Uh oh!
There was an error while loading. Please reload this page.
stream: cut promise churn in webstreams hot paths - #65138
Merged
Conversation
nodejs-github-bot
commented
Aug 8, 2026
Collaborator
Review requested:
|
jasnell
reviewed
Aug 8, 2026
Uh oh!
There was an error while loading. Please reload this page.
jasnell
reviewed
Aug 8, 2026
Uh oh!
There was an error while loading. Please reload this page.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@## main #65138 +/- ##
==========================================
+ Coverage 90.30% 90.34% +0.03%
==========================================
Files 751 751 Lines 250235 250330 +95 Branches 47305 47321 +16 ==========================================
+ Hits 225987 226171 +184 + Misses 15619 15557 -62 + Partials 8629 8602 -27
🚀 New features to boost your workflow:
|
bjohansebas
approved these changes
Aug 9, 2026
gurgunday
requested changes
Aug 9, 2026
mcollina
commented
Aug 9, 2026
MemberAuthor
As long as WPT passes, I'm good if we can avoid a promise allocation. That specifically seems non-observable behavior. |
gurgunday
approved these changes
Aug 10, 2026
nodejs-github-bot
commented
Aug 12, 2026
Collaborator
aduh95
reviewed
Aug 12, 2026
Uh oh!
There was an error while loading. Please reload this page.
anonrig
approved these changes
Aug 14, 2026
Contributor
Benchmark GHA (webstreams / pipe-to): https://github.com/nodejs/node/actions/runs/31826854651 Results
Benchmark results:
|
This comment was marked as outdated.
This comment was marked as outdated.
gurgunday
approved these changes
Aug 14, 2026
nodejs-github-bot
commented
Aug 15, 2026
Collaborator
aduh95
approved these changes
Aug 15, 2026
nodejs-github-bot
commented
Aug 15, 2026
Collaborator
Commit Queue failed- Loading data for nodejs/node/pull/65138 ✔ Done loading data for nodejs/node/pull/65138 ----------------------------------- PR info ------------------------------------ Title stream: cut promise churn in webstreams hot paths (#65138) Author Matteo Collina <matteo.collina@gmail.com> (@mcollina) Branch mcollina:webstream-perf-round12 -> nodejs:main Labels author ready, needs-ci, commit-queue, web streams Commits 4 - benchmark: apply highWaterMark in webstreams pipe-to - stream: cut promise churn in webstreams hot paths - stream: consolidate non-op algorithm callbacks - Update lib/internal/webstreams/readablestream.js Committers 2 - Matteo Collina <hello@matteocollina.com> - GitHub <noreply@github.com> PR-URL: https://github.com/nodejs/node/pull/65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com> ------------------------------ Generated metadata ------------------------------ PR-URL: https://github.com/nodejs/node/pull/65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com> -------------------------------------------------------------------------------- ℹ This PR was created on Sat, 08 Aug 2026 15:01:49 GMT ✔ Approvals: 4 ✔ - James M Snell (@jasnell) (TSC): https://github.com/nodejs/node/pull/65138#pullrequestreview-4889126958 ✔ - Gürgün Dayıoğlu (@gurgunday): https://github.com/nodejs/node/pull/65138#pullrequestreview-4941618229 ✔ - Yagiz Nizipli (@anonrig) (TSC): https://github.com/nodejs/node/pull/65138#pullrequestreview-4937946571 ✔ - Antoine du Hamel (@aduh95) (TSC): https://github.com/nodejs/node/pull/65138#pullrequestreview-4943235120 ✔ Last GitHub CI successful ℹ Last Full PR CI on 2026-08-15T04:53:41Z: https://ci.nodejs.org/job/node-test-pull-request/75855/ - Querying data for job/node-test-pull-request/75855/ ✔ Build data downloaded ✔ Last Jenkins CI successful -------------------------------------------------------------------------------- ✔ No git cherry-pick in progress ✔ No git am in progress ✔ No git rebase in progress -------------------------------------------------------------------------------- - Bringing origin/main up to date... From https://github.com/nodejs/node * branch main -> FETCH_HEAD ✔ origin/main is now up-to-date - Downloading patch for 65138 From https://github.com/nodejs/node * branch refs/pull/65138/merge -> FETCH_HEAD ✔ Fetched commits as 9e23066b8af4..0c0d5beaa88d -------------------------------------------------------------------------------- [main eaaf996401] benchmark: apply highWaterMark