.8%	0.74 / 0.69 / +6.0%
tpcds_q77/datafusion:parquet	57406387 / 57123894 / +0.5%	72548657 / 75735250 / -4.2%	0.79 / 0.75 / +4.9%
tpcds_q78/datafusion:parquet	85969347 / 85268329 / +0.8%	117323029 / 115062750 / +2.0%	0.73 / 0.74 / -1.1%
tpcds_q79/datafusion:parquet	27177558 / 27016716 / +0.6%	42922658 / 41406123 / +3.7%	0.63 / 0.65 / -3.0%
tpcds_q80/datafusion:parquet	81033820 / 80177621 / +1.1%	108540026 / 107359603 / +1.1%	0.75 / 0.75 / -0.0%
tpcds_q81/datafusion:parquet	38258438 / 36689468 / +4.3%	50525983 / 51731213 / -2.3%	0.76 / 0.71 / +6.8%
tpcds_q82/datafusion:parquet	21992970 / 22566992 / -2.5%	36416237 / 35373497 / +2.9%	0.60 / 0.64 / -5.3%
tpcds_q83/datafusion:parquet	58612768 / 58063186 / +0.9%	72090611 / 75073381 / -4.0%	0.81 / 0.77 / +5.1%
tpcds_q84/datafusion:parquet	25745566 / 26433304 / -2.6%	42240610 / 38836633 / +8.8%	0.61 / 0.68 / -10.5%
tpcds_q85/datafusion:parquet	81259810 / 80894374 / +0.5%	104035948 / 101416266 / +2.6%	0.78 / 0.80 / -2.1%
tpcds_q86/datafusion:parquet	20718298 / 20851596 / -0.6%	33238157 / 33115555 / +0.4%	0.62 / 0.63 / -1.0%
tpcds_q87/datafusion:parquet	45126260 / 44118442 / +2.3%	64269253 / 64350399 / -0.1%	0.70 / 0.69 / +2.4%
tpcds_q88/datafusion:parquet	37932268 / 37626844 / +0.8%	51889881 / 52415287 / -1.0%	0.73 / 0.72 / +1.8%
tpcds_q89/datafusion:parquet	25932154 / 26014490 / -0.3%	41544970 / 41199555 / +0.8%	0.62 / 0.63 / -1.1%
tpcds_q90/datafusion:parquet	14548414 / 15252398 / -4.6%	27518450 / 27255033 / +1.0%	0.53 / 0.56 / -5.5%
tpcds_q91/datafusion:parquet	31562374 / 32264378 / -2.2%	48220232 / 47253760 / +2.0%	0.65 / 0.68 / -4.1%
tpcds_q92/datafusion:parquet	26492787 / 27016796 / -1.9%	43617719 / 41389571 / +5.4%	0.61 / 0.65 / -6.9%
tpcds_q93/datafusion:parquet	26920136 / 27882475 / -3.5%	47368689 / 45762040 / +3.5%	0.57 / 0.61 / -6.7%
tpcds_q94/datafusion:parquet	34643391 / 35073608 / -1.2%	48508173 / 48864859 / -0.7%	0.71 / 0.72 / -0.5%
tpcds_q95/datafusion:parquet	64734618 / 63547685 / +1.9%	79208956 / 82201726 / -3.6%	0.82 / 0.77 / +5.7%
tpcds_q96/datafusion:parquet	12460749 / 12581218 / -1.0%	25101698 / 24073556 / +4.3%	0.50 / 0.52 / -5.0%
tpcds_q97/datafusion:parquet	35315365 / 35233132 / +0.2%	52052786 / 49607060 / +4.9%	0.68 / 0.71 / -4.5%
tpcds_q98/datafusion:parquet	24895042 / 25180659 / -1.1%	40584194 / 39568618 / +2.6%	0.61 / 0.64 / -3.6%
tpcds_q99/datafusion:parquet	29327538 / 29883085 / -1.9%	43770878 / 42943170 / +1.9%	0.67 / 0.70 / -3.7%

