Tell duckdb if we're blocked on IO - #8068
Closed
myrrc wants to merge 1 commit into
Closed
Conversation
Contributor
Benchmarks: TPC-H SF=10 on S3 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.975x ➖, 2↑ 1↓)
datafusion / vortex-compact / ns (1.044x ➖, 0↑ 2↓)
datafusion / parquet / ns (0.868x ➖, 4↑ 0↓)
duckdb / vortex-file-compressed / ns (1.011x ➖, 0↑ 0↓)
duckdb / vortex-compact / ns (1.034x ➖, 0↑ 0↓)
duckdb / parquet / ns (1.097x ➖, 0↑ 1↓)
|
Contributor
Benchmarks: FineWeb NVMe 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.991x ➖, 0↑ 0↓)
datafusion / vortex-compact / ns (1.007x ➖, 0↑ 0↓)
datafusion / parquet / ns (0.971x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (0.965x ➖, 2↑ 2↓)
duckdb / vortex-compact / ns (1.006x ➖, 0↑ 1↓)
duckdb / parquet / ns (0.957x ➖, 0↑ 0↓)
No file size changes detected. |
Contributor
Benchmarks: FineWeb S3 📖Verdict: No clear signal (environment too noisy confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.898x ➖, 1↑ 0↓)
datafusion / parquet / ns (0.940x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (1.023x ➖, 0↑ 0↓)
duckdb / parquet / ns (1.004x ➖, 1↑ 1↓)
|
Contributor
Benchmarks: TPC-H SF=10 on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.005x ➖, 0↑ 0↓)
datafusion / vortex-compact / ns (1.015x ➖, 0↑ 0↓)
datafusion / parquet / ns (1.007x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (1.040x ➖, 0↑ 0↓)
duckdb / vortex-compact / ns (1.031x ➖, 0↑ 0↓)
duckdb / parquet / ns (0.983x ➖, 2↑ 0↓)
duckdb / duckdb / ns (1.008x ➖, 0↑ 1↓)
No file size changes detected. |
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 (1.013x ➖, 0↑ 1↓)
datafusion / parquet / ns (1.020x ➖, 1↑ 1↓)
duckdb / vortex-file-compressed / ns (1.033x ➖, 0↑ 2↓)
duckdb / parquet / ns (1.014x ➖, 0↑ 0↓)
|
Contributor
Benchmarks: Statistical and Population Genetics 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
duckdb / vortex-file-compressed / ns (1.040x ➖, 0↑ 2↓)
duckdb / parquet / ns (1.001x ➖, 0↑ 0↓)
File Size Changes (2 files changed, -32.3% overall, 0↑ 2↓)
Totals:
|
Contributor
Benchmarks: TPC-DS SF=1 on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.997x ➖, 2↑ 1↓)
datafusion / parquet / ns (1.006x ➖, 0↑ 1↓)
duckdb / vortex-file-compressed / ns (1.060x ➖, 1↑ 22↓)
duckdb / parquet / ns (0.995x ➖, 4↑ 0↓)
File Size Changes (25 files changed, -43.5% overall, 0↑ 25↓)
Totals:
|
Contributor
Benchmarks: Clickbench on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.006x ➖, 0↑ 1↓)
datafusion / parquet / ns (1.000x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (1.033x ➖, 1↑ 7↓)
duckdb / parquet / ns (1.003x ➖, 0↑ 1↓)
File Size Changes (101 files changed, -39.2% overall, 0↑ 101↓)
Totals:
|
Contributor
Benchmarks: TPC-H SF=1 on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.994x ➖, 0↑ 0↓)
datafusion / parquet / ns (0.986x ➖, 2↑ 0↓)
duckdb / vortex-file-compressed / ns (1.024x ➖, 0↑ 2↓)
duckdb / parquet / ns (0.995x ➖, 0↑ 0↓)
File Size Changes (9 files changed, -43.9% overall, 0↑ 9↓)
Totals:
|
myrrc
force-pushed
the
myrrc/duckdb-async-scan
branch
from
May 22, 2026 16:20
ac58d6e to
281de76
Compare
Contributor
Polar Signals Profiling ResultsLatest Run
Previous Runs (1)
Powered by Polar Signals Cloud |
myrrc
added
changelog/fix
myrrc
force-pushed
the
myrrc/duckdb-async-scan
branch
2 times, most recently
from
July 31, 2026 13:21
45e50d1 to
4863391
Compare
Signed-off-by: Mikhail Kot <mikhail@spiraldb.com>
myrrc
force-pushed
the
myrrc/duckdb-async-scan
branch
from
July 31, 2026 13:21
4863391 to
abbfa58
Compare
myrrc
marked this pull request as ready for review
July 31, 2026 13:22
Comment on lines
+544
to
+584
| /// Poll next item without blocking caller thread | ||
| fn poll_next_item(local_state: &mut TableFunctionLocal) -> NextItem { | ||