in webstreams pipe-to Author: Matteo Collina <hello@matteocollina.com> Date: Sat Aug 8 17:00:33 2026 +0200 1 file changed, 4 insertions(+), 6 deletions(-) Auto-merging lib/internal/webstreams/readablestream.js [main 1d82311a65] stream: cut promise churn in webstreams hot paths Author: Matteo Collina <hello@matteocollina.com> Date: Sat Aug 8 17:00:48 2026 +0200 3 files changed, 126 insertions(+), 25 deletions(-) Auto-merging lib/internal/webstreams/readablestream.js [main 6bb61acbad] stream: consolidate non-op algorithm callbacks Author: Matteo Collina <hello@matteocollina.com> Date: Sat Aug 8 19:32:35 2026 +0200 3 files changed, 19 insertions(+), 25 deletions(-) Auto-merging lib/internal/webstreams/readablestream.js [main e67bb74dca] Update lib/internal/webstreams/readablestream.js Author: Matteo Collina <matteo.collina@gmail.com> Date: Fri Aug 14 10:03:20 2026 -0400 1 file changed, 1 insertion(+), 1 deletion(-) ✔ Patches applied There are 4 commits in the PR. Attempting autorebase. (node:400) [DEP0190] DeprecationWarning: Passing args to a child process with shell option true can lead to security vulnerabilities, as the arguments are not escaped, only concatenated. (Use `node --trace-deprecation ...` to show where the warning was created) Rebasing (2/8) Executing: git node land --amend --yes --------------------------------- New Message ---------------------------------- benchmark: apply highWaterMark in webstreams pipe-tohttps://github.com/nodejs/node/actions/runs/31872533736 |
The highWaterMark values were passed as properties of the underlying source and sink dictionaries, where they are ignored: a queuing strategy's highWaterMark is read from the constructors' second argument. Every configuration therefore measured the identical workload at the default highWaterMark of 1, which also explains the historically high run-to-run variance of this benchmark family. Pass the strategies as the constructors' second argument and cover the default (1) alongside buffered (1024, 4096) configurations. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: nodejs#65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
Three related reductions on the per-chunk paths: Wrap user sink.write and source.pull callbacks without coercing their result into a promise. When the callback returns a non-thenable (the common synchronous case), fulfillment is guaranteed and no then() lookup is observable, so the fulfilled reaction is enqueued through a single shared resolved promise at the exact microtask position the coerced promise's reaction would have had, skipping the implicit async-wrapper promise per chunk. Thenable results go through PromiseResolve(), which matches the spec's "a promise resolved with" conversion (identity for native promises). Park pipeTo's pump on backpressure by installing a record that duck-types the writer's lazily-materialized [[readyPromise]] record and whose resolve function is the pump continuation itself. Backpressure clearing then resumes the pump directly instead of materializing a fresh promise record plus reaction per flip, and the pump no longer schedules a microtask per batch. writableStreamUpdateBackpressure publishes the new backpressure state before resolving the ready record so the pump observes the updated value. Replace queueMicrotask() on the pipeTo and tee chunk-forwarding paths with a reaction on the shared resolved promise, which enqueues the continuation at the same position without the per-call scheduling overhead. pipe-to improves by 8-14% across all benchmark configurations, with readable-read and tee also improving in spot runs. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: nodejs#65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
The start, pull, and write non-op algorithms are all raw callbacks with an identical empty body now, so a single shared nonOpCallback replaces nonOpStart, nonOpPull, and nonOpWrite. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: nodejs#65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
aduh95force-pushed
the
webstream-perf-round12
branch
from
August 15, 2026 08:00
0c0d5be to
2929417Compareaduh95
commented
Aug 15, 2026
Contributor
Landed in 4551732...2929417 |
Uh oh!