duckdb / vortex-file-compressed / ns (0.994x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpcds_q01/duckdb:vortex-file-compressed	17404511 / 17439644 / -0.2%	25572649 / 27219297 / -6.0%	0.68 / 0.64 / +6.2%
tpcds_q02/duckdb:vortex-file-compressed	16131546 / 16083914 / +0.3%	23267313 / 23502917 / -1.0%	0.69 / 0.68 / +1.3%
tpcds_q03/duckdb:vortex-file-compressed	7174012 / 7170842 / +0.0%	15730922 / 16666916 / -5.6%	0.46 / 0.43 / +6.0%
tpcds_q04/duckdb:vortex-file-compressed	48353047 / 48707472 / -0.7%	57582886 / 58047581 / -0.8%	0.84 / 0.84 / +0.1%
tpcds_q05/duckdb:vortex-file-compressed	23226534 / 23618420 / -1.7%	31708250 / 31794352 / -0.3%	0.73 / 0.74 / -1.4%
tpcds_q06/duckdb:vortex-file-compressed	28154117 / 28492640 / -1.2%	42126842 / 42541896 / -1.0%	0.67 / 0.67 / -0.2%
tpcds_q07/duckdb:vortex-file-compressed	14348598 / 14174695 / +1.2%	25325702 / 25347357 / -0.1%	0.57 / 0.56 / +1.3%
tpcds_q08/duckdb:vortex-file-compressed	18761526 / 18900235 / -0.7%	27301000 / 27718311 / -1.5%	0.69 / 0.68 / +0.8%
tpcds_q09/duckdb:vortex-file-compressed	18147568 / 18138864 / +0.0%	28605571 / 29221507 / -2.1%	0.63 / 0.62 / +2.2%
tpcds_q10/duckdb:vortex-file-compressed	33066213 / 33029630 / +0.1%	45987871 / 46829434 / -1.8%	0.72 / 0.71 / +1.9%
tpcds_q11/duckdb:vortex-file-compressed	39645610 / 39458654 / +0.5%	47785988 / 47656455 / +0.3%	0.83 / 0.83 / +0.2%
tpcds_q12/duckdb:vortex-file-compressed	10439530 / 10685926 / -2.3%	17733416 / 17944220 / -1.2%	0.59 / 0.60 / -1.1%
tpcds_q13/duckdb:vortex-file-compressed	21096500 / 21505498 / -1.9%	34220960 / 34571928 / -1.0%	0.62 / 0.62 / -0.9%
tpcds_q14/duckdb:vortex-file-compressed	58371658 / 59354654 / -1.7%	68593787 / 69010908 / -0.6%	0.85 / 0.86 / -1.1%
tpcds_q15/duckdb:vortex-file-compressed	27651018 / 27865054 / -0.8%	35414039 / 35466133 / -0.1%	0.78 / 0.79 / -0.6%
tpcds_q16/duckdb:vortex-file-compressed	36306352 / 36263682 / +0.1%	46711722 / 46410167 / +0.6%	0.78 / 0.78 / -0.5%
tpcds_q17/duckdb:vortex-file-compressed	26167291 / 26014247 / +0.6%	39962111 / 39653790 / +0.8%	0.65 / 0.66 / -0.2%
tpcds_q18/duckdb:vortex-file-compressed	32793346 / 32597166 / +0.6%	42992300 / 43159512 / -0.4%	0.76 / 0.76 / +1.0%
tpcds_q19/duckdb:vortex-file-compressed	89598144 / 89557922 / +0.0%	100843233 / 101850245 / -1.0%	0.89 / 0.88 / +1.0%
tpcds_q20/duckdb:vortex-file-compressed	11189797 / 11335046 / -1.3%	19147645 / 18866297 / +1.5%	0.58 / 0.60 / -2.7%
tpcds_q21/duckdb:vortex-file-compressed	10597598 / 10550480 / +0.4%	22654304 / 22047254 / +2.8%	0.47 / 0.48 / -2.2%
tpcds_q22/duckdb:vortex-file-compressed	18991054 / 18974138 / +0.1%	29219747 / 28581449 / +2.2%	0.65 / 0.66 / -2.1%
tpcds_q23/duckdb:vortex-file-compressed	39475751 / 39314652 / +0.4%	50396263 / 51213966 / -1.6%	0.78 / 0.77 / +2.0%
tpcds_q24/duckdb:vortex-file-compressed	28575868 / 28884934 / -1.1%	40476583 / 40024470 / +1.1%	0.71 / 0.72 / -2.2%
tpcds_q25/duckdb:vortex-file-compressed	24165326 / 24429660 / -1.1%	36496004 / 36379295 / +0.3%	0.66 / 0.67 / -1.4%
tpcds_q26/duckdb:vortex-file-compressed	13592394 / 13690386 / -0.7%	21358202 / 22307538 / -4.3%	0.64 / 0.61 / +3.7%
tpcds_q27/duckdb:vortex-file-compressed	17049474 / 17139150 / -0.5%	27899615 / 28790767 / -3.1%	0.61 / 0.60 / +2.7%
tpcds_q28/duckdb:vortex-file-compressed	15540588 / 15418094 / +0.8%	24249595 / 24049146 / +0.8%	0.64 / 0.64 / -0.0%
tpcds_q29/duckdb:vortex-file-compressed	24183800 / 24086816 / +0.4%	36504672 / 36218944 / +0.8%	0.66 / 0.67 / -0.4%
tpcds_q30/duckdb:vortex-file-compressed	16738330 / 16423360 / +1.9%	25623015 / 25162420 / +1.8%	0.65 / 0.65 / +0.1%
tpcds_q31/duckdb:vortex-file-compressed	19025982 / 19038258 / -0.1%	27887868 / 27773983 / +0.4%	0.68 / 0.69 / -0.5%
tpcds_q32/duckdb:vortex-file-compressed	13459838 / 13370864 / +0.7%	23069677 / 21994982 / +4.9%	0.58 / 0.61 / -4.0%
tpcds_q33/duckdb:vortex-file-compressed	17850380 / 18050817 / -1.1%	26129808 / 25375692 / +3.0%	0.68 / 0.71 / -4.0%
tpcds_q34/duckdb:vortex-file-compressed	12879598 / 12846442 / +0.3%	23140737 / 22710999 / +1.9%	0.56 / 0.57 / -1.6%
tpcds_q35/duckdb:vortex-file-compressed	29239876 / 29170485 / +0.2%	40387655 / 40616529 / -0.6%	0.72 / 0.72 / +0.8%
tpcds_q36/duckdb:vortex-file-compressed	15438031 / 15500460 / -0.4%	25467594 / 25928680 / -1.8%	0.61 / 0.60 / +1.4%
tpcds_q37/duckdb:vortex-file-compressed	9594712 / 9584120 / +0.1%	23587324 / 23057283 / +2.3%	0.41 / 0.42 / -2.1%
tpcds_q38/duckdb:vortex-file-compressed	25403988 / 25402442 / +0.0%	32327690 / 32979596 / -2.0%	0.79 / 0.77 / +2.0%
tpcds_q39/duckdb:vortex-file-compressed	20376152 / 21551021 / -5.5%	33247789 / 31761303 / +4.7%	0.61 / 0.68 / -9.7%
tpcds_q40/duckdb:vortex-file-compressed	14479482 / 14691824 / -1.4%	21954078 / 21942694 / +0.1%	0.66 / 0.67 / -1.5%
tpcds_q41/duckdb:vortex-file-compressed	15621592 / 15635824 / -0.1%	23857314 / 24539561 / -2.8%	0.65 / 0.64 / +2.8%
tpcds_q42/duckdb:vortex-file-compressed	7309903 / 7127577 / +2.6%	15479365 / 16028414 / -3.4%	0.47 / 0.44 / +6.2%
tpcds_q43/duckdb:vortex-file-compressed	10342268 / 10416310 / -0.7%	19199499 / 19702965 / -2.6%	0.54 / 0.53 / +1.9%
tpcds_q44/duckdb:vortex-file-compressed	14978118 / 15288050 / -2.0%	26480421 / 27312336 / -3.0%	0.57 / 0.56 / +1.1%
tpcds_q45/duckdb:vortex-file-compressed	25390806 / 25522246 / -0.5%	32705093 / 33381185 / -2.0%	0.78 / 0.76 / +1.5%
tpcds_q46/duckdb:vortex-file-compressed	20296122 / 20097704 / +1.0%	28901683 / 29393995 / -1.7%	0.70 / 0.68 / +2.7%
tpcds_q47/duckdb:vortex-file-compressed	36830337 / 36720506 / +0.3%	45946956 / 47230154 / -2.7%	0.80 / 0.78 / +3.1%
tpcds_q48/duckdb:vortex-file-compressed	24335514 / 24168216 / +0.7%	37363788 / 36654957 / +1.9%	0.65 / 0.66 / -1.2%
tpcds_q49/duckdb:vortex-file-compressed	20499200 / 20488396 / +0.1%	33447172 / 33177951 / +0.8%	0.61 / 0.62 / -0.8%
