Optimize buffer ops#8322
Draft
AdamGS wants to merge 8 commits into
Draft
Conversation
Merging this PR will degrade performance by 12.18%
|
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ❌ | Simulation | chunked_bool_canonical_into[(1000, 10)] |
20.7 µs | 35.5 µs | -41.8% |
| ❌ | Simulation | compare[48] |
213 µs | 300.5 µs | -29.11% |
| ❌ | Simulation | compare[50] |
227.7 µs | 319 µs | -28.63% |
| ❌ | Simulation | compare[49] |
228.2 µs | 317.1 µs | -28.04% |
| ❌ | Simulation | compare[44] |
207.5 µs | 287.6 µs | -27.87% |
| ❌ | Simulation | compare[47] |
223.5 µs | 309.1 µs | -27.68% |
| ❌ | Simulation | compare[46] |
218.5 µs | 302.1 µs | -27.67% |
| ❌ | Simulation | compare[40] |
190.7 µs | 263.4 µs | -27.61% |
| ❌ | Simulation | compare[44] |
212.1 µs | 292.2 µs | -27.4% |
| ❌ | Simulation | compare[45] |
218.9 µs | 300.8 µs | -27.23% |
| ❌ | Simulation | compare[43] |
209.2 µs | 287.5 µs | -27.21% |
| ❌ | Simulation | compare[42] |
204.5 µs | 281 µs | -27.21% |
| ❌ | Simulation | compare[40] |
195.6 µs | 268.1 µs | -27.04% |
| ❌ | Simulation | compare[43] |
214.2 µs | 292.4 µs | -26.74% |
| ❌ | Simulation | compare[41] |
204.5 µs | 279.1 µs | -26.73% |
| ❌ | Simulation | compare[42] |
209.4 µs | 284.9 µs | -26.5% |
| ❌ | Simulation | compare[41] |
209.3 µs | 283.8 µs | -26.23% |
| ❌ | Simulation | compare[31] |
157.7 µs | 213.6 µs | -26.18% |
| ❌ | Simulation | compare[39] |
199.9 µs | 270.7 µs | -26.16% |
| ❌ | Simulation | compare[38] |
195.5 µs | 264.2 µs | -26.01% |
| ... | ... | ... | ... | ... | ... |
ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.
Tip
Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.
Comparing adamg/buffer-slice-fast (11bb86c) with develop (f67b594)
Footnotes
-
10 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. ↩
connortsui20
pushed a commit
that referenced
this pull request
Jun 9, 2026
## Summary Adds a basic benchmark for slicing, including an Arrow baseline. Hopefully building up to #8322, but I want a baseline first. Signed-off-by: Adam Gutglick <adam@spiraldb.com>
AdamGS
force-pushed
the
adamg/buffer-slice-fast
branch
from
fd451bf to
341039a
Compare
github-actions
Bot
removed
the
action/benchmark
Contributor
Polar Signals Profiling ResultsLatest Run
Previous Runs (2)
Powered by Polar Signals Cloud |
Contributor
Benchmarks: PolarSignals ProfilingVortex (geomean): 1.090x ➖ How to read Verdict and Engines
datafusion / vortex-file-compressed (1.090x ➖, 0↑ 4↓)
No file size changes detected. |
Contributor
Benchmarks: FineWeb NVMeVerdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed (1.112x ❌, 0↑ 6↓)
datafusion / vortex-compact (1.080x ➖, 0↑ 2↓)
datafusion / parquet (1.121x ❌, 0↑ 5↓)
duckdb / vortex-file-compressed (1.141x ❌, 0↑ 8↓)
duckdb / vortex-compact (1.083x ➖, 0↑ 4↓)
duckdb / parquet (1.102x ❌, 0↑ 3↓)
File Size Changes (1 files changed, +0.0% overall, 1↑ 0↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on NVMEVerdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed (1.062x ➖, 0↑ 2↓)
datafusion / vortex-compact (0.932x ➖, 10↑ 0↓)
datafusion / parquet (1.041x ➖, 1↑ 4↓)
datafusion / arrow (0.960x ➖, 4↑ 3↓)
duckdb / vortex-file-compressed (1.041x ➖, 0↑ 6↓)
duckdb / vortex-compact (1.000x ➖, 1↑ 0↓)
duckdb / parquet (0.990x ➖, 0↑ 0↓)
duckdb / duckdb (1.006x ➖, 0↑ 0↓)
File Size Changes (9 files changed, +0.2% overall, 9↑ 0↓)
Totals:
|
Contributor
Benchmarks: TPC-DS SF=1 on NVMEVerdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed (0.991x ➖, 6↑ 1↓)
datafusion / vortex-compact (1.003x ➖, 0↑ 2↓)
datafusion / parquet (0.996x ➖, 3↑ 0↓)
duckdb / vortex-file-compressed (0.994x ➖, 2↑ 1↓)
duckdb / vortex-compact (0.994x ➖, 0↑ 1↓)
duckdb / parquet (0.999x ➖, 1↑ 1↓)
duckdb / duckdb (0.997x ➖, 0↑ 3↓)
File Size Changes (7 files changed, +0.0% overall, 7↑ 0↓)
Totals:
|
Contributor
Benchmarks: Statistical and Population GeneticsVerdict: No clear signal (low confidence) How to read Verdict and Engines
duckdb / vortex-file-compressed (1.022x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.020x ➖, 0↑ 0↓)
duckdb / parquet (1.025x ➖, 0↑ 0↓)
File Size Changes (1 files changed, +0.0% overall, 1↑ 0↓)
Totals:
|
Contributor
Benchmarks: FineWeb S3Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed (0.971x ➖, 1↑ 1↓)
datafusion / vortex-compact (1.014x ➖, 2↑ 1↓)
datafusion / parquet (0.920x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (0.853x ➖, 2↑ 0↓)