| if let Some(future) = local_state.pending.as_mut() { | ||
| let mut ctx = Context::from_waker(std::task::Waker::noop()); | ||
| return match future.poll_unpin(&mut ctx) { | ||
| Poll::Ready(item) => { | ||
| local_state.pending = None; | ||
| if let Some(item) = item { | ||
| NextItem::Ready(item) | ||
| } else { | ||
| NextItem::None | ||
| } | ||
| } | ||
| Poll::Pending => NextItem::BlockedOnIO, | ||
| }; | ||
| } | ||
|
|
||
| match local_state.iterator.try_recv() { | ||
| TryRecv::Item(item) => NextItem::Ready(item), | ||
| TryRecv::Closed => NextItem::None, | ||
| TryRecv::Empty => { | ||
| let rx = local_state.iterator.receiver(); | ||
| local_state.pending = Some(Box::pin(async move { rx.recv().await.ok() })); | ||
| NextItem::BlockedOnIO | ||
| } | ||
| } | ||
| } | ||
|
|
||
| /// Finish array receive started by scan() | ||
| pub fn wait_and_fetch(local_state: &mut TableFunctionLocal) { | ||
| if let Some(mut future) = local_state.pending.take() { | ||
| // Duckdb calls this on a background task thread pool which is separate | ||
| // from table function processing pool. This call therefore doesn't | ||
| // block the query worker | ||
| let item = RUNTIME.block_on(&mut future); | ||
| // When this assignment is happening, thread related to "local_state" | ||
| // is suspended. It will be resumed only when wait_and_fetch returns | ||
| // so there's no race condition | ||
| local_state.pending = Some(Box::pin(std::future::ready(item))); | ||
| }; | ||
| } |
Contributor
There was a problem hiding this comment.
this needs very careful explanation.
It also shouldn't be in this mod I would this in the iterator def?
Contributor
Benchmarks: Clickbench Sorted on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.977x ➖, 1↑ 1↓)
datafusion / parquet / ns (1.025x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (1.018x ➖, 0↑ 1↓)
duckdb / parquet / ns (0.994x ➖, 0↑ 0↓)
duckdb / duckdb / ns (1.022x ➖, 0↑ 0↓)
File Size Changes (201 files changed, +0.0% overall, 94↑ 107↓)
Totals:
|
| type ScanItemFuture = BoxFuture<'static, Option<ScanItem>>; | ||
|
|
||
| pub struct TableFunctionGlobal { | ||
| iterator: DataSourceIterator, |
Contributor
There was a problem hiding this comment.
is this used in the scan?
Contributor
Benchmarks: Appian on NVME 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (1.003x ➖, 0↑ 0↓)
datafusion / parquet / ns (0.996x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (0.982x ➖, 1↑ 0↓)
duckdb / parquet / ns (1.001x ➖, 0↑ 0↓)
duckdb / duckdb / ns (0.998x ➖, 0↑ 0↓)
File Size Changes (1 files changed, -0.0% overall, 0↑ 1↓)
Totals:
|
Contributor
Benchmarks: Vortex queries 📖Verdict: No clear signal (low confidence) How to read Verdict and Engines
datafusion / vortex-file-compressed / ns (0.999x ➖, 0↑ 0↓)
datafusion / parquet / ns (1.003x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed / ns (1.028x ➖, 0↑ 0↓)
duckdb / parquet / ns (1.004x ➖, 0↑ 0↓)
No file size changes detected. |
Contributor
Author
|
In addition to the comments ^ benchmarks likely show no impact because we can't really reproduce slow IO even on our S3. I'll think whether we can get a good reproduction where we will see a speedup, otherwise there's still no sense to merge this |
Contributor
Author
|
This PR is superceded by #9355 |
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.
If we're running Duckdb over slow storage (S3), getting ArrayRef's may block. Before this PR, scan function would block until getting an array. Now we can tell Duckdb we're waiting for IO, and it will suspend the worker thread and reschedule it when we get the array.
This doesn't have any impact if we're running over fast storage like NVME.