tpcds_q50/duckdb:vortex-file-compressed	17439118 / 17931386 / -2.7%	29288979 / 29638240 / -1.2%	0.60 / 0.61 / -1.6%
tpcds_q51/duckdb:vortex-file-compressed	51990104 / 52500616 / -1.0%	79052091 / 80923259 / -2.3%	0.66 / 0.65 / +1.4%
tpcds_q52/duckdb:vortex-file-compressed	7331032 / 7319667 / +0.2%	15771966 / 16698794 / -5.6%	0.46 / 0.44 / +6.0%
tpcds_q53/duckdb:vortex-file-compressed	12003345 / 12050183 / -0.4%	21708316 / 20427654 / +6.3%	0.55 / 0.59 / -6.3%
tpcds_q54/duckdb:vortex-file-compressed	20586837 / 20459934 / +0.6%	29539980 / 30721248 / -3.8%	0.70 / 0.67 / +4.6%
tpcds_q55/duckdb:vortex-file-compressed	7074685 / 7112698 / -0.5%	15444556 / 16245077 / -4.9%	0.46 / 0.44 / +4.6%
tpcds_q56/duckdb:vortex-file-compressed	18127792 / 18217444 / -0.5%	26556002 / 26742044 / -0.7%	0.68 / 0.68 / +0.2%
tpcds_q57/duckdb:vortex-file-compressed	27269762 / 27667744 / -1.4%	35424847 / 35748223 / -0.9%	0.77 / 0.77 / -0.5%
tpcds_q58/duckdb:vortex-file-compressed	20616182 / 20594944 / +0.1%	29308071 / 29427011 / -0.4%	0.70 / 0.70 / +0.5%
tpcds_q59/duckdb:vortex-file-compressed	20792138 / 20750972 / +0.2%	28438046 / 29050640 / -2.1%	0.73 / 0.71 / +2.4%
tpcds_q60/duckdb:vortex-file-compressed	18809594 / 19135727 / -1.7%	28485010 / 27636985 / +3.1%	0.66 / 0.69 / -4.6%
tpcds_q61/duckdb:vortex-file-compressed	24022872 / 24108312 / -0.4%	33513610 / 34077043 / -1.7%	0.72 / 0.71 / +1.3%
tpcds_q62/duckdb:vortex-file-compressed	12553318 / 12456039 / +0.8%	20108587 / 20304566 / -1.0%	0.62 / 0.61 / +1.8%
tpcds_q63/duckdb:vortex-file-compressed	11310577 / 11689796 / -3.2%	20474684 / 19906462 / +2.9%	0.55 / 0.59 / -5.9%
tpcds_q64/duckdb:vortex-file-compressed	51472976 / 51378027 / +0.2%	71103705 / 72381847 / -1.8%	0.72 / 0.71 / +2.0%
tpcds_q65/duckdb:vortex-file-compressed	15411203 / 16437374 / -6.2%	24669347 / 25748689 / -4.2%	0.62 / 0.64 / -2.1%
tpcds_q66/duckdb:vortex-file-compressed	30971708 / 33208336 / -6.7%	41144699 / 42336058 / -2.8%	0.75 / 0.78 / -4.0%
tpcds_q67/duckdb:vortex-file-compressed	64869358 / 65435933 / -0.9%	74457887 / 73848023 / +0.8%	0.87 / 0.89 / -1.7%
tpcds_q68/duckdb:vortex-file-compressed	19818416 / 20078928 / -1.3%	29118520 / 29170901 / -0.2%	0.68 / 0.69 / -1.1%
tpcds_q69/duckdb:vortex-file-compressed	23214264 / 23241534 / -0.1%	38785749 / 38850516 / -0.2%	0.60 / 0.60 / +0.0%
tpcds_q70/duckdb:vortex-file-compressed	16265244 / 16538370 / -1.7%	27731393 / 25211757 / +10.0%	0.59 / 0.66 / -10.6%
tpcds_q71/duckdb:vortex-file-compressed	13412551 / 13516254 / -0.8%	21897476 / 22743346 / -3.7%	0.61 / 0.59 / +3.1%
tpcds_q72/duckdb:vortex-file-compressed	58140911 / 58587628 / -0.8%	72945027 / 72788882 / +0.2%	0.80 / 0.80 / -1.0%
tpcds_q73/duckdb:vortex-file-compressed	11643922 / 11713432 / -0.6%	21626945 / 22329312 / -3.1%	0.54 / 0.52 / +2.6%
tpcds_q74/duckdb:vortex-file-compressed	27514647 / 27674992 / -0.6%	34843328 / 35033246 / -0.5%	0.79 / 0.79 / -0.0%
tpcds_q75/duckdb:vortex-file-compressed	28857836 / 29288680 / -1.5%	40086345 / 40611473 / -1.3%	0.72 / 0.72 / -0.2%
tpcds_q76/duckdb:vortex-file-compressed	12931946 / 13116972 / -1.4%	22419274 / 21974451 / +2.0%	0.58 / 0.60 / -3.4%
tpcds_q77/duckdb:vortex-file-compressed	20884846 / 21265524 / -1.8%	31791468 / 32589973 / -2.5%	0.66 / 0.65 / +0.7%
tpcds_q78/duckdb:vortex-file-compressed	43891219 / 44440038 / -1.2%	53151424 / 53650764 / -0.9%	0.83 / 0.83 / -0.3%
tpcds_q79/duckdb:vortex-file-compressed	13691468 / 13880707 / -1.4%	21756558 / 22258998 / -2.3%	0.63 / 0.62 / +0.9%
tpcds_q80/duckdb:vortex-file-compressed	38165928 / 37344568 / +2.2%	49341661 / 49823572 / -1.0%	0.77 / 0.75 / +3.2%
tpcds_q81/duckdb:vortex-file-compressed	22210950 / 22079180 / +0.6%	29527624 / 29931506 / -1.3%	0.75 / 0.74 / +2.0%
tpcds_q82/duckdb:vortex-file-compressed	9899054 / 10058612 / -1.6%	23500839 / 23175922 / +1.4%	0.42 / 0.43 / -2.9%
tpcds_q83/duckdb:vortex-file-compressed	17615406 / 17976734 / -2.0%	24638871 / 25007513 / -1.5%	0.71 / 0.72 / -0.5%
tpcds_q84/duckdb:vortex-file-compressed	15769548 / 16074242 / -1.9%	24191678 / 24634746 / -1.8%	0.65 / 0.65 / -0.1%
tpcds_q85/duckdb:vortex-file-compressed	25003022 / 25467392 / -1.8%	38182982 / 38371984 / -0.5%	0.65 / 0.66 / -1.3%
tpcds_q86/duckdb:vortex-file-compressed	10282238 / 10327764 / -0.4%	18317849 / 18271613 / +0.3%	0.56 / 0.57 / -0.7%
tpcds_q87/duckdb:vortex-file-compressed	27124155 / 27195582 / -0.3%	33603920 / 34082461 / -1.4%	0.81 / 0.80 / +1.2%
tpcds_q88/duckdb:vortex-file-compressed	38393330 / 38891858 / -1.3%	47562458 / 48534789 / -2.0%	0.81 / 0.80 / +0.7%
tpcds_q89/duckdb:vortex-file-compressed	12750930 / 13064133 / -2.4%	21833981 / 22099014 / -1.2%	0.58 / 0.59 / -1.2%
tpcds_q90/duckdb:vortex-file-compressed	8612808 / 8715810 / -1.2%	15381020 / 15807495 / -2.7%	0.56 / 0.55 / +1.6%
tpcds_q91/duckdb:vortex-file-compressed	14550112 / 14795734 / -1.7%	21591017 / 21648977 / -0.3%	0.67 / 0.68 / -1.4%
tpcds_q92/duckdb:vortex-file-compressed	13875932 / 14046564 / -1.2%	23217487 / 23426387 / -0.9%	0.60 / 0.60 / -0.3%
tpcds_q93/duckdb:vortex-file-compressed	11209942 / 11185930 / +0.2%	24701245 / 24912610 / -0.8%	0.45 / 0.45 / +1.1%
tpcds_q94/duckdb:vortex-file-compressed	32612539 / 32675963 / -0.2%	42087524 / 42873121 / -1.8%	0.77 / 0.76 / +1.7%
tpcds_q95/duckdb:vortex-file-compressed	43635252 / 43801560 / -0.4%	53368443 / 54021690 / -1.2%	0.82 / 0.81 / +0.8%
tpcds_q96/duckdb:vortex-file-compressed	6899549 / 6795696 / +1.5%	15160593 / 15466802 / -2.0%	0.46 / 0.44 / +3.6%
tpcds_q97/duckdb:vortex-file-compressed	23958687 / 24096507 / -0.6%	31022334 / 31722642 / -2.2%	0.77 / 0.76 / +1.7%
tpcds_q98/duckdb:vortex-file-compressed	11936775 / 11999761 / -0.5%	21159560 / 20860478 / +1.4%	0.56 / 0.58 / -1.9%
tpcds_q99/duckdb:vortex-file-compressed	15342880 / 15267530 / +0.5%	23531620 / 23326632 / +0.9%	0.65 / 0.65 / -0.4%

