Skip to content

Vortex 0.90.0 for DataFusion and DuckDB with automated results (replaces #3056, #3057) - #3355

Merged
alexey-milovidov merged 7 commits into
mainfrom
vortex-0.90-automated-results
Nov 15, 2026
Merged

alexey-milovidov merged 7 commits into
mainfrom
vortex-0.90-automated-results

Conversation

@alexey-milovidov

@alexey-milovidov alexey-milovidov commented Nov 15, 2026 •

Copy link
Copy Markdown
Member

Combines #3056 and #3057 by @joseph-isaacs (Vortex 0.90.0 for datafusion-vortex(-partitioned) and duckdb-vortex(-partitioned)), and replaces the manually submitted results with the results of the automated benchmark runs. The automation could not commit them to the original PRs because they come from an organization-owned fork, where GitHub does not offer maintainer edits.

  • duckdb-vortex: automated results for c6a.4xlarge, c6a.metal, c7a.metal-48xl, c8g.4xlarge, c8g.metal-48xl; the manually added 20261102 files are removed.
  • duckdb-vortex-partitioned: automated results for c6a.large, c6a.xlarge, c6a.2xlarge, c6a.4xlarge, c6a.metal, c7a.metal-48xl, c8g.4xlarge, c8g.metal-48xl; the manually added 20261102 files are removed.
  • datafusion-vortex: automated results for c6a.xlarge, c6a.2xlarge, c6a.4xlarge, c6a.metal, c7a.metal-48xl, c8g.4xlarge, c8g.metal-48xl; the manually added 20261102 files are removed.
  • datafusion-vortex-partitioned: automated results for c6a.xlarge, c6a.2xlarge, c6a.4xlarge, c6a.metal, c7a.metal-48xl, c8g.4xlarge, c8g.metal-48xl; the manually added 20261102 files are removed. The first automated run on DataFusion (Vortex): bump to Vortex 0.90.0 and update results #3056 failed because the 0.90.0-55.2.0 tag of vortex-datafusion-cli did not exist yet; these come from the rerun on 2026-11-15.

Related: #3056
Related: #3057

🤖 Generated with Claude Code

joseph-isaacs and others added 6 commits November 4, 2026 17:38
Build vortex-datafusion-cli 0.90.0-55.2.0 (Vortex 0.90.0, DataFusion
55.2.0) for both the single-file and the partitioned variant, and add
results for c8g.metal-48xl, c7a.metal-48xl and c8g.4xlarge. The READMEs
still named the 0.70.0-53.1.0 tag, so they now match the install scripts.

Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Build duckdb-vortex from source for both the single-file and the
partitioned variant, pinned to a commit with Vortex 0.90.0 on DuckDB
v2.0.0, and add results for c8g.metal-48xl, c7a.metal-48xl and
c8g.4xlarge. DuckDB's own `INSTALL vortex` ships an older Vortex, and the
partitioned variant was still on v0.44.0.

Building against DuckDB 2.0 needs three fixes: link vortex into the shell
explicitly (STATICALLY_LINK_EXTENSIONS), skip C++20 module scanning, which
needs clang-scan-deps, and raise the open-file limit for many-core hosts.
The install script now checks for vortex_version() rather than for the
binary, since the plain shell is linked before the extension is built.

Rebased onto main after the DuckDB 2.0 bump (#3186): pass -no-agent to
the custom binary in load and query for both variants, so the CLI keeps
its plain output when run from an AI coding agent.

Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UeniBANoYjsc8bCuAtmkMC
The automated benchmark runs of #3057
produced results for `duckdb-vortex` (5 machines) and `duckdb-vortex-partitioned`
(8 machines). They could not be committed to the PR because it comes from an
organization-owned fork, so they are added here, and the manually added result
files are removed, as the collector does for PRs it can push to.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown
Contributor

@alexey-milovidov
alexey-milovidov marked this pull request as ready for review November 15, 2026 14:03

@neuroalexey neuroalexey left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Status: Needs a fix — the new duckdb-vortex results will not show

Main already has duckdb-vortex/results/20261110/ for all five machines of this PR (7f7648d). Those runs use the stock DuckDB 2.0 with INSTALL vortex. The website shows only the latest dated copy of each <system>/<machine> pair (generate-results.sh:57, README § How To Add a New Result). So the 20261105 files stay hidden.

After the merge, the site shows numbers that these scripts no longer produce. On c7a.metal-48xl the hot geomean is 0.196 sec in 20261105 and 0.241 sec in 20261110. Let's re-run duckdb-vortex on this branch, so the newest results come from the Vortex 0.90.0 build.

I ran main's validator on this head: RC 0, no lines for the PR's files. The other added results are the latest for their machines, with 43 rows, no nulls and matching dates.

Automated review by neuroalexey

@alexey-milovidov

Copy link
Copy Markdown
Member Author

The results appeared - let's add.

The rerun of the automated benchmark of #3056
(after the `0.90.0-55.2.0` tag of `vortex-datafusion-cli` was published) produced
results for `datafusion-vortex` and `datafusion-vortex-partitioned` on 7 machines each.
They are added here, and the manually added result files are removed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions

Copy link
Copy Markdown
Contributor

@alexey-milovidov
alexey-milovidov merged commit fac622a into main Nov 15, 2026
3 of 4 checks passed
@alexey-milovidov
alexey-milovidov deleted the vortex-0.90-automated-results branch November 15, 2026 16:41
@alexey-milovidov alexey-milovidov self-assigned this Nov 15, 2026

This branch is waiting to be deployed

1 waiting deployment
benchmark-approval — a280f669 Waiting Nov 15, 2026 by alexey-milovidov via launch #1305
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants