CI: run each benchmark query in a separate process - #9385
Open
myrrc wants to merge 1 commit into
Open
Conversation
Signed-off-by: Mikhail Kot <mikhail@spiraldb.com>
myrrc
enabled auto-merge (squash)
August 12, 2026 16:03
myrrc
added
the
action/bench-sql
github-actions
Bot
removed
the
action/bench-sql
Merging this PR will degrade performance by 20.38%
|
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ❌ | Simulation | decompress[u64, (10000, 4)] |
310.1 µs | 403.3 µs | -23.12% |
| ❌ | Simulation | cold_misaligned[(64, 256)] |
4.4 ms | 5.3 ms | -17.53% |
| 🆕 | WallTime | words_gather_dispatch[1024] |
N/A | 7 ns | N/A |
| 🆕 | WallTime | words_gather_dispatch[65536] |
N/A | 1.3 µs | N/A |
| 🆕 | WallTime | words_gather_scalar[1024] |
N/A | 100 ns | N/A |
| 🆕 | WallTime | words_gather_scalar[65536] |
N/A | 6.4 µs | N/A |
| 🆕 | WallTime | words_gather_dispatch[1024] |
N/A | 17 ns | N/A |
| 🆕 | WallTime | words_gather_dispatch[65536] |
N/A | 1.3 µs | N/A |
| 🆕 | WallTime | words_gather_scalar[1024] |
N/A | 147 ns | N/A |
| 🆕 | WallTime | words_gather_scalar[65536] |
N/A | 9.4 µs | N/A |
| 🆕 | WallTime | words_gather_dispatch[1024] |
N/A | 30 ns | N/A |
| 🆕 | WallTime | words_gather_dispatch[65536] |
N/A | 2.1 µs | N/A |
| 🆕 | WallTime | words_gather_scalar[1024] |
N/A | 360 ns | N/A |
| 🆕 | WallTime | words_gather_scalar[65536] |
N/A | 22.5 µs | N/A |
Tip
Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.
Comparing myrrc/bench-process-per-query (eac735c) with develop (7ba790f)2
Footnotes
-
89 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. ↩
-
No successful run was found on
develop(204d1d4) during the generation of this report, so 7ba790f was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩
Contributor
Polar Signals Profiling ResultsLatest Run
Powered by Polar Signals Cloud |
Contributor
Benchmarks: PolarSignals Profiling 📖Vortex (geomean): 2.055x ❌ datafusion / vortex-file-compressed / ns (2.055x ❌, 0↑ 10↓)
No file size changes detected. |
Contributor
Benchmarks: FineWeb NVMe 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (2.162x ❌, 0↑ 9↓)
datafusion / parquet / ns (1.449x ❌, 0↑ 9↓)
duckdb / vortex-file-compressed / ns (1.697x ❌, 0↑ 9↓)
duckdb / parquet / ns (1.063x ➖, 1↑ 3↓)
File Size Changes (2 files changed, -46.3% overall, 0↑ 2↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on NVME 📖Verdict: Likely regression (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (2.167x ❌, 0↑ 22↓)
datafusion / parquet / ns (1.329x ❌, 0↑ 21↓)
duckdb / vortex-file-compressed / ns (1.820x ❌, 0↑ 22↓)
duckdb / parquet / ns (1.054x ➖, 0↑ 2↓)
File Size Changes (9 files changed, -43.9% overall, 0↑ 9↓)
Totals:
|
Contributor
Benchmarks: FineWeb S3 📖Verdict: Likely regression (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.026x ➖, 0↑ 1↓)
datafusion / parquet / ns (1.193x ➖, 0↑ 2↓)
duckdb / vortex-file-compressed / ns (1.000x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.168x ✅, 9↑ 0↓)
|
Contributor
Benchmarks: Statistical and Population Genetics 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
duckdb / vortex-file-compressed / ns (1.066x ➖, 3↑ 4↓)
duckdb / parquet / ns (1.047x ➖, 0↑ 2↓)
File Size Changes (2 files changed, -32.3% overall, 0↑ 2↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=10 on NVME 📖Verdict: Likely regression (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.423x ❌, 0↑ 22↓)
datafusion / parquet / ns (1.202x ❌, 0↑ 19↓)
duckdb / vortex-file-compressed / ns (1.363x ❌, 0↑ 19↓)
duckdb / parquet / ns (1.045x ➖, 1↑ 3↓)
File Size Changes (9 files changed, -44.0% overall, 0↑ 9↓)
Totals:
|
Contributor
Benchmarks: Clickbench on NVME 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.722x ❌, 0↑ 38↓)
datafusion / parquet / ns (2.150x ❌, 0↑ 39↓)
duckdb / vortex-file-compressed / ns (1.761x ❌, 0↑ 41↓)
duckdb / parquet / ns (1.421x ❌, 0↑ 42↓)
File Size Changes (101 files changed, -39.2% overall, 0↑ 101↓)
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 (2.359x ❌, 0↑ 10↓)
datafusion / parquet / ns (2.076x ❌, 0↑ 9↓)
duckdb / vortex-file-compressed / ns (2.188x ❌, 0↑ 10↓)
duckdb / parquet / ns (1.563x ❌, 0↑ 9↓)
File Size Changes (201 files changed, -42.8% overall, 49↑ 152↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on S3 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.634x ❌, 0↑ 20↓)
datafusion / parquet / ns (1.559x ❌, 0↑ 21↓)
duckdb / vortex-file-compressed / ns (1.120x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.998x ➖, 0↑ 0↓)
|
Contributor
Benchmarks: TPC-DS SF=1 on NVME 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.895x ❌, 0↑ 97↓)
datafusion / parquet / ns (1.682x ❌, 0↑ 99↓)
duckdb / vortex-file-compressed / ns (1.624x ❌, 0↑ 97↓)
duckdb / parquet / ns (1.163x ❌, 2↑ 74↓)
File Size Changes (25 files changed, -43.5% overall, 0↑ 25↓)
Totals:
|
Contributor
|
Should we also clear the cache between runs? |
Contributor
Author
I think no, because then we'd measure cold runs only, and I find ClickBench's combined metric or similar metric of that kind) useful. In the end, in production you'd likely operate on a warm server |
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.
Currently all queries in a benchmark suite are run in a same process.
This is an issue for #9381 since
memory pressure for earlier queries influences latter queries. This in turn
makes results very different.
The distinction between now and before is that cache was dropped per benchmark target (i.e. clickbench) and all queries per benchmark ran in same process. Now every query runs 10 times, each time in a separate process, and cache is dropped after every query (i.e. after every 10 runs) to make sure we don't benefit from cached pages from previous request.
Now we spawn a new process for every query. This also brings benchmark results
closer to ClickBench.
As result, duckdb-bench and datafusion-bench no longer need
--iterationsflag.Resolves: #9381