duckdb / parquet / ns (0.994x ➖, 0↑ 0↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpcds_q01/duckdb:parquet	27204633 / 27347317 / -0.5%	34959400 / 35303502 / -1.0%	0.78 / 0.77 / +0.5%
tpcds_q02/duckdb:parquet	19377860 / 19487888 / -0.6%	25869869 / 26543727 / -2.5%	0.75 / 0.73 / +2.0%
tpcds_q03/duckdb:parquet	11018104 / 11096145 / -0.7%	17796395 / 18150226 / -1.9%	0.62 / 0.61 / +1.3%
tpcds_q04/duckdb:parquet	131972152 / 131379761 / +0.5%	137768200 / 141158352 / -2.4%	0.96 / 0.93 / +2.9%
tpcds_q05/duckdb:parquet	31081359 / 30833834 / +0.8%	36447201 / 37028776 / -1.6%	0.85 / 0.83 / +2.4%
tpcds_q06/duckdb:parquet	32339862 / 32150086 / +0.6%	43108883 / 42644460 / +1.1%	0.75 / 0.75 / -0.5%
tpcds_q07/duckdb:parquet	32978242 / 33298061 / -1.0%	42116772 / 42603548 / -1.1%	0.78 / 0.78 / +0.2%
tpcds_q08/duckdb:parquet	22455378 / 22679677 / -1.0%	28452162 / 29862720 / -4.7%	0.79 / 0.76 / +3.9%
tpcds_q09/duckdb:parquet	28507844 / 29306652 / -2.7%	34010941 / 34056098 / -0.1%	0.84 / 0.86 / -2.6%
tpcds_q10/duckdb:parquet	41215024 / 41164660 / +0.1%	49446129 / 50308642 / -1.7%	0.83 / 0.82 / +1.9%
tpcds_q11/duckdb:parquet	74813478 / 75149775 / -0.4%	81245244 / 83075332 / -2.2%	0.92 / 0.90 / +1.8%
tpcds_q12/duckdb:parquet	16147020 / 16362310 / -1.3%	22023646 / 22513717 / -2.2%	0.73 / 0.73 / +0.9%
tpcds_q13/duckdb:parquet	34005784 / 33044183 / +2.9%	43286099 / 44182828 / -2.0%	0.79 / 0.75 / +5.0%
tpcds_q14/duckdb:parquet	70527760 / 71174578 / -0.9%	78028579 / 79471883 / -1.8%	0.90 / 0.90 / +0.9%
tpcds_q15/duckdb:parquet	18512554 / 18668916 / -0.8%	25815528 / 26114769 / -1.1%	0.72 / 0.71 / +0.3%
tpcds_q16/duckdb:parquet	43775804 / 44263705 / -1.1%	51123128 / 51154541 / -0.1%	0.86 / 0.87 / -1.0%
tpcds_q17/duckdb:parquet	37697273 / 37858576 / -0.4%	45742299 / 45437532 / +0.7%	0.82 / 0.83 / -1.1%
tpcds_q18/duckdb:parquet	51614556 / 51899668 / -0.5%	59095696 / 59726302 / -1.1%	0.87 / 0.87 / +0.5%
tpcds_q19/duckdb:parquet	20478813 / 20570085 / -0.4%	28090077 / 29074077 / -3.4%	0.73 / 0.71 / +3.0%
tpcds_q20/duckdb:parquet	17201504 / 17192398 / +0.1%	23238307 / 23405915 / -0.7%	0.74 / 0.73 / +0.8%
tpcds_q21/duckdb:parquet	12009552 / 12069398 / -0.5%	19256452 / 19193985 / +0.3%	0.62 / 0.63 / -0.8%
tpcds_q22/duckdb:parquet	22447210 / 22786454 / -1.5%	29438567 / 29371010 / +0.2%	0.76 / 0.78 / -1.7%
tpcds_q23/duckdb:parquet	46067616 / 46531502 / -1.0%	52227559 / 52932614 / -1.3%	0.88 / 0.88 / +0.3%
tpcds_q24/duckdb:parquet	33467704 / 33533062 / -0.2%	41294658 / 41847818 / -1.3%	0.81 / 0.80 / +1.1%
tpcds_q25/duckdb:parquet	33705246 / 34025074 / -0.9%	41715340 / 42410854 / -1.6%	0.81 / 0.80 / +0.7%
tpcds_q26/duckdb:parquet	31265500 / 31858504 / -1.9%	39167863 / 38845727 / +0.8%	0.80 / 0.82 / -2.7%
tpcds_q27/duckdb:parquet	36123361 / 35889354 / +0.7%	45565378 / 45815253 / -0.5%	0.79 / 0.78 / +1.2%
tpcds_q28/duckdb:parquet	28908190 / 30018104 / -3.7%	37237606 / 37590008 / -0.9%	0.78 / 0.80 / -2.8%
tpcds_q29/duckdb:parquet	35538666 / 35842821 / -0.8%	44210684 / 43915527 / +0.7%	0.80 / 0.82 / -1.5%
tpcds_q30/duckdb:parquet	40734312 / 40950778 / -0.5%	50479073 / 53761078 / -6.1%	0.81 / 0.76 / +5.9%
tpcds_q31/duckdb:parquet	38834650 / 38876682 / -0.1%	46901645 / 47252166 / -0.7%	0.83 / 0.82 / +0.6%
tpcds_q32/duckdb:parquet	17453800 / 17467670 / -0.1%	24665626 / 24704014 / -0.2%	0.71 / 0.71 / +0.1%
tpcds_q33/duckdb:parquet	24689911 / 25273202 / -2.3%	33066253 / 32285324 / +2.4%	0.75 / 0.78 / -4.6%
tpcds_q34/duckdb:parquet	18465751 / 18557908 / -0.5%	26330146 / 26435074 / -0.4%	0.70 / 0.70 / -0.1%
tpcds_q35/duckdb:parquet	39670731 / 39911215 / -0.6%	49652748 / 48425176 / +2.5%	0.80 / 0.82 / -3.1%
tpcds_q36/duckdb:parquet	42538156 / 42703088 / -0.4%	52057353 / 52222944 / -0.3%	0.82 / 0.82 / -0.1%
tpcds_q37/duckdb:parquet	14681338 / 14975561 / -2.0%	22420762 / 22350433 / +0.3%	0.65 / 0.67 / -2.3%
tpcds_q38/duckdb:parquet	32230516 / 32298115 / -0.2%	37178194 / 37193818 / -0.0%	0.87 / 0.87 / -0.2%
tpcds_q39/duckdb:parquet	25169127 / 25590262 / -1.6%	33201616 / 33087117 / +0.3%	0.76 / 0.77 / -2.0%
tpcds_q40/duckdb:parquet	18048562 / 18153762 / -0.6%	24575901 / 24397925 / +0.7%	0.73 / 0.74 / -1.3%
tpcds_q41/duckdb:parquet	14118046 / 14246674 / -0.9%	21925581 / 21943308 / -0.1%	0.64 / 0.65 / -0.8%
tpcds_q42/duckdb:parquet	10475015 / 10579660 / -1.0%	17664764 / 17727575 / -0.4%	0.59 / 0.60 / -0.6%
tpcds_q43/duckdb:parquet	13587414 / 13619888 / -0.2%	21116878 / 20808750 / +1.5%	0.64 / 0.65 / -1.7%
tpcds_q44/duckdb:parquet	23190675 / 24287447 / -4.5%	30732627 / 28739059 / +6.9%	0.75 / 0.85 / -10.7%
tpcds_q45/duckdb:parquet	23429548 / 23696610 / -1.1%	30994657 / 31357422 / -1.2%	0.76 / 0.76 / +0.0%
tpcds_q46/duckdb:parquet	38548740 / 38500332 / +0.1%	47293501 / 47718020 / -0.9%	0.82 / 0.81 / +1.0%
tpcds_q47/duckdb:parquet	37378397 / 37871174 / -1.3%	45151303 / 46527552 / -3.0%	0.83 / 0.81 / +1.7%
tpcds_q48/duckdb:parquet	30549300 / 30891434 / -1.1%	40461243 / 40608555 / -0.4%	0.76 / 0.76 / -0.7%
tpcds_q49/duckdb:parquet	32186469 / 31652608 / +1.7%	40499848 / 40689618 / -0.5%	0.79 / 0.78 / +2.2%
tpcds_q50/duckdb:parquet	21511290 / 21500242 / +0.1%	29485934 / 29406199 / +0.3%	0.73 / 0.73 / -0.2%
tpcds_q51/duckdb:parquet	55946892 / 56409870 / -0.8%	63210141 / 64688708 / -2.3%	0.89 / 0.87 / +1.5%
tpcds_q52/duckdb:parquet	10668042 / 10763580 / -0.9%	18088328 / 18112062 / -0.1%	0.59 / 0.59 / -0.8%
tpcds_q53/duckdb:parquet	15463727 / 15476084 / -0.1%	23149291 / 23446814 / -1.3%	0.67 / 0.66 / +1.2%
tpcds_q54/duckdb:parquet	27319394 / 27484921 / -0.6%	35635495 / 35644381 / -0.0%	0.77 / 0.77 / -0.6%
tpcds_q55/duckdb:parquet	10380050 / 10666912 / -2.7%	17799885 / 17759741 / +0.2%	0.58 / 0.60 / -2.9%
tpcds_q56/duckdb:parquet	26871574 / 26985425 / -0.4%	33910971 / 34252694 / -1.0%	0.79 / 0.79 / +0.6%
tpcds_q57/duckdb:parquet	33737474 / 33679152 / +0.2%	40095136 / 40326070 / -0.6%	0.84 / 0.84 / +0.8%
tpcds_q58/duckdb:parquet	26322729 / 26536364 / -0.8%	33488186 / 33483367 / +0.0%	0.79 / 0.79 / -0.8%
tpcds_q59/duckdb:parquet	24953730 / 25114642 / -0.6%	32219599 / 32525005 / -0.9%	0.77 / 0.77 / +0.3%
tpcds_q60/duckdb:parquet	28327502 / 28401478 / -0.3%	35474202 / 35608805 / -0.4%	0.80 / 0.80 / +0.1%
tpcds_q61/duckdb:parquet	26385498 / 27084780 / -2.6%	31860945 / 32072144 / -0.7%	0.83 / 0.84 / -1.9%
tpcds_q62/duckdb:parquet	14372380 / 14452104 / -0.6%	20358953 / 20694763 / -1.6%	0.71 / 0.70 / +1.1%
tpcds_q63/duckdb:parquet	14718138 / 14807718 / -0.6%	22283964 / 22419485 / -0.6%	0.66 / 0.66 / -0.0%
tpcds_q64/duckdb:parquet	77443932 / 77044416 / +0.5%	88426432 / 88989658 / -0.6%	0.88 / 0.87 / +1.2%
tpcds_q65/duckdb:parquet	21623700 / 21999061 / -1.7%	27386487 / 27453346 / -0.2%	0.79 / 0.80 / -1.5%
tpcds_q66/duckdb:parquet	39987762 / 40164515 / -0.4%	48149102 / 48418907 / -0.6%	0.83 / 0.83 / +0.1%
tpcds_q67/duckdb:parquet	67556590 / 67849475 / -0.4%	76912417 / 76309207 / +0.8%	0.88 / 0.89 / -1.2%
tpcds_q68/duckdb:parquet	30466498 / 30772986 / -1.0%	40173019 / 39496453 / +1.7%	0.76 / 0.78 / -2.7%
tpcds_q69/duckdb:parquet	28708868 / 28968358 / -0.9%	38092903 / 38537349 / -1.2%	0.75 / 0.75 / +0.3%
tpcds_q70/duckdb:parquet	21380098 / 21713808 / -1.5%	29845245 / 29551750 / +1.0%	0.72 / 0.73 / -2.5%
tpcds_q71/duckdb:parquet	18748244 / 18364015 / +2.1%	26170219 / 26083380 / +0.3%	0.72 / 0.70 / +1.8%
tpcds_q72/duckdb:parquet	60706188 / 60419342 / +0.5%	68641055 / 70373632 / -2.5%	0.88 / 0.86 / +3.0%
tpcds_q73/duckdb:parquet	15949272 / 15972982 / -0.1%	24241942 / 24157600 / +0.3%	0.66 / 0.66 / -0.5%
tpcds_q74/duckdb:parquet	93473558 / 92831064 / +0.7%	101539143 / 101609503 / -0.1%	0.92 / 0.91 / +0.8%
tpcds_q75/duckdb:parquet	47330510 / 46978390 / +0.7%	56223583 / 56255548 / -0.1%	0.84 / 0.84 / +0.8%
tpcds_q76/duckdb:parquet	19889589 / 20723146 / -4.0%	25188122 / 25748937 / -2.2%	0.79 / 0.80 / -1.9%
tpcds_q77/duckdb:parquet	26753238 / 26867437 / -0.4%	33738826 / 34106605 / -1.1%	0.79 / 0.79 / +0.7%
tpcds_q78/duckdb:parquet	81781919 / 82669150 / -1.1%	89710073 / 90493928 / -0.9%	0.91 / 0.91 / -0.2%
tpcds_q79/duckdb:parquet	23205682 / 23370406 / -0.7%	32034233 / 31826809 / +0.7%	0.72 / 0.73 / -1.3%
tpcds_q80/duckdb:parquet	41457538 / 39822459 / +4.1%	47469111 / 47482389 / -0.0%	0.87 / 0.84 / +4.1%
tpcds_q81/duckdb:parquet	40762924 / 41480985 / -1.7%	50397464 / 50605131 / -0.4%	0.81 / 0.82 / -1.3%
tpcds_q82/duckdb:parquet	15597605 / 15761252 / -1.0%	23414340 / 23587700 / -0.7%	0.67 / 0.67 / -0.3%
tpcds_q83/duckdb:parquet	21980659 / 22334745 / -1.6%	28519304 / 28123425 / +1.4%	0.77 / 0.79 / -3.0%
tpcds_q84/duckdb:parquet	24601234 / 24533518 / +0.3%	31257904 / 31456642 / -0.6%	0.79 / 0.78 / +0.9%
tpcds_q85/duckdb:parquet	35480059 / 35836572 / -1.0%	43878658 / 44209113 / -0.7%	0.81 / 0.81 / -0.2%
tpcds_q86/duckdb:parquet	13590028 / 13810191 / -1.6%	19960152 / 19817337 / +0.7%	0.68 / 0.70 / -2.3%
tpcds_q87/duckdb:parquet	34295591 / 34206630 / +0.3%	39965619 / 39386514 / +1.5%	0.86 / 0.87 / -1.2%
tpcds_q88/duckdb:parquet	43149835 / 43280864 / -0.3%	48546313 / 54118964 / -10.3% 🟢	0.89 / 0.80 / +11.1%
tpcds_q89/duckdb:parquet	15579984 / 15691137 / -0.7%	23622864 / 23659831 / -0.2%	0.66 / 0.66 / -0.6%
tpcds_q90/duckdb:parquet	10876534 / 10859698 / +0.2%	17862991 / 17619035 / +1.4%	0.61 / 0.62 / -1.2%
tpcds_q91/duckdb:parquet	23159430 / 23136540 / +0.1%	32204928 / 31639361 / +1.8%	0.72 / 0.73 / -1.7%
tpcds_q92/duckdb:parquet	18784124 / 19125308 / -1.8%	26512067 / 26458391 / +0.2%	0.71 / 0.72 / -2.0%
tpcds_q93/duckdb:parquet	15423936 / 15483261 / -0.4%	23138695 / 23411350 / -1.2%	0.67 / 0.66 / +0.8%
tpcds_q94/duckdb:parquet	37746878 / 37727360 / +0.1%	45876421 / 45479438 / +0.9%	0.82 / 0.83 / -0.8%
tpcds_q95/duckdb:parquet	48366638 / 47991017 / +0.8%	53615029 / 54105684 / -0.9%	0.90 / 0.89 / +1.7%
tpcds_q96/duckdb:parquet	9015411 / 9126438 / -1.2%	15498478 / 15644767 / -0.9%	0.58 / 0.58 / -0.3%
tpcds_q97/duckdb:parquet	29220275 / 29273749 / -0.2%	36585504 / 37201939 / -1.7%	0.80 / 0.79 / +1.5%
tpcds_q98/duckdb:parquet	18918603 / 18846741 / +0.4%	26622867 / 26462560 / +0.6%	0.71 / 0.71 / -0.2%
tpcds_q99/duckdb:parquet	20578274 / 20298226 / +1.4%	26496075 / 26502711 / -0.0%	0.78 / 0.77 / +1.4%