There was an error while loading. Please reload this page.
aduh95 pushed a commit
that referenced
this pull request
Aug 25, 2026
The highWaterMark values were passed as properties of the underlying source and sink dictionaries, where they are ignored: a queuing strategy's highWaterMark is read from the constructors' second argument. Every configuration therefore measured the identical workload at the default highWaterMark of 1, which also explains the historically high run-to-run variance of this benchmark family. Pass the strategies as the constructors' second argument and cover the default (1) alongside buffered (1024, 4096) configurations. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: #65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
aduh95 pushed a commit
that referenced
this pull request
Aug 25, 2026
Three related reductions on the per-chunk paths: Wrap user sink.write and source.pull callbacks without coercing their result into a promise. When the callback returns a non-thenable (the common synchronous case), fulfillment is guaranteed and no then() lookup is observable, so the fulfilled reaction is enqueued through a single shared resolved promise at the exact microtask position the coerced promise's reaction would have had, skipping the implicit async-wrapper promise per chunk. Thenable results go through PromiseResolve(), which matches the spec's "a promise resolved with" conversion (identity for native promises). Park pipeTo's pump on backpressure by installing a record that duck-types the writer's lazily-materialized [[readyPromise]] record and whose resolve function is the pump continuation itself. Backpressure clearing then resumes the pump directly instead of materializing a fresh promise record plus reaction per flip, and the pump no longer schedules a microtask per batch. writableStreamUpdateBackpressure publishes the new backpressure state before resolving the ready record so the pump observes the updated value. Replace queueMicrotask() on the pipeTo and tee chunk-forwarding paths with a reaction on the shared resolved promise, which enqueues the continuation at the same position without the per-call scheduling overhead. pipe-to improves by 8-14% across all benchmark configurations, with readable-read and tee also improving in spot runs. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: #65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
aduh95 pushed a commit
that referenced
this pull request
Aug 25, 2026
The start, pull, and write non-op algorithms are all raw callbacks with an identical empty body now, so a single shared nonOpCallback replaces nonOpStart, nonOpPull, and nonOpWrite. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: #65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
aduh95 pushed a commit
that referenced
this pull request
Aug 25, 2026
The highWaterMark values were passed as properties of the underlying source and sink dictionaries, where they are ignored: a queuing strategy's highWaterMark is read from the constructors' second argument. Every configuration therefore measured the identical workload at the default highWaterMark of 1, which also explains the historically high run-to-run variance of this benchmark family. Pass the strategies as the constructors' second argument and cover the default (1) alongside buffered (1024, 4096) configurations. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: #65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
aduh95 pushed a commit
that referenced
this pull request
Aug 25, 2026
Three related reductions on the per-chunk paths: Wrap user sink.write and source.pull callbacks without coercing their result into a promise. When the callback returns a non-thenable (the common synchronous case), fulfillment is guaranteed and no then() lookup is observable, so the fulfilled reaction is enqueued through a single shared resolved promise at the exact microtask position the coerced promise's reaction would have had, skipping the implicit async-wrapper promise per chunk. Thenable results go through PromiseResolve(), which matches the spec's "a promise resolved with" conversion (identity for native promises). Park pipeTo's pump on backpressure by installing a record that duck-types the writer's lazily-materialized [[readyPromise]] record and whose resolve function is the pump continuation itself. Backpressure clearing then resumes the pump directly instead of materializing a fresh promise record plus reaction per flip, and the pump no longer schedules a microtask per batch. writableStreamUpdateBackpressure publishes the new backpressure state before resolving the ready record so the pump observes the updated value. Replace queueMicrotask() on the pipeTo and tee chunk-forwarding paths with a reaction on the shared resolved promise, which enqueues the continuation at the same position without the per-call scheduling overhead. pipe-to improves by 8-14% across all benchmark configurations, with readable-read and tee also improving in spot runs. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: #65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
aduh95 pushed a commit
that referenced
this pull request
Aug 25, 2026
The start, pull, and write non-op algorithms are all raw callbacks with an identical empty body now, so a single shared nonOpCallback replaces nonOpStart, nonOpPull, and nonOpWrite. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: #65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
