perf(physical-plan): optimize byte view append#21586
Open
kumarUjjawal wants to merge 3 commits intoapache:mainfrom
Open
perf(physical-plan): optimize byte view append#21586kumarUjjawal wants to merge 3 commits intoapache:mainfrom
kumarUjjawal wants to merge 3 commits intoapache:mainfrom
Conversation
martin-g
reviewed
Apr 13, 2026
| } | ||
|
|
||
| let start_idx = self.views.len(); | ||
| self.views.extend(rows.iter().map(|&row| source_views[row])); |
Member
There was a problem hiding this comment.
This could be moved at line 202 and this will re-use the iteration over rows.
| let start_idx = self.views.len(); | ||
| self.views.extend(rows.iter().map(|&row| source_views[row])); | ||
|
|
||
| let mut pending = Vec::with_capacity(rows.len()); |
Member
There was a problem hiding this comment.
This most probably will over-allocate because only the rows with more than 12 bytes will be appended. Maybe use half of the rows length as initial capacity and let it grow if needed ?!
| rows: &[usize], | ||
| ) { | ||
| let source_views = array.views(); | ||
| let mut pending = Vec::with_capacity(rows.len()); |
Member
There was a problem hiding this comment.
Same - consider using a smaller initial capacity.
Contributor
Author
There was a problem hiding this comment.
I think half or rows.len() would be good as you said.
Contributor
|
run benchmarks |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing feat/vectorized_append (bd0a959) to 98d280f (merge-base) diff using: clickbench_partitioned File an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing feat/vectorized_append (bd0a959) to 98d280f (merge-base) diff using: tpch File an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing feat/vectorized_append (bd0a959) to 98d280f (merge-base) diff using: tpcds File an issue against this benchmark runner |
|
🤖 Benchmark completed (GKE) | trigger Instance: CPU Details (lscpu)Details
Resource Usagetpcds — base (merge-base)
tpcds — branch
File an issue against this benchmark runner |
|
🤖 Benchmark completed (GKE) | trigger Instance: CPU Details (lscpu)Details
Resource Usageclickbench_partitioned — base (merge-base)
clickbench_partitioned — branch
File an issue against this benchmark runner |
martin-g
approved these changes
Apr 18, 2026
alamb
reviewed
Apr 19, 2026
Contributor
alamb
left a comment
alamb
left a comment
There was a problem hiding this comment.
Thanks for this @kumarUjjawal and @martin-g
Do we have benchmarks that show it making the operation faster? The bot benchmark runner results seem mixed
I am also surprised that this would be faster given how much extra work it is doing (allocating another buffer, and a bunch of calculations in batch_copy_long_views) 🤔
| for (idx, &row) in rows.iter().enumerate() { | ||
| let view = source_views[row]; | ||
| self.views.push(view); | ||
| if (view as u32) > 12 { |
Contributor
There was a problem hiding this comment.
we culd replace 12 with MAX_INLINE_VIEW_LEN maybe to make it clearer what is going on here
| } | ||
|
|
||
| let start_idx = self.views.len(); | ||
| let mut pending = Vec::with_capacity(rows.len().saturating_add(1) / 2); |
Contributor
There was a problem hiding this comment.
You could probably save this allocation by using some sort of temp scratch space on ByteViewGroupValueBuilder
Something lik
pub struct ByteViewGroupValueBuilder<B: ByteViewType> {
...
scratch: Vec<PendingByteViewCopy>,}
So then it wouldn't reallocate on each batch
Contributor
There was a problem hiding this comment.
I don't think we need to have this scratch space
|
|
||
| const BYTE_VIEW_MAX_BLOCK_SIZE: usize = 2 * 1024 * 1024; | ||
|
|
||
| #[derive(Clone, Copy)] |
Contributor
There was a problem hiding this comment.
I think it would help to add some explaining what this is and what it is used for so a future reader doesn't have to read the rest of the code to figure it out
| self.batch_copy_long_views(array.data_buffers(), &pending); | ||
| } | ||
|
|
||
| fn batch_copy_long_views( |
Contributor
There was a problem hiding this comment.
it would help me to understand this if it has some more comments explaining the rationale for postponing the copy (to detect contiguous views?)
Dandandan
reviewed
Apr 19, 2026
| let mut pending = Vec::with_capacity(rows.len().saturating_add(1) / 2); | ||
| self.views.reserve(rows.len()); | ||
|
|
||
| for &row in rows { |
Contributor
There was a problem hiding this comment.
We should use extend more in this code path which reduces capacity checks / improves code gen.
It also reduces reallocations
Dandandan
requested changes
Apr 19, 2026
Contributor
Dandandan
left a comment
Dandandan
left a comment
There was a problem hiding this comment.
I think we probably need some good benchmark first / or show improvement on the aggregation benchmarks.
Contributor
Author
|
@Dandandan Can you run the benchmark |
Contributor
|
run benchmarks |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing feat/vectorized_append (1fed6b5) to 98d280f (merge-base) diff using: tpcds File an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing feat/vectorized_append (1fed6b5) to 98d280f (merge-base) diff using: tpch File an issue against this benchmark runner |
|
🤖 Benchmark running (GKE) | trigger CPU Details (lscpu)Comparing feat/vectorized_append (1fed6b5) to 98d280f (merge-base) diff using: clickbench_partitioned File an issue against this benchmark runner |
|
Benchmark for this request failed. Last 20 lines of output: Click to expandFile an issue against this benchmark runner |
|
🤖 Benchmark completed (GKE) | trigger Instance: CPU Details (lscpu)Details
Resource Usageclickbench_partitioned — base (merge-base)
clickbench_partitioned — branch
File an issue against this benchmark runner |
|
🤖 Benchmark completed (GKE) | trigger Instance: CPU Details (lscpu)Details
Resource Usagetpcds — base (merge-base)
tpcds — branch
File an issue against this benchmark runner |
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
5 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.
Which issue does this PR close?
Rationale for this change
ByteViewGroupValueBuilder::vectorized_appenddid extraper-value work. It rebuilt views and copied long values one by one even when the input already had reusable byte views.What changes are included in this PR?
vectorized_appendextendon the no-null fast pathtake_nafter vectorized appendAre these changes tested?
Yes
Are there any user-facing changes?
No