No file size changes detected.

The stream a scan returns now owns the planning run and drives it as it is polled. Every CPU step of the run, the planners' and morsels' compute, is a `spawn_cpu` task; every read is already a `spawn_io` task of the segment source; the stream's poll awaits one or the other and does no scan work on the engine's thread. The scan advances only as fast as its consumer pulls, so the task and channel that ran ahead of it go. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

A prune planner no longer knows about chunks or selections. It is built with the planner it continues into, as a function of the rows its proofs keep, and emits that planner with the kept mask once every proof has run. `PreparedScan::planner` supplies the chunk planner. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

github-actions · 2026-10-22T12:39:44Z

Benchmarks: TPC-H SF=1 on S3 📖

Commits: PR da8a7c8a vs base 31027fc2
Verdict: No clear signal (environment too noisy confidence)
Attributed Vortex impact: -9.1%
Engines: DataFusion No clear signal (-25.5%, environment too noisy confidence) · DuckDB No clear signal (+10.9%, environment too noisy confidence)
Vortex (geomean): hot 0.964x ➖ · cold 0.993x ➖
Parquet (geomean): hot 1.061x ➖ · cold 1.090x ➖
Shifts: Parquet (control) +6.1% · Median polish +3.9%

How to read Verdict and Engines

Verdict: Overall PR-level signal after subtracting baseline drift estimated from Parquet control rows. It can be Likely improvement, Likely regression, or No clear signal.
Engines: Per-engine attribution. DataFusion is compared against DataFusion/Parquet controls; DuckDB is compared against DuckDB/Parquet controls. This answers whether each engine improved or regressed independently.
Confidence: Based on directional consistency, share of rows above the noise floor, and control-run noise.
Hot vs cold: Every measurement is run several times. The first run is reported as the cold run, and the median of the runs after it is reported as the hot run. The verdict and significance use hot runs; each target's geomean reports hot and cold beside each other where the individual runs were recorded and hot alone where they were not, the cold column shows first-run cost per row, and hot/cold is how much of each run the warm path saves. Rows whose results predate per-run reporting show only one value, taken from the value the runner reported.
Table cells: Each cell reads PR / base / %diff. The hot and cold columns mark a change as 🔴 slower or 🟢 faster once it clears this suite's threshold; anything smaller is noise here and is left unmarked, as is hot/cold, because a shift in the warm-up ratio is not a win or a loss by itself.

datafusion / vortex-file-compressed / ns (0.799x ➖, 5↑ 1↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/datafusion:vortex-file-compressed	250811611 / 267707148 / -6.3%	756460408 / 1286047577 / -41.2% 🟢	0.33 / 0.21 / +59.3%
tpch_q02/datafusion:vortex-file-compressed	326318286 / 416827196 / -21.7%	767744471 / 964511425 / -20.4%	0.43 / 0.43 / -1.6%
tpch_q03/datafusion:vortex-file-compressed	392678427 / 448775125 / -12.5%	1197139491 / 1484291328 / -19.3%	0.33 / 0.30 / +8.5%
tpch_q04/datafusion:vortex-file-compressed	207035886 / 242301955 / -14.6%	451726946 / 695566848 / -35.1% 🟢	0.46 / 0.35 / +31.6%
tpch_q05/datafusion:vortex-file-compressed	770077008 / 446320311 / +72.5% 🔴	768862692 / 696805981 / +10.3%	1.00 / 0.64 / +56.4%
tpch_q06/datafusion:vortex-file-compressed	160082699 / 428614632 / -62.7% 🟢	416194959 / 494628790 / -15.9%	0.38 / 0.87 / -55.6%
tpch_q07/datafusion:vortex-file-compressed	400768840 / 435168624 / -7.9%	569216288 / 545378043 / +4.4%	0.70 / 0.80 / -11.8%
tpch_q08/datafusion:vortex-file-compressed	458939171 / 521383232 / -12.0%	684200509 / 663032925 / +3.2%	0.67 / 0.79 / -14.7%
tpch_q09/datafusion:vortex-file-compressed	295343816 / 487434394 / -39.4% 🟢	515984725 / 646172232 / -20.1%	0.57 / 0.75 / -24.1%
tpch_q10/datafusion:vortex-file-compressed	352139369 / 416314867 / -15.4%	539446922 / 854774764 / -36.9% 🟢	0.65 / 0.49 / +34.0%
tpch_q11/datafusion:vortex-file-compressed	355124475 / 364692198 / -2.6%	526227292 / 478470177 / +10.0%	0.67 / 0.76 / -11.5%
tpch_q12/datafusion:vortex-file-compressed	255281112 / 752496903 / -66.1% 🟢	769731253 / 1208065625 / -36.3% 🟢	0.33 / 0.62 / -46.8%
tpch_q13/datafusion:vortex-file-compressed	130488924 / 147340225 / -11.4%	409168189 / 350492133 / +16.7%	0.32 / 0.42 / -24.1%
tpch_q14/datafusion:vortex-file-compressed	175820888 / 197745573 / -11.1%	310726718 / 362499493 / -14.3%	0.57 / 0.55 / +3.7%
tpch_q15/datafusion:vortex-file-compressed	343137747 / 342774677 / +0.1%	561033856 / 612815149 / -8.4%	0.61 / 0.56 / +9.3%
tpch_q16/datafusion:vortex-file-compressed	134095123 / 156820251 / -14.5%	267453483 / 272808337 / -2.0%	0.50 / 0.57 / -12.8%
tpch_q17/datafusion:vortex-file-compressed	282066308 / 637566919 / -55.8% 🟢	427577458 / 696135197 / -38.6% 🟢	0.66 / 0.92 / -28.0%
tpch_q18/datafusion:vortex-file-compressed	288146693 / 271840370 / +6.0%	881960862 / 462638470 / +90.6% 🔴	0.33 / 0.59 / -44.4%
tpch_q19/datafusion:vortex-file-compressed	275593142 / 456263692 / -39.6% 🟢	474653521 / 554818440 / -14.4%	0.58 / 0.82 / -29.4%
tpch_q20/datafusion:vortex-file-compressed	294683080 / 304001086 / -3.1%	481040737 / 536483639 / -10.3%	0.61 / 0.57 / +8.1%
tpch_q21/datafusion:vortex-file-compressed	526848816 / 593234606 / -11.2%	700437150 / 838985520 / -16.5%	0.75 / 0.71 / +6.4%
tpch_q22/datafusion:vortex-file-compressed	176920033 / 196364260 / -9.9%	435421017 / 380134121 / +14.5%	0.41 / 0.52 / -21.3%

