Optimize split calculation in DF source - #9167
Open
AdamGS wants to merge 2 commits into
Open
Conversation
AdamGS
force-pushed
the
adamg/splits-optimization
branch
from
August 4, 2026 11:20
2fbb89b to
60a0ff2
Compare
AdamGS
force-pushed
the
adamg/splits-optimization
branch
from
August 4, 2026 11:48
60a0ff2 to
ea6d06e
Compare
Contributor
Polar Signals Profiling ResultsLatest Run
Previous Runs (10)
Powered by Polar Signals Cloud |
Contributor
Benchmarks: PolarSignals Profiling 📖Vortex (geomean): 0.975x ➖ datafusion / vortex-file-compressed / ns (0.975x ➖, 1↑ 0↓)
No file size changes detected. |
Contributor
Benchmarks: FineWeb NVMe 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.972x ➖, 2↑ 0↓)
datafusion / parquet / ns (0.987x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (1.027x ➖, 0↑ 0↓)
duckdb / parquet / ns (1.002x ➖, 0↑ 0↓)
File Size Changes (2 files changed, -46.3% overall, 0↑ 2↓)
Totals:
|
Contributor
Benchmarks: FineWeb S3 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.263x ➖, 0↑ 5↓)
datafusion / parquet / ns (1.210x ➖, 0↑ 3↓)
duckdb / vortex-file-compressed / ns (1.042x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.974x ➖, 1↑ 1↓)
|
Contributor
Benchmarks: Clickbench Sorted on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.032x ➖, 1↑ 1↓)
datafusion / parquet / ns (0.983x ➖, 2↑ 0↓)
duckdb / vortex-file-compressed / ns (1.001x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.991x ➖, 1↑ 0↓)
File Size Changes (201 files changed, -42.8% overall, 56↑ 145↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=10 on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.879x ✅, 9↑ 0↓)
datafusion / parquet / ns (1.003x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (1.016x ➖, 0↑ 0↓)
duckdb / parquet / ns (1.000x ➖, 0↑ 0↓)
File Size Changes (9 files changed, -44.0% overall, 0↑ 9↓)
Totals:
|
Contributor
Benchmarks: TPC-DS SF=1 on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.991x ➖, 1↑ 0↓)
datafusion / parquet / ns (1.001x ➖, 1↑ 2↓)
duckdb / vortex-file-compressed / ns (0.973x ➖, 13↑ 5↓)
duckdb / parquet / ns (1.014x ➖, 4↑ 9↓)
File Size Changes (25 files changed, -43.5% overall, 0↑ 25↓)
Totals:
|
Contributor
Benchmarks: Statistical and Population Genetics 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
duckdb / vortex-file-compressed / ns (1.003x ➖, 3↑ 4↓)
duckdb / parquet / ns (0.997x ➖, 0↑ 0↓)
File Size Changes (2 files changed, -32.3% overall, 0↑ 2↓)
Totals:
|
Contributor
Benchmarks: Clickbench on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.009x ➖, 2↑ 5↓)
datafusion / parquet / ns (1.012x ➖, 1↑ 2↓)
duckdb / vortex-file-compressed / ns (1.030x ➖, 1↑ 6↓)
duckdb / parquet / ns (0.991x ➖, 0↑ 1↓)
File Size Changes (101 files changed, -39.2% overall, 0↑ 101↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on S3 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.006x ➖, 0↑ 0↓)
datafusion / parquet / ns (0.984x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (0.961x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.958x ➖, 0↑ 0↓)
|
Contributor
Benchmarks: TPC-H SF=1 on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.951x ➖, 3↑ 0↓)
datafusion / parquet / ns (1.026x ➖, 0↑ 3↓)
duckdb / vortex-file-compressed / ns (0.976x ➖, 1↑ 0↓)
duckdb / parquet / ns (0.999x ➖, 0↑ 0↓)
File Size Changes (9 files changed, -43.9% overall, 0↑ 9↓)
Totals:
|
AdamGS
force-pushed
the
adamg/splits-optimization
branch
2 times, most recently
from
August 5, 2026 13:33
57c29f4 to
10cbfb8
Compare
AdamGS
marked this pull request as ready for review
August 5, 2026 13:37
AdamGS
force-pushed
the
adamg/splits-optimization
branch
from
August 5, 2026 15:53
10cbfb8 to
a127540
Compare
Signed-off-by: Adam Gutglick <adam@spiraldb.com>
Signed-off-by: Adam Gutglick <adam@spiraldb.com>
AdamGS
force-pushed
the
adamg/splits-optimization
branch
from
August 5, 2026 16:52
bd3954d to
762c28f
Compare
AdamGS
added
the
action/bench-sql
github-actions
Bot
removed
the
action/bench-sql
Merging this PR will degrade performance by 11.64%
Warning Please fix the performance issues or acknowledge them on CodSpeed. Performance Changes
Tip Investigate this regression by commenting Comparing 
Footnotes
|
Contributor
Benchmarks: PolarSignals Profiling 📖Vortex (geomean): 0.954x ➖ datafusion / vortex-file-compressed / ns (0.954x ➖, 2↑ 0↓)
No file size changes detected. |
Contributor
Benchmarks: Clickbench on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.914x ➖, 10↑ 1↓)
datafusion / parquet / ns (0.998x ➖, 0↑ 1↓)
duckdb / vortex-file-compressed / ns (0.992x ➖, 2↑ 2↓)
duckdb / parquet / ns (1.005x ➖, 0↑ 2↓)
