Add ability to override function behaviour via registry in VortexSession#7588
Open
robert3005 wants to merge 10 commits intodevelopfrom
Open
Add ability to override function behaviour via registry in VortexSession#7588robert3005 wants to merge 10 commits intodevelopfrom
robert3005 wants to merge 10 commits intodevelopfrom
Conversation
robert3005
commented
Apr 22, 2026
robert3005
commented
Apr 22, 2026
robert3005
force-pushed
the
rk/castkernels
branch
from
363a27c to
ea25a77
Compare
joseph-isaacs
added
the
action/benchmark
github-actions
Bot
removed
the
action/benchmark
Contributor
joseph-isaacs
left a comment
joseph-isaacs
left a comment
There was a problem hiding this comment.
let's run benchmarks.
Contributor
Polar Signals Profiling ResultsLatest Run
Previous Runs (1)
Powered by Polar Signals Cloud |
Contributor
Benchmarks: PolarSignals ProfilingVortex (geomean): 0.929x ➖ datafusion / vortex-file-compressed (0.929x ➖, 4↑ 0↓)
|
Contributor
File Sizes: PolarSignals ProfilingNo file size changes detected. |
Contributor
Benchmarks: FineWeb NVMeVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.947x ➖, 1↑ 0↓)
datafusion / vortex-compact (0.973x ➖, 0↑ 0↓)
datafusion / parquet (0.974x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (0.991x ➖, 2↑ 1↓)
duckdb / vortex-compact (0.990x ➖, 0↑ 0↓)
duckdb / parquet (0.946x ➖, 2↑ 0↓)
Full attributed analysis
|
Contributor
File Sizes: FineWeb NVMeNo file size changes detected. |
Contributor
Benchmarks: TPC-H SF=1 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.053x ➖, 0↑ 0↓)
datafusion / vortex-compact (1.049x ➖, 0↑ 0↓)
datafusion / parquet (1.064x ➖, 0↑ 3↓)
datafusion / arrow (1.083x ➖, 0↑ 7↓)
duckdb / vortex-file-compressed (1.063x ➖, 0↑ 2↓)
duckdb / vortex-compact (1.048x ➖, 0↑ 1↓)
duckdb / parquet (1.008x ➖, 3↑ 2↓)
duckdb / duckdb (1.053x ➖, 0↑ 2↓)
Full attributed analysis
|
Contributor
File Sizes: TPC-H SF=1 on NVMENo file size changes detected. |
Contributor
Benchmarks: TPC-DS SF=1 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.015x ➖, 1↑ 2↓)
datafusion / vortex-compact (1.012x ➖, 0↑ 1↓)
datafusion / parquet (1.013x ➖, 1↑ 2↓)
duckdb / vortex-file-compressed (1.001x ➖, 1↑ 1↓)
duckdb / vortex-compact (1.000x ➖, 2↑ 2↓)
duckdb / parquet (1.005x ➖, 0↑ 0↓)
duckdb / duckdb (1.015x ➖, 0↑ 3↓)
Full attributed analysis
|
Contributor
File Sizes: TPC-DS SF=1 on NVMENo file size changes detected. |
Contributor
Benchmarks: FineWeb S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.827x ➖, 2↑ 0↓)
datafusion / vortex-compact (1.024x ➖, 0↑ 1↓)
datafusion / parquet (0.977x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (0.841x ➖, 1↑ 0↓)
duckdb / vortex-compact (0.898x ➖, 0↑ 0↓)
duckdb / parquet (0.949x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: Random AccessVortex (geomean): 0.949x ➖ unknown / unknown (0.959x ➖, 4↑ 0↓)
|
Contributor
Benchmarks: Statistical and Population GeneticsVerdict: No clear signal (low confidence) duckdb / vortex-file-compressed (1.007x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.010x ➖, 0↑ 0↓)
duckdb / parquet (0.992x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: TPC-H SF=10 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.049x ➖, 0↑ 0↓)
datafusion / vortex-compact (1.055x ➖, 0↑ 1↓)
datafusion / parquet (1.039x ➖, 0↑ 0↓)
datafusion / arrow (1.039x ➖, 0↑ 2↓)
duckdb / vortex-file-compressed (1.069x ➖, 0↑ 3↓)
duckdb / vortex-compact (1.068x ➖, 0↑ 2↓)
duckdb / parquet (1.036x ➖, 0↑ 0↓)