datafusion / parquet / ns (1.073x ➖, 0↑ 3↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/datafusion:parquet	107529052 / 108985789 / -1.3%	397709406 / 388303497 / +2.4%	0.27 / 0.28 / -3.7%
tpch_q02/datafusion:parquet	5544604413 / 5615566732 / -1.3%	5643042724 / 6272268641 / -10.0%	0.98 / 0.90 / +9.7%
tpch_q03/datafusion:parquet	594633926 / 315366453 / +88.6% 🔴	908565638 / 676120001 / +34.4% 🔴	0.65 / 0.47 / +40.3%
tpch_q04/datafusion:parquet	244939115 / 219699700 / +11.5%	523596309 / 385841975 / +35.7% 🔴	0.47 / 0.57 / -17.8%
tpch_q05/datafusion:parquet	3540493612 / 2975299663 / +19.0%	3306462282 / 3405014240 / -2.9%	1.07 / 0.87 / +22.5%
tpch_q06/datafusion:parquet	183555250 / 116087462 / +58.1% 🔴	502796761 / 230485772 / +118.1% 🔴	0.37 / 0.50 / -27.5%
tpch_q07/datafusion:parquet	3151030818 / 3652918284 / -13.7%	3220031576 / 3129960199 / +2.9%	0.98 / 1.17 / -16.2%
tpch_q08/datafusion:parquet	3039289423 / 3261344197 / -6.8%	3085969359 / 3460035446 / -10.8%	0.98 / 0.94 / +4.5%
tpch_q09/datafusion:parquet	3053369354 / 3193894735 / -4.4%	3129167861 / 3165740637 / -1.2%	0.98 / 1.01 / -3.3%
tpch_q10/datafusion:parquet	292263674 / 311770813 / -6.3%	501339826 / 457433990 / +9.6%	0.58 / 0.68 / -14.5%
tpch_q11/datafusion:parquet	5478314943 / 5485462724 / -0.1%	6208784111 / 5439356571 / +14.1%	0.88 / 1.01 / -12.5%
tpch_q12/datafusion:parquet	386149722 / 219515797 / +75.9% 🔴	696054649 / 353540494 / +96.9% 🔴	0.55 / 0.62 / -10.7%
tpch_q13/datafusion:parquet	224900691 / 179562615 / +25.2%	335310528 / 440611745 / -23.9%	0.67 / 0.41 / +64.6%
tpch_q14/datafusion:parquet	198993340 / 195907293 / +1.6%	328020083 / 379968760 / -13.7%	0.61 / 0.52 / +17.7%
tpch_q15/datafusion:parquet	2850109751 / 2912710831 / -2.1%	2973356102 / 3553580199 / -16.3%	0.96 / 0.82 / +16.9%
tpch_q16/datafusion:parquet	2679453541 / 2800592128 / -4.3%	2871737942 / 2731497775 / +5.1%	0.93 / 1.03 / -9.0%
tpch_q17/datafusion:parquet	286884107 / 329983591 / -13.1%	545232244 / 447417787 / +21.9%	0.53 / 0.74 / -28.7%
tpch_q18/datafusion:parquet	351825026 / 396053617 / -11.2%	563596600 / 553482367 / +1.8%	0.62 / 0.72 / -12.8%
tpch_q19/datafusion:parquet	259840968 / 224212601 / +15.9%	472415277 / 304969421 / +54.9% 🔴	0.55 / 0.74 / -25.2%
tpch_q20/datafusion:parquet	2881780011 / 2874525016 / +0.3%	3217376673 / 3029607551 / +6.2%	0.90 / 0.95 / -5.6%
tpch_q21/datafusion:parquet	3089403965 / 2896165949 / +6.7%	3064538560 / 3788385940 / -19.1%	1.01 / 0.76 / +31.9%
tpch_q22/datafusion:parquet	424691883 / 502695029 / -15.5%	446627805 / 647722641 / -31.0% 🟢	0.95 / 0.78 / +22.5%

duckdb / vortex-file-compressed / ns (1.162x ➖, 0↑ 4↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/duckdb:vortex-file-compressed	198335353 / 212170199 / -6.5%	287664222 / 417486268 / -31.1% 🟢	0.69 / 0.51 / +35.7%
tpch_q02/duckdb:vortex-file-compressed	228166567 / 225704631 / +1.1%	295853494 / 255415474 / +15.8%	0.77 / 0.88 / -12.7%
tpch_q03/duckdb:vortex-file-compressed	551774395 / 400479101 / +37.8% 🔴	665824406 / 429449138 / +55.0% 🔴	0.83 / 0.93 / -11.1%
tpch_q04/duckdb:vortex-file-compressed	215792641 / 208866401 / +3.3%	294596627 / 275077405 / +7.1%	0.73 / 0.76 / -3.5%
tpch_q05/duckdb:vortex-file-compressed	829739910 / 431572870 / +92.3% 🔴	578650498 / 498909098 / +16.0%	1.43 / 0.87 / +65.8%
tpch_q06/duckdb:vortex-file-compressed	325937072 / 321586325 / +1.4%	323690007 / 311084295 / +4.1%	1.01 / 1.03 / -2.6%
tpch_q07/duckdb:vortex-file-compressed	381860780 / 312826181 / +22.1%	518370792 / 352106299 / +47.2% 🔴	0.74 / 0.89 / -17.1%
tpch_q08/duckdb:vortex-file-compressed	503985625 / 468481423 / +7.6%	534961054 / 503729645 / +6.2%	0.94 / 0.93 / +1.3%
tpch_q09/duckdb:vortex-file-compressed	411397993 / 354663942 / +16.0%	461306614 / 393317006 / +17.3%	0.89 / 0.90 / -1.1%
tpch_q10/duckdb:vortex-file-compressed	460634753 / 423180082 / +8.9%	559052921 / 494454690 / +13.1%	0.82 / 0.86 / -3.7%
tpch_q11/duckdb:vortex-file-compressed	192684684 / 172615825 / +11.6%	217031492 / 210317145 / +3.2%	0.89 / 0.82 / +8.2%
tpch_q12/duckdb:vortex-file-compressed	679360401 / 437263266 / +55.4% 🔴	642922208 / 581248292 / +10.6%	1.06 / 0.75 / +40.5%
tpch_q13/duckdb:vortex-file-compressed	306125800 / 272711896 / +12.3%	275428982 / 221427711 / +24.4%	1.11 / 1.23 / -9.8%
tpch_q14/duckdb:vortex-file-compressed	306371290 / 197200811 / +55.4% 🔴	243234046 / 298794927 / -18.6%	1.26 / 0.66 / +90.8%
tpch_q15/duckdb:vortex-file-compressed	188929609 / 187068995 / +1.0%	282142504 / 218923011 / +28.9%	0.67 / 0.85 / -21.6%
tpch_q16/duckdb:vortex-file-compressed	138533541 / 129667478 / +6.8%	161675155 / 165462258 / -2.3%	0.86 / 0.78 / +9.3%
tpch_q17/duckdb:vortex-file-compressed	370634211 / 384689541 / -3.7%	377361031 / 380988777 / -1.0%	0.98 / 1.01 / -2.7%
tpch_q18/duckdb:vortex-file-compressed	327761026 / 297389355 / +10.2%	498363052 / 334448442 / +49.0% 🔴	0.66 / 0.89 / -26.0%
tpch_q19/duckdb:vortex-file-compressed	457815859 / 364134618 / +25.7%	452614335 / 414685202 / +9.1%	1.01 / 0.88 / +15.2%
tpch_q20/duckdb:vortex-file-compressed	405165317 / 370054109 / +9.5%	513564530 / 451797355 / +13.7%	0.79 / 0.82 / -3.7%
tpch_q21/duckdb:vortex-file-compressed	712864255 / 561206678 / +27.0%	670860586 / 563983743 / +19.0%	1.06 / 1.00 / +6.8%
tpch_q22/duckdb:vortex-file-compressed	165928720 / 160119572 / +3.6%	200903114 / 191379818 / +5.0%	0.83 / 0.84 / -1.3%

duckdb / parquet / ns (1.048x ➖, 1↑ 3↓)

name	hot ns (PR / base / %diff)	cold ns (PR / base / %diff)	hot/cold (PR / base / %diff)
tpch_q01/duckdb:parquet	187873910 / 138502921 / +35.6% 🔴	453611890 / 177305776 / +155.8% 🔴	0.41 / 0.78 / -47.0%
tpch_q02/duckdb:parquet	416448005 / 426957352 / -2.5%	434579589 / 534202633 / -18.6%	0.96 / 0.80 / +19.9%
tpch_q03/duckdb:parquet	536479549 / 475792397 / +12.8%	562259967 / 552872261 / +1.7%	0.95 / 0.86 / +10.9%
tpch_q04/duckdb:parquet	354212173 / 310016445 / +14.3%	392083406 / 389724946 / +0.6%	0.90 / 0.80 / +13.6%
tpch_q05/duckdb:parquet	680852476 / 621459893 / +9.6%	622923985 / 646958910 / -3.7%	1.09 / 0.96 / +13.8%
tpch_q06/duckdb:parquet	175361004 / 190001524 / -7.7%	220123026 / 254943682 / -13.7%	0.80 / 0.75 / +6.9%
tpch_q07/duckdb:parquet	561122758 / 556321217 / +0.9%	593243520 / 527569475 / +12.4%	0.95 / 1.05 / -10.3%
tpch_q08/duckdb:parquet	718338868 / 681392973 / +5.4%	779217797 / 800967629 / -2.7%	0.92 / 0.85 / +8.4%
tpch_q09/duckdb:parquet	628011370 / 672271671 / -6.6%	674453368 / 659438316 / +2.3%	0.93 / 1.02 / -8.7%
tpch_q10/duckdb:parquet	624373444 / 557699757 / +12.0%	788746233 / 614578057 / +28.3%	0.79 / 0.91 / -12.8%
tpch_q11/duckdb:parquet	253663510 / 298371800 / -15.0%	330413826 / 1266956609 / -73.9% 🟢	0.77 / 0.24 / +226.0%
tpch_q12/duckdb:parquet	538276446 / 351495069 / +53.1% 🔴	1449773788 / 484645496 / +199.1% 🔴	0.37 / 0.73 / -48.8%
tpch_q13/duckdb:parquet	356085388 / 332590739 / +7.1%	437327422 / 348579089 / +25.5%	0.81 / 0.95 / -14.7%
tpch_q14/duckdb:parquet	359448109 / 362923328 / -1.0%	451138856 / 445643531 / +1.2%	0.80 / 0.81 / -2.2%
tpch_q15/duckdb:parquet	263797906 / 231548955 / +13.9%	305396270 / 251613746 / +21.4%	0.86 / 0.92 / -6.1%
tpch_q16/duckdb:parquet	358711045 / 615657818 / -41.7% 🟢	425803998 / 1569573083 / -72.9% 🟢	0.84 / 0.39 / +114.8%
tpch_q17/duckdb:parquet	362805762 / 413544607 / -12.3%	423314434 / 537406489 / -21.2%	0.86 / 0.77 / +11.4%
tpch_q18/duckdb:parquet	478889311 / 425986187 / +12.4%	483283919 / 468945283 / +3.1%	0.99 / 0.91 / +9.1%
tpch_q19/duckdb:parquet	382250184 / 344498828 / +11.0%	1921578652 / 333092103 / +476.9% 🔴	0.20 / 1.03 / -80.8%
tpch_q20/duckdb:parquet	558586836 / 514154054 / +8.6%	545272882 / 531220937 / +2.6%	1.02 / 0.97 / +5.8%
tpch_q21/duckdb:parquet	568710454 / 551584435 / +3.1%	612927570 / 717665038 / -14.6%	0.93 / 0.77 / +20.7%
tpch_q22/duckdb:parquet	457931017 / 349008457 / +31.2% 🔴	1082361290 / 430769821 / +151.3% 🔴	0.42 / 0.81 / -47.8%