File Size Changes (101 files changed, -39.2% overall, 0↑ 101↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.939x ➖, 5↑ 1↓)
datafusion / parquet / ns (1.016x ➖, 0↑ 3↓)
duckdb / vortex-file-compressed / ns (1.004x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.996x ➖, 0↑ 0↓)
File Size Changes (9 files changed, -43.9% overall, 0↑ 9↓)
Totals:
|
Contributor
Benchmarks: FineWeb NVMe 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.052x ➖, 0↑ 1↓)
datafusion / parquet / ns (1.033x ➖, 0↑ 1↓)
duckdb / vortex-file-compressed / ns (1.020x ➖, 0↑ 2↓)
duckdb / parquet / ns (1.014x ➖, 0↑ 0↓)
File Size Changes (2 files changed, -46.3% overall, 0↑ 2↓)
Totals:
|
Contributor
Benchmarks: TPC-DS SF=1 on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.980x ➖, 3↑ 0↓)
datafusion / parquet / ns (1.002x ➖, 0↑ 4↓)
duckdb / vortex-file-compressed / ns (1.012x ➖, 3↑ 9↓)
duckdb / parquet / ns (1.000x ➖, 3↑ 3↓)
File Size Changes (25 files changed, -43.5% overall, 0↑ 25↓)
Totals:
|
Contributor
Benchmarks: Clickbench Sorted on NVME 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.922x ➖, 5↑ 0↓)
datafusion / parquet / ns (1.027x ➖, 0↑ 1↓)
duckdb / vortex-file-compressed / ns (1.021x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.983x ➖, 0↑ 0↓)
File Size Changes (201 files changed, -42.8% overall, 58↑ 143↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on S3 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.907x ➖, 0↑ 0↓)
datafusion / parquet / ns (0.961x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (0.988x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.952x ➖, 0↑ 0↓)
|
Contributor
Benchmarks: FineWeb S3 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.287x ➖, 0↑ 4↓)
datafusion / parquet / ns (1.050x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (0.953x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.782x ➖, 1↑ 0↓)
|
robert3005
reviewed
Aug 5, 2026
robert3005
reviewed
Aug 5, 2026
Comment on lines
+531
to
+534
| debug_assert_eq!( | ||
| assignment_bytes.len() + usize::from(!row_boundaries.is_empty()), | ||
| row_boundaries.len() | ||
| ); |
Contributor
There was a problem hiding this comment.
this is incorrect if row_count == 0 but there's no test that hits this
Contributor
Author
There was a problem hiding this comment.
if there are 0 rows, do we report a single split at 0?
Contributor
There was a problem hiding this comment.
I think this is only ever possible to generate in tests that's why this is confusing, no production path would do this but tests can. I think we can have a single split at 0 or no splits?
Contributor
Author
There was a problem hiding this comment.
no splits works here IMO, but I'll try a test with one split.
| /// Shared full-file natural split ranges keyed by path. | ||
| natural_split_ranges: Arc<DashMap<Path, Arc<[Range<u64>]>>>, | ||
| /// Shared full-file natural splits keyed by path. | ||
| natural_splits: Arc<DashMap<Path, Arc<NaturalSplits>>>, |
Contributor
There was a problem hiding this comment.
This is scoped to a single table read right?
Contributor
Author
There was a problem hiding this comment.
Yes.
I think there's a way to make this more robust which also applies to other file-specific stuff we cache, I'll do it as a follow up.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Rationale for this change
When DataFusion splits a Vortex file into byte ranges, every partition open recomputed the file's natural splits and linearly scanned them to translate its byte range into row boundaries — identical work repeated N times per file, with the layout walks contending on lazily-initialized layout children.
Now the splits are computed once per file (for only the fields the scan references), cached with precomputed assignment bytes so the translation is a binary search, and handed back to the
ScanBuilderso per-partitionprepare()skips its layout walk entirely.What changes are included in this PR?
vortex-datafusion: per-fileNaturalSplitscache (boundaries + sorted assignment bytes), computed under theDashMapentry so concurrent partitions wait for the winner instead of racing the walk.vortex-layout: newScanBuilder::full_file_splits()/with_natural_splits();Splits::Naturalnow holdsArc<[u64]>. FixedRepeatedScan::executeemitting an empty leading split when the row range starts on a natural boundary (the common case for translated ranges).vortex-layoutandvortex-datafusionsuites pass.What APIs are changed? Are there any user-facing changes?
New public
ScanBuilder::full_file_splits()andwith_natural_splits().Splits::Natural's type change is in a private module. No user-facing behavior changes.