Use approximate cardinality to decide whether to use dict compression#7759
Draft
robert3005 wants to merge 1 commit intodevelopfrom
Draft
Use approximate cardinality to decide whether to use dict compression#7759robert3005 wants to merge 1 commit intodevelopfrom
robert3005 wants to merge 1 commit intodevelopfrom
Conversation
robert3005
added
action/benchmark
github-actions
Bot
removed
the
action/benchmark
Merging this PR will degrade performance by 24.9%
|
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ❌ | Simulation | varbinview_zip_fragmented_mask |
6.5 ms | 7.3 ms | -10.29% |
| ❌ | Simulation | decode_primitives[u8, (1000, 2)] |
15.8 µs | 17.6 µs | -10.29% |
| ⚡ | Simulation | encode_varbinview[(1000, 2)] |
243.8 µs | 168.1 µs | +45.05% |
| ⚡ | Simulation | encode_varbin[(10000, 512)] |
1,029.1 µs | 934.6 µs | +10.11% |
| ❌ | Simulation | varbinview_zip_block_mask |
2.9 ms | 3.7 ms | -21.65% |
| ❌ | Simulation | new_bp_prim_test_between[i64, 32768] |
177.3 µs | 236.1 µs | -24.9% |
| ⚡ | Simulation | dict_compress_string |
8.2 ms | 7.4 ms | +10.54% |
Comparing rk/cardinality-estimator (7fa9309) with develop (5e5572b)
Footnotes
-
138 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. ↩
Contributor
Polar Signals Profiling ResultsLatest Run
Previous Runs (3)
Powered by Polar Signals Cloud |
Contributor
Benchmarks: CompressionVortex (geomean): 1.002x ➖ unknown / unknown (0.992x ➖, 4↑ 5↓)
|
Contributor
Benchmarks: PolarSignals ProfilingVortex (geomean): 0.933x ➖ datafusion / vortex-file-compressed (0.933x ➖, 2↑ 0↓)
|
Contributor
Benchmarks: FineWeb NVMeVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.983x ➖, 1↑ 0↓)
datafusion / vortex-compact (1.006x ➖, 0↑ 0↓)
datafusion / parquet (0.981x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (0.978x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.056x ➖, 0↑ 1↓)
duckdb / parquet (0.965x ➖, 1↑ 0↓)
Full attributed analysis
|
Contributor
File Sizes: FineWeb NVMeNo file size changes detected. |
Contributor
Benchmarks: TPC-DS SF=1 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.917x ➖, 42↑ 0↓)
datafusion / vortex-compact (0.955x ➖, 11↑ 0↓)
datafusion / parquet (0.925x ➖, 32↑ 0↓)
duckdb / vortex-file-compressed (1.003x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.008x ➖, 1↑ 1↓)
duckdb / parquet (1.006x ➖, 0↑ 3↓)
duckdb / duckdb (0.991x ➖, 3↑ 1↓)
Full attributed analysis
|
Contributor
Benchmarks: Statistical and Population GeneticsVerdict: No clear signal (low confidence) duckdb / vortex-file-compressed (0.995x ➖, 0↑ 0↓)
duckdb / vortex-compact (0.990x ➖, 0↑ 0↓)
duckdb / parquet (0.986x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: FineWeb S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (1.001x ➖, 0↑ 0↓)
datafusion / vortex-compact (0.983x ➖, 0↑ 0↓)
datafusion / parquet (0.985x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (0.979x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.015x ➖, 0↑ 0↓)
duckdb / parquet (0.948x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: Clickbench on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.997x ➖, 1↑ 0↓)
datafusion / parquet (1.002x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (0.975x ➖, 2↑ 0↓)
duckdb / parquet (0.996x ➖, 0↑ 0↓)
duckdb / duckdb (0.967x ➖, 6↑ 0↓)
Full attributed analysis
|
Contributor
File Sizes: Clickbench on NVMEFile Size Changes (105 files changed, +0.0% overall, 95↑ 10↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.014x ➖, 0↑ 0↓)
datafusion / vortex-compact (1.014x ➖, 0↑ 0↓)
datafusion / parquet (0.979x ➖, 2↑ 0↓)
datafusion / arrow (1.007x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (1.016x ➖, 0↑ 1↓)