duckdb / duckdb (1.038x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
File Sizes: Statistical and Population GeneticsNo file size changes detected. |
Contributor
File Sizes: TPC-H SF=10 on NVMENo file size changes detected. |
Contributor
Benchmarks: TPC-H SF=1 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.911x ➖, 2↑ 1↓)
datafusion / vortex-compact (0.999x ➖, 3↑ 3↓)
datafusion / parquet (1.153x ➖, 0↑ 4↓)
duckdb / vortex-file-compressed (0.975x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.007x ➖, 0↑ 0↓)
duckdb / parquet (0.966x ➖, 0↑ 0↓)
Full attributed analysis
|
Contributor
Benchmarks: Clickbench on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.938x ➖, 4↑ 0↓)
datafusion / parquet (0.955x ➖, 1↑ 0↓)
duckdb / vortex-file-compressed (0.982x ➖, 3↑ 3↓)
duckdb / parquet (0.991x ➖, 0↑ 0↓)
duckdb / duckdb (0.984x ➖, 2↑ 0↓)
Full attributed analysis
|
Contributor
File Sizes: Clickbench on NVMEFile Size Changes (1 files changed, -0.0% overall, 0↑ 1↓)
Totals:
|
Contributor
Benchmarks: CompressionVortex (geomean): 1.013x ➖ unknown / unknown (1.027x ➖, 0↑ 12↓)
|
Contributor
Benchmarks: TPC-H SF=10 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.767x ➖, 7↑ 1↓)
datafusion / vortex-compact (0.942x ➖, 1↑ 1↓)
datafusion / parquet (0.939x ➖, 1↑ 1↓)
duckdb / vortex-file-compressed (0.968x ➖, 0↑ 0↓)
duckdb / vortex-compact (0.922x ➖, 0↑ 0↓)
duckdb / parquet (0.954x ➖, 0↑ 0↓)
Full attributed analysis
|
Signed-off-by: Robert Kruszewski <github@robertk.io>
Signed-off-by: Robert Kruszewski <github@robertk.io>
robert3005
force-pushed
the
rk/castkernels
branch
from
3186c7c to
167b2bb
Compare
github-actions
Bot
removed
the
action/benchmark-sql
Comment on lines
+89
to
+92
| pub fn find_reduce_parent(&self, outer: Id, child: Id) -> Option<Arc<ReduceParentFn>> { | ||
| self.registry | ||
| .find(self.hash_fn_ids(outer, child, FnKind::ReduceParent)) | ||
| } |
Contributor
There was a problem hiding this comment.
There can be more than one?
Comment on lines
+95
to
+98
| pub fn contains_reduce_parent(&self, outer: Id, child: Id) -> bool { | ||
| self.registry | ||
| .contains(self.hash_fn_ids(outer, child, FnKind::ReduceParent)) | ||
| } |
Contributor
There was a problem hiding this comment.
I think we want one registry per find?
| /// clone the current map and swap in a new snapshot, so externally serialize concurrent | ||
| /// [`FnRegistry::register`] calls if every write must be retained. | ||
| #[derive(Debug, Default)] | ||
| pub struct FnRegistry(ArcSwap<HashMap<u64, Arc<dyn Any + Send + Sync>>>); |
Contributor
There was a problem hiding this comment.
Shall we store one per fn type to avoid the downcast to the known type up extraction?
Contributor
|
I really think we want one lookup for either of the kernels types execute/reduce? |
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.
This logic isn't used yet but will be used to allow us to customise behaviour of
functions depending on an integration point, i.e. Datafusion can have it's
casting logic that is different from arrow casting logic while everyone using
vortex can still continue calling
castand not specialize for the enginethey're using
Thing to consider is whether we want require passing session to optimise or whether we should remove the implicit optimise calls and defer them to execute loop
The next pr will replace struct casting logic with Arrow and DF specific
behaviour.
Signed-off-by: Robert Kruszewski github@robertk.io