…h other A planner may now emit a child with `PlannerOutput::Spawn`: the run does not queue it but hands it to its owner as `Progress::Spawn`, and the planner continues at once. The owner runs it anywhere; the blocking driver admits it as another root of the same run. The chunk planner spawns every chunk. The scan's stream runs each on a run and a session of its own, stepped on `spawn_cpu`, up to `concurrency` chunks per worker thread at once, merging their batches as they finish. A scan with fewer engine partitions than cores had lost the parallelism of one task per split, 2 to 4 times slower on selective queries with one partition; it is back, without a planner knowing about threads. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

Hints were deduplicated by scanning the batch for each segment, which is quadratic in a chunk's segments: a wide projection over a filtered chunk announces hundreds. A set of the segments seen replaces the scan. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

…es not A spawned chunk runs on one task of the runtime that drives its run to the end and awaits only its reads, instead of a `spawn_cpu` task per step: a chunk's run waits for reads many times, and a hop at each wait was latency on its critical path that no other chunk's work filled. At a concurrency of one, chunks run on the scan's own run as further roots: the run's steps already interleave their work, and a run, a session and a task per chunk only added a hand-off per chunk, measured at 2 to 3 percent on TPC-H, for parallelism the engine provides. DataFusion sets the concurrency, unless configured, to the cores per partition, so a scan over one partition per core runs its chunks inline and a single-partition scan runs them beside each other. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

A planner says what it needs, compute or reads. Where either runs is an engine's decision, so this crate stops driving: `ScanBuilder::prepare` lowers and plans a scan without IO into a `PreparedScan`, which exposes its rows, selection, filter plans and IO service, and `PrunePlanner` and `ChunkPlanner` are public for an engine to compose, the prune planner continuing into whatever it is built with and passing `None` when no proof ran. The builder's ordering, concurrency, mapping and streams go with the driver, and the layout's tests drive their scans on a run of their own. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

The opener composes a scan's planners, a prune planner into a chunk planner when the scan has a filter and the chunk planner alone when it does not, and a driver of this crate runs them: the root run stepped on `spawn_cpu` tasks as DataFusion polls the stream, each spawned chunk on a task of its own up to the concurrency, or on the root run when the engine's partitions already fill the cores, with the concurrency derived from cores per partition unless configured. Batches are mapped to Arrow on the task that produced them. Every spawning decision of a V2 scan under DataFusion is in this crate. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

`vortex_file::v2::ScanDriver` composes a scan's planners and runs them on the session's runtime, the way an engine's driver does, for the file API's callers: the compress and random-access benchmarks move to it. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

The opener prepares the scan where compute runs, composes its planners itself, a prune planner continuing into a chunk planner when there is a filter and the chunk planner alone when there is not, and hands the driver the first planner. The driver only drives. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

The prune planner skipped prefetching the zone maps of proofs an earlier scan of the file had proved. The segment list now comes from compiling the proof, and a proved proof compiles to its kept result without a segment scan, so the check repeated what compile already knew. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

The scan code named the filter's zone-map plans proofs. They are the filter's pruning expressions evaluated over the file's zone map, so the layout, file and DataFusion code now says pruning, matching prune_zones, PrunedZones and the pruning field they already used. Behaviour is unchanged. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

…steps A planner that waits can now say what it will read later: the run asks every parked planner to speculate when nothing else can compute, and registers the answer with its root's session as reads that are never delivered. The prune planner speculates the predicate columns of the next chunk of its partition on each ask, so they are read while the zone maps are, as its filter planner would read them once pruning keeps the chunk; a scan selecting rows by index speculates nothing. The segment IO service keeps speculated reads for whichever session fetches a segment first, so a chunk on its own session takes the read the prune planner started. The root run's CPU step now runs until the run has a batch, waits, or is done, collecting every chunk it spawns, instead of one spawn_cpu hop per chunk. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

The prune planner ran its pruning plans one after another, each over the rows the previous one kept, so a filter with three plans waited for three reads in turn, each wait costing a round trip and a hop of the root run. It now builds every plan's adaptor in its first compute, publishes all their zone-map reads as one batch, waits once, and intersects the kept rows. Each adaptor numbers its requests from its own range, so the planner routes deliveries by request id. The separate compile of each plan to prefetch its segments is gone: the adaptors' fetches are the reads. Debug events mark when the pruning reads go out, when a chunk's reads are speculated, and when a chunk's run starts. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

The root run was stepped on a spawn_cpu task per step, so a partition paid a hop to take its zone-map completion, another to step the run, and a task spawn for its chunk, each a few hundred microseconds when every core is busy. The root run now runs on one task that awaits its reads in place, like a chunk's, and the first chunk a partition spawns runs on that run as another root. Later chunks are spawned as before, so one partition with many chunks stays parallel. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

Planning cuts a partition's rows from the footer's row count on the writer's 8192-row grid, but the writer buffers chunks to megabytes, so at fourteen partitions nearly every boundary fell inside a chunk and both neighbours decoded it: q19 at SF1 retired 17% more instructions than the stacked V2 scan, which aligned partitions to natural splits, and none more at one partition. The opener now snaps the planning-time rows up to the natural split boundaries of the columns the scan reads, computed once per file as the byte-range fallback already does, so the partitions still tile the file and no chunk is decoded twice. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

…reads Speculation prefetched only the predicate columns of the next chunk, so the projection columns interleaved between them in the file were read as gap bytes and then read again on their own: q6 at SF1 issued 23 physical reads for 164 requests where the stacked scan issued 14, and 14% more bytes. The prune planner now speculates the batch the chunk's filter planner would publish, predicate prefetches and projection announcements together, and the segment IO service keeps speculated announcements, requested and never polled, for whichever session fetches the segment first. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

…erlap The session polled a fetch as it was submitted, which marked it wanted and let the read driver dispatch the zone map alone before the run had speculated the chunk's reads microseconds later: q6 at SF1 still issued 25 physical reads for the stacked scan's 14. The run now speculates before it polls its sources, and the session registers a fetch on submit and polls it when the run polls, so what a planner speculates is registered beside the reads it waits for and a coalescing source folds them into one read. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

A chunk's filter planner announced its projection through its own session, which registered a second request for a segment speculation had already announced beside the chunk's reads. The second request was dispatched alone when the projection fetched, after the chunk's read had gone: two of each partition's eleven requests came back as a separate read inside the same range. A session announcement now takes the speculated announcement, and a fetch prefers it too. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

A chunk straddling two partitions is fetched by both. A partition's chunk took its speculated read out of the service's map, so when the neighbouring partition speculated the same segment a moment later it started another read, and once the first had completed the file's shared source no longer folded them: on q6 at SF1, 26 segments were read twice. A speculated read now stays in the service's map for as long as the service lives, cloned by every session that fetches it. Trace events name each new segment request by the path that made it. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

Filtered chunks were cut on a fixed 1M-row grid over the file, so a partition of 430K rows straddling a grid line got a small second chunk, speculated only on the next idle wait, after its neighbour's read had gone: on q6 at SF1 those tails came back as separate reads of ranges already read. Chunks now start at the partition's own first row, so a partition shorter than a chunk is one chunk and its reads go out together. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

One chunk per idle wait left a partition of five chunks at SF10 reading its last three only once their tasks started, where the stacked scan had every split prefetching from the start. The prune planner now speculates every chunk's predicate prefetches and projection announcements in one batch on the first ask, registered together so the partition's reads coalesce as one scan's. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

Two chunks of one partition can straddle a writer chunk. The first chunk's session took the speculated announcement for it and read it; the second found nothing speculated, announced a request of its own and read the segment again at the end of the scan: q12 at SF10 issued 30 single-request reads of ranges already read. A speculated announcement now stays with the service, a session's own announcement of the segment adds nothing, and the first fetch turns it into the shared read every later session finds. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

A projection-only partition of seventeen 64K-row chunks handed each chunk through the channel on its own, each needing a poll, a hop and a task spawn, so its chunks started their first reads a few hundred microseconds apart and the last ones no longer coalesced: q12 at SF10 issued 27 single-request reads of the orders table at the end of the scan, where the stacked scan spawned every split of a partition within microseconds. The root task now collects every chunk the run spawns until it has a batch, waits, or finishes, and sends them as one group the consumer spawns at once. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

…ments A read only announced is not started until the projection fetches it, after the filter, with nothing to overlap, and in a file whose columns are written in bursts a wide projection column shares no read with the narrow predicate beside it. On q14 at SF1 the projection waited 2.4 to 4 ms on reads of its own while its CPU was under 0.1 ms, the whole gap to V1, which polls all of a split's reads at open. The prune planner now speculates the projection's prefetches after the predicates' in the same batch, and the session registers a batch's announcements before it starts any read, so what is announced is dispatched with the reads it coalesces with. Debug events mark the projection's start, reads and finish, and each batch handed to the engine. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