duckdb / vortex-compact (0.968x ➖, 0↑ 1↓)
duckdb / parquet (0.972x ➖, 0↑ 0↓)
|
Contributor
BENCHMARK FAILEDBenchmark |
Contributor
Benchmarks: TPC-H SF=10 on NVMEVerdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed (0.993x ➖, 0↑ 0↓)
datafusion / vortex-compact (1.006x ➖, 0↑ 0↓)
datafusion / parquet (0.997x ➖, 0↑ 0↓)
datafusion / arrow (0.926x ➖, 7↑ 0↓)
duckdb / vortex-file-compressed (1.002x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.003x ➖, 0↑ 0↓)
duckdb / parquet (1.000x ➖, 0↑ 0↓)
duckdb / duckdb (0.998x ➖, 0↑ 0↓)
File Size Changes (26 files changed, -0.0% overall, 8↑ 18↓)
Totals:
|
Contributor
Benchmarks: Clickbench on NVMEVerdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed (0.969x ➖, 13↑ 2↓)
datafusion / parquet (0.913x ➖, 13↑ 0↓)
duckdb / vortex-file-compressed (1.011x ➖, 4↑ 0↓)
duckdb / parquet (1.002x ➖, 0↑ 0↓)
duckdb / duckdb (1.013x ➖, 0↑ 1↓)
File Size Changes (107 files changed, -0.0% overall, 56↑ 51↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on S3Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed (0.873x ➖, 3↑ 0↓)
datafusion / vortex-compact (0.867x ➖, 5↑ 0↓)
datafusion / parquet (1.267x ➖, 1↑ 13↓)
duckdb / vortex-file-compressed (0.875x ➖, 1↑ 0↓)
duckdb / vortex-compact (0.933x ➖, 0↑ 0↓)
duckdb / parquet (0.917x ➖, 0↑ 0↓)
|
Contributor
Benchmarks: Appian on NVMEVerdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed (1.008x ➖, 0↑ 0↓)
datafusion / parquet (1.007x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (1.031x ➖, 0↑ 0↓)
duckdb / parquet (1.006x ➖, 0↑ 0↓)
duckdb / duckdb (1.007x ➖, 0↑ 0↓)
File Size Changes (4 files changed, -0.0% overall, 1↑ 3↓)
Totals:
|
Contributor
Benchmarks: CompressionVortex (geomean): 1.016x ➖ How to read Verdict and Engines
unknown / unknown (1.045x ➖, 2↑ 29↓)
|
Contributor
Benchmarks: TPC-H SF=10 on S3Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed (0.870x ➖, 4↑ 1↓)
datafusion / vortex-compact (0.867x ➖, 3↑ 0↓)
datafusion / parquet (1.044x ➖, 0↑ 4↓)
duckdb / vortex-file-compressed (0.929x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.065x ➖, 0↑ 2↓)
duckdb / parquet (0.921x ➖, 0↑ 0↓)
|
Contributor
|
I made #8162 that optimizes some of these code paths |
AdamGS
force-pushed
the
adamg/buffer-slice-fast
branch
from
06ac2f8 to
dde48e4
Compare
github-actions
Bot
removed
the
action/benchmark
Contributor
Author
|
@robert3005 do you want to merge that first? |
Contributor
|
that would be ideal, the pr I made is a revival of an older pr already |
Contributor
Author
|
I'll review it |
AdamGS
force-pushed
the
adamg/buffer-slice-fast
branch
from
dde48e4 to
ae9bc12
Compare
github-actions
Bot
removed
the
action/benchmark
AdamGS
force-pushed
the
adamg/buffer-slice-fast
branch
3 times, most recently
from
d4f17f3 to
64fb797
Compare
robert3005
reviewed
Jun 11, 2026
| } | ||
| } else { | ||
| // Use bitvec for unaligned bit copying. | ||
| let self_slice = self |
Contributor
There was a problem hiding this comment.
We have benchmarked this and bitvec is faster for anything that's bigger than 128 bits. I think we want to keep it
AdamGS
force-pushed
the
adamg/buffer-slice-fast
branch
3 times, most recently
from
bf88b71 to
b8ad2b4
Compare
Signed-off-by: Adam Gutglick <adam@spiraldb.com>
Signed-off-by: Adam Gutglick <adam@spiraldb.com>
Signed-off-by: Adam Gutglick <adam@spiraldb.com>
Signed-off-by: Adam Gutglick <adam@spiraldb.com>
Signed-off-by: Adam Gutglick <adam@spiraldb.com>
Signed-off-by: Adam Gutglick <adam@spiraldb.com>
AdamGS
force-pushed
the
adamg/buffer-slice-fast
branch
from
b8ad2b4 to
a7bdc00
Compare
Signed-off-by: Adam Gutglick <adam@spiraldb.com>
AdamGS
force-pushed
the
adamg/buffer-slice-fast
branch
from
e465b55 to
ad6949b
Compare
AdamGS
force-pushed
the
adamg/buffer-slice-fast
branch
from
be356c3 to
11bb86c
Compare
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
2 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.
Summary
This PR includes a few optimization for buffer-level ops:
BitBufferMut::append_bufferuses arrow's word-sized append for unaligned bitbuffers instead of bitvec which is 1 bit a time.Alignment, instead of having less specific checks in different callsites.After this PR is merged, I'll follow up and remove
bitvecas a dependency, its currently used in a couple of pretty random places and I suspect there's nothing special about them compared to our ownBitBuffer.