aduh95 pushed a commit
that referenced
this pull request
Aug 27, 2026
The highWaterMark values were passed as properties of the underlying source and sink dictionaries, where they are ignored: a queuing strategy's highWaterMark is read from the constructors' second argument. Every configuration therefore measured the identical workload at the default highWaterMark of 1, which also explains the historically high run-to-run variance of this benchmark family. Pass the strategies as the constructors' second argument and cover the default (1) alongside buffered (1024, 4096) configurations. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: #65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
aduh95 pushed a commit
that referenced
this pull request
Aug 27, 2026
Three related reductions on the per-chunk paths: Wrap user sink.write and source.pull callbacks without coercing their result into a promise. When the callback returns a non-thenable (the common synchronous case), fulfillment is guaranteed and no then() lookup is observable, so the fulfilled reaction is enqueued through a single shared resolved promise at the exact microtask position the coerced promise's reaction would have had, skipping the implicit async-wrapper promise per chunk. Thenable results go through PromiseResolve(), which matches the spec's "a promise resolved with" conversion (identity for native promises). Park pipeTo's pump on backpressure by installing a record that duck-types the writer's lazily-materialized [[readyPromise]] record and whose resolve function is the pump continuation itself. Backpressure clearing then resumes the pump directly instead of materializing a fresh promise record plus reaction per flip, and the pump no longer schedules a microtask per batch. writableStreamUpdateBackpressure publishes the new backpressure state before resolving the ready record so the pump observes the updated value. Replace queueMicrotask() on the pipeTo and tee chunk-forwarding paths with a reaction on the shared resolved promise, which enqueues the continuation at the same position without the per-call scheduling overhead. pipe-to improves by 8-14% across all benchmark configurations, with readable-read and tee also improving in spot runs. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: #65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
aduh95 pushed a commit
that referenced
this pull request
Aug 27, 2026
The start, pull, and write non-op algorithms are all raw callbacks with an identical empty body now, so a single shared nonOpCallback replaces nonOpStart, nonOpPull, and nonOpWrite. Signed-off-by: Matteo Collina <hello@matteocollina.com> PR-URL: #65138 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Gürgün Dayıoğlu <hey@gurgun.day> Reviewed-By: Yagiz Nizipli <yagiz@nizipli.com> Reviewed-By: Antoine du Hamel <duhamelantoine1995@gmail.com>
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 freeto join this conversation on GitHub.
Already have an account?
Sign in to comment
7 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.
Twelfth round of pure-JS webstreams optimizations, following #64890.
Profiling
pipeToat the defaulthighWaterMarkshowed roughly 10% of the profile inqueueMicrotaskplus its native binding (one call per pump batch), a freshwriter.readypromise record plus reaction per backpressure flip on thepipeThroughshape, and an implicit async-wrapper promise persink.write/source.pullinvocation. Three changes, one commit:sink.writeandsource.pullcallbacks are wrapped without coercing their result into a promise. A non-thenable result (the common synchronous case) means fulfillment is guaranteed and nothen()lookup is observable, so the fulfilled reaction is enqueued through a single shared resolved promise at the exact microtask position the coerced promise's reaction would have had. Thenable results go throughPromiseResolve(), matching the reference implementation'spromiseCall(identity for native promises).[[readyPromise]]record;writableStreamUpdateBackpressureresolving it re-enters the pump directly. This removes the per-flip promise record + reaction and the per-batchqueueMicrotask. The backpressure state field is now published before the ready record is resolved so the hook observes the new value (the resolve of a real ready record only settles a promise, so the reorder is unobservable otherwise).forwardChunkhops and the pump's between-batch yield use a reaction on the shared resolved promise instead ofqueueMicrotask, which enqueues at the same position with less overhead.A separate first commit fixes the
pipe-to.jsbenchmark: thehighWaterMarkvalues were passed inside the underlying source/sink dictionaries where they are ignored, so all 16 configurations measured the identical workload at the defaulthighWaterMarkof 1. The strategies are now passed as the constructors' second argument, with the matrix covering the default (1) and buffered (1024, 4096) configurations.Benchmark results with the fixed benchmark (30 runs):
The full-suite run showed no regressions in any other family;
readable-read normal,tee normal, andpipeThroughpassthrough spot runs also improve (~+6-18%). Verified with the WPT streams/compression/encoding suites, the full parallel webstream/whatwg test set, and a shutdown-ordering stress (abort mid-write, close with pending writes, sync-throwing and rejecting sinks, error propagation) whose event log is byte-identical tomain.