A prepared scan carries a set of metrics its planners and morsels update, registered with the engine's registry under the partition's labels like the IO metrics: pruning duration and rows, speculated requests, chunks planned and skipped, filter duration, conjunct evaluations and their rows, projection duration and the wait from its reads going out to its first piece, rows projected and batches handed over. The DataFusion opener registers them, so they print beside the engine's own per-operator metrics. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

…hunks A partition without a filter had each 64K-row chunk fetch its own segments as it started, so the coalescer saw them piecemeal: q9's lineitem scan at SF1 issued 62 reads and 18% more bytes than V1, which registers a split's reads at open. The chunk planner now prefetches the partition's projection segments before it emits a chunk, and a prefetch is shared by every session of the service, like a speculated read, since chunks run on sessions of their own. Listing a plan's segments now shares the scan's value cache, so a dictionary's values already held list no segment and are not read ahead again. Timer metrics gain a total. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

Debug events now bracket each phase a partition goes through: open prepared and stream returned, pruning started, reads out, speculated and finished, filter started and each conjunct evaluated, projection started, reads out, first piece and finished, and each batch sent, so a trace gives the timeline of a scan without instrumenting it again. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

The driver takes driver metrics from the opener and records, under the partition's labels, the time from the stream's creation to its first batch and to its end, each spawned chunk's run, and how many chunks it spawned, so --show-metrics gives a partition's timeline end to end beside the stages' own durations. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

…ions waiting for it Every partition of a file rebuilt the filter plans and optimized the projection on every query: all fourteen open together before any has filled the file's cache, so the cache never hit within a query. On q6 at SF1 that was 448 us of filter plans and 92 us of projection per partition, on q14 177 and 142, at open, on every partition, fourteen times over. Each plan is now a once-cell in the cache: the first partition to reach it builds it and the others wait for that plan alone, not for the cache. Debug events time prepare's steps. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

…nd values The scan cache keys plans, pruned zones and values by the identity of a bound expression's tree, and every DataFusion partition binds the filter and projection on its own, so no two partitions ever hit the same entry: every partition built every plan, proved its own zones and decoded every dictionary's values for itself. A prepared scan now interns its filter and projection in the file's cache, structurally, and uses the shared tree, so the once-cells build once and the values are decoded once per file. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

A filtered partition of fewer rows than a chunk ran as one chunk, on one task from its first read to its last batch, and an engine's partitions started together straggled on uneven cores: identical 430K-row filter chunks took 0.5 to 5 ms at SF1, and the query waited for the slowest while V1's finer futures were stolen by the idle cores. A filtered chunk is now a quarter of the partition's rows on the row grid, between 256K and 1M rows, so a partition's chunks after its first are spawned where the runtime has room. SF1 partitions get two chunks; SF10 partitions keep their 1M-row chunks. A fixed 256K cost SF10 q1 8%. Local, 14 partitions, against the previous head: SF1 0.993 median (q15 0.976, q19 0.982), SF10 1.008. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

Interning hashes an expression's whole tree, and under a mutex the partitions opening a file together serialized on it: 3 ms of lock spinning over 15 runs of q15. Readers probe concurrently; only a miss takes the write lock, and checks again under it. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

… beside A zone-map fetch is the read everything else in a partition waits on, yet the file's read driver coalesced it with the chunk reads speculated beside it, so pruning waited for a 3 MB read from the page cache and, over an object store, for a 9.5 MB one behind 150 others. An urgent intent reads a segment on its own as soon as it is published, coalesced only with other urgent reads: a segment source may request a segment alone, the read driver keeps such requests apart, and the prune planner's adaptors publish their reads urgent. Over MinIO with 20 ms of round trip, SF10 q6 pruning finished at 25 to 50 ms instead of 320 to 390. Local: neutral (SF1 0.996, SF10 1.005 against the previous commit). MinIO with 20 ms round trip and 100 MB/s per connection: SF1 0.60 of V1, SF10 1.2 (the SF10 loss is addressed next). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

A partition speculated every chunk's reads at once, and the file's read driver dispatches in registration order, so with an engine's partitions started together the last partitions' first chunks queued behind the first partitions' last: over MinIO at SF10 every partition's zone maps and first batches came at once, and the batches of the rest tailed for 300 ms while V1, reading one split at a time per partition, shared the bandwidth evenly. Each partition now reads ahead only its first chunks when the store is not a local file, two by default; the later chunks fetch their own reads as their tasks start, interleaved across partitions. A local file keeps reading every chunk ahead: reading two ahead there cost SF10 6%. MinIO, 20 ms round trip, 100 MB/s per connection, 10 partitions at SF10: 0.90 of V1 at the median (q1 0.79, q19 0.66, q6 1.23 bimodal), from 1.27 with everything read ahead; SF1 at 14 partitions 0.61 of V1. Local unchanged (SF1 1.001, SF10 1.001). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

A V2 scan cache lived only as long as the scans of one query holding it, so every query lowered the file's layout, optimized its projection and filter plans, proved its zones and decoded its dictionary values again: q2 at SF1 ran 35 prepares a query, 1.6 ms of CPU, with the first per table on the critical path of a five-table join chain. The cache now lives beside the footer in DataFusion's session-level file metadata cache, so a later query of the file plans nothing again; partitions opening a file the cache misses still share one cache, put with the entry the first of them writes, and a rewritten file invalidates both. The entry's reported size does not count the cache, which grows after the entry is put. Local, 14 partitions, against the previous commit: SF1 0.965 (q19 0.90, q6 0.93, q2 0.96, q13 0.96), SF10 0.962 (q6 0.95, q12 0.95, q19 0.94). Against V1: SF1 0.955, SF10 0.81 on the targeted sets. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vr65ZwPZz4LPjkvDA6ZUuE Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>

joseph-isaacs and others added 6 commits October 22, 2026 12:04

Add a switchable planner-driven scan v2 path for DataFusion

8b80703

Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> (cherry picked from commit 6e27e30)

joseph-isaacs added the changelog/performance A performance improvement label Oct 22, 2026

joseph-isaacs added the action/bench-sql Run SQL benchmarks without Vortex Compact on this PR label Oct 22, 2026

github-actions Bot removed the action/bench-sql Run SQL benchmarks without Vortex Compact on this PR label Oct 22, 2026

joseph-isaacs and others added 2 commits October 22, 2026 13:35

joseph-isaacs and others added 7 commits October 22, 2026 13:58

joseph-isaacs added the action/bench-sql Run SQL benchmarks without Vortex Compact on this PR label Oct 22, 2026

github-actions Bot removed the action/bench-sql Run SQL benchmarks without Vortex Compact on this PR label Oct 22, 2026

joseph-isaacs added the action/bench-sql Run SQL benchmarks without Vortex Compact on this PR label Oct 22, 2026

github-actions Bot removed the action/bench-sql Run SQL benchmarks without Vortex Compact on this PR label Oct 22, 2026

joseph-isaacs and others added 24 commits October 22, 2026 18:16

Navigation Menu

Repository navigation

Plan a V2 scan as a chain of planners from one prepared scan - #10694

Plan a V2 scan as a chain of planners from one prepared scan#10694
joseph-isaacs wants to merge 41 commits into
developfrom
ji/scan-v2-planner-chain

joseph-isaacs commented Oct 22, 2026 •

edited

Loading

Uh oh!

codspeed Bot commented Oct 22, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 22, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 22, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 22, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 22, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 22, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 22, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 22, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Oct 22, 2026 •

edited

Loading

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

joseph-isaacs commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What

Benchmarks

Planning once per file, across queries

Object store: MinIO behind toxiproxy

What was slower, and why

Metrics and phase timing

What was tried and backed out

Checks

Uh oh!

codspeed Bot commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will regress 2 benchmarks

Performance Changes

Footnotes

Uh oh!

github-actions Bot commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Polar Signals Profiling Results

Latest Run

Uh oh!

github-actions Bot commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmarks: PolarSignals Profiling 📖

Uh oh!

github-actions Bot commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmarks: FineWeb NVMe 📖

Uh oh!

github-actions Bot commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmarks: TPC-H SF=1 on NVME 📖

Uh oh!

github-actions Bot commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmarks: TPC-H SF=10 on NVME 📖

Uh oh!

github-actions Bot commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmarks: Clickbench on NVME 📖

Uh oh!

github-actions Bot commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmarks: TPC-DS SF=1 on NVME 📖

Uh oh!

github-actions Bot commented Oct 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Benchmarks: TPC-H SF=1 on S3 📖

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

joseph-isaacs commented Oct 22, 2026 •

edited

Loading

codspeed Bot commented Oct 22, 2026 •

edited

Loading

github-actions Bot commented Oct 22, 2026 •

edited

Loading

github-actions Bot commented Oct 22, 2026 •

edited

Loading

github-actions Bot commented Oct 22, 2026 •

edited

Loading

github-actions Bot commented Oct 22, 2026 •

edited

Loading

github-actions Bot commented Oct 22, 2026 •

edited

Loading

github-actions Bot commented Oct 22, 2026 •

edited

Loading

github-actions Bot commented Oct 22, 2026 •

edited

Loading

github-actions Bot commented Oct 22, 2026 •

edited

Loading