duckdb / vortex-compact (1.003x ➖, 0↑ 0↓)
duckdb / parquet (1.004x ➖, 0↑ 1↓)
duckdb / duckdb (1.006x ➖, 0↑ 1↓)
Full attributed analysis
|
Contributor
File Sizes: TPC-H SF=1 on NVMEFile Size Changes (195 files changed, -98.4% overall, 0↑ 195↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=10 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.000x ➖, 0↑ 0↓)
datafusion / vortex-compact (0.997x ➖, 0↑ 0↓)
datafusion / parquet (0.994x ➖, 0↑ 0↓)
datafusion / arrow (0.958x ➖, 2↑ 0↓)
duckdb / vortex-file-compressed (0.992x ➖, 0↑ 0↓)
duckdb / vortex-compact (0.997x ➖, 0↑ 0↓)
duckdb / parquet (1.003x ➖, 0↑ 0↓)
duckdb / duckdb (0.998x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
File Sizes: TPC-H SF=10 on NVMEFile Size Changes (2 files changed, +0.0% overall, 2↑ 0↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.946x ➖, 0↑ 0↓)
datafusion / vortex-compact (0.877x ➖, 0↑ 0↓)
datafusion / parquet (0.887x ➖, 1↑ 0↓)
duckdb / vortex-file-compressed (0.963x ➖, 0↑ 0↓)
duckdb / vortex-compact (0.968x ➖, 0↑ 0↓)
duckdb / parquet (0.948x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: TPC-H SF=10 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.962x ➖, 0↑ 0↓)
datafusion / vortex-compact (1.000x ➖, 0↑ 1↓)
datafusion / parquet (0.909x ➖, 2↑ 1↓)
duckdb / vortex-file-compressed (0.976x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.033x ➖, 0↑ 0↓)
duckdb / parquet (0.998x ➖, 0↑ 0↓)
Full attributed analysis
|
github-actions
Bot
removed
the
action/benchmark
Contributor
File Sizes: PolarSignals ProfilingNo file size changes detected. |
Replace the exact `HashMap`/`HashSet` previously used to compute distinct-value counts during compression stats generation with Cloudflare's `cardinality-estimator` crate. The estimator gives us a bounded-memory approximation (exact up to ~128 distinct values, then HyperLogLog++) so high-cardinality arrays no longer require an O(n) auxiliary hash table to answer the single question "how many unique values does this have?". - Integer stats swap the hash map for a `CardinalityEstimator` and track the most frequent value via a Boyer-Moore majority candidate plus a second-pass exact count. Sparse/dict schemes only care about the heavy hitter (>= 90% threshold) or a rough distinct ratio, so this is behaviourally equivalent for the decisions they make. - Float and string stats likewise drop their hash sets in favor of the estimator. - The integer and float dictionary encoders now rebuild the exact set of distinct values from the source array at compress time, since they need the values themselves and the stats layer no longer retains them. - `SequenceScheme`'s fast-path check for "all values are distinct" now tolerates the estimator's small approximation error; the deferred callback still validates sequences exactly. Signed-off-by: Robert Kruszewski <github@robertk.io>
robert3005
force-pushed
the
rk/cardinality-estimator
branch
from
fc38e36 to
7fa9309
Compare
Contributor
File Sizes: TPC-DS SF=1 on NVMEFile Size Changes (4 files changed, +0.0% overall, 2↑ 2↓)
Totals:
|
github-actions
Bot
removed
the
action/benchmark
Contributor
File Sizes: Statistical and Population GeneticsFile Size Changes (2 files changed, +0.0% overall, 1↑ 1↓)
Totals:
|
Contributor
Benchmarks: Random AccessVortex (geomean): 0.878x ✅ unknown / unknown (0.920x ➖, 13↑ 0↓)
|
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
1 participant
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.
Since we no longer precompute all unique values the layout and compressor dict
encoding logic get unified