stream: cache minimum cursor count in broadcast by trivikr · Pull Request #63322 · nodejs/node

trivikr · 2026-05-15T05:44:59Z

Description

This ports the cached minimum-cursor consumer count used by share to
broadcast.

broadcast previously used a dirty min-cursor flag, causing trimming to call
getMinCursor() and scan all consumers whenever the cached minimum might have
changed. Under high fan-out, that can make normal consumer advancement pay an
O(consumers) cost.

This change tracks how many consumers are currently at the cached minimum
cursor. broadcast now only recomputes the minimum when the last such consumer
advances or detaches. The drop-oldest path still recomputes after forcing
lagging cursors forward.

Benchmark

                                                                                       confidence improvement accuracy (*)    (**)   (***)
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=1 api='classic'             *     10.81 %       ±9.94% ±13.96% ±19.78%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=1 api='iter'               **     19.27 %      ±11.61% ±16.16% ±22.56%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=1 api='webstream'         ***     67.64 %       ±9.73% ±13.67% ±19.40%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=2 api='classic'                    5.67 %       ±5.87%  ±8.09% ±11.11%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=2 api='iter'              ***     17.14 %       ±5.43%  ±7.56% ±10.56%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=2 api='webstream'         ***     46.36 %       ±5.99%  ±8.24% ±11.30%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=4 api='classic'             *     12.86 %      ±10.57% ±14.82% ±20.97%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=4 api='iter'              ***     15.57 %       ±4.32%  ±5.96%  ±8.21%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=4 api='webstream'         ***     43.94 %       ±3.79%  ±5.23%  ±7.23%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=1 api='classic'                   4.88 %       ±6.51%  ±8.92% ±12.16%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=1 api='iter'             ***     14.77 %       ±5.67%  ±7.88% ±10.96%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=1 api='webstream'        ***     45.84 %      ±16.70% ±23.78% ±34.49%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=2 api='classic'                   2.65 %       ±4.17%  ±5.74%  ±7.87%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=2 api='iter'              **     26.89 %      ±18.15% ±25.98% ±37.98%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=2 api='webstream'        ***     21.67 %       ±7.80% ±10.93% ±15.42%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=4 api='classic'                   3.63 %       ±3.64%  ±5.02%  ±6.91%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=4 api='iter'              **      6.26 %       ±4.46%  ±6.15%  ±8.44%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=4 api='webstream'        ***     18.26 %       ±5.36%  ±7.60% ±10.93%

Be aware that when doing many comparisons the risk of a false-positive
result increases. In this case, there are 18 comparisons, you can thus
expect the following amount of false-positive results:
  0.90 false positives, when considering a   5% risk acceptance (*, **, ***),
  0.18 false positives, when considering a   1% risk acceptance (**, ***),
  0.02 false positives, when considering a 0.1% risk acceptance (***)

Assisted-by: openai:gpt-5.5

Track how many broadcast consumers share the cached minimum cursor, matching the share implementation. This lets buffer trimming avoid calling getMinCursor() until the last minimum-cursor consumer advances or detaches. Add coverage for fan-out trimming behavior. Signed-off-by: Kamat, Trivikram <16024985+trivikr@users.noreply.github.com> Assisted-by: openai:gpt-5.5

nodejs-github-bot · 2026-05-15T05:45:05Z

Review requested:

@nodejs/streams

trivikr · 2026-05-15T05:49:22Z

I'm not sure why I didn't see improvements in the previous PR in #63262 (comment)
I might have swapped old and new binaries while running that specific benchmark.
It was my code contribution in Node.js core after multiple years.

In this PR, I validated that benchmarks are run correctly

$ ./configure --ninja

$ make JOBS=8

$ ./node benchmark/compare.js \
  --old ../node_old \
  --new ./node \
  --runs 10 \
  --filter iter-throughput-broadcast \
  --no-progress \
  streams > /tmp/broadcast.csv

$ cat /tmp/broadcast.csv | Rscript benchmark/compare.R

trivikr · 2026-05-15T05:53:10Z

Re-run benchmarks with 30 runs

                                                                                       confidence improvement accuracy (*)   (**)   (***)
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=1 api='classic'           ***     10.67 %       ±2.96% ±3.94%  ±5.13%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=1 api='iter'              ***     13.41 %       ±2.64% ±3.51%  ±4.57%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=1 api='webstream'         ***     49.68 %       ±3.89% ±5.18%  ±6.76%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=2 api='classic'           ***      8.68 %       ±3.12% ±4.16%  ±5.43%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=2 api='iter'              ***     11.49 %       ±3.00% ±3.99%  ±5.20%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=2 api='webstream'         ***     33.26 %       ±3.24% ±4.33%  ±5.69%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=4 api='classic'           ***      9.85 %       ±3.28% ±4.38%  ±5.72%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=4 api='iter'              ***     15.16 %       ±3.49% ±4.65%  ±6.05%
streams/iter-throughput-broadcast.js n=5 datasize=1048576 consumers=4 api='webstream'         ***     35.17 %       ±2.90% ±3.86%  ±5.03%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=1 api='classic'           **      4.34 %       ±3.19% ±4.25%  ±5.55%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=1 api='iter'             ***     14.81 %       ±5.90% ±7.90% ±10.37%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=1 api='webstream'        ***     28.71 %       ±2.92% ±3.91%  ±5.13%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=2 api='classic'          ***      6.16 %       ±2.46% ±3.28%  ±4.27%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=2 api='iter'             ***     15.17 %       ±6.29% ±8.43% ±11.11%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=2 api='webstream'        ***     14.40 %       ±3.90% ±5.20%  ±6.78%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=4 api='classic'            *      9.01 %       ±6.96% ±9.31% ±12.21%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=4 api='iter'                      2.17 %       ±2.98% ±3.96%  ±5.16%
streams/iter-throughput-broadcast.js n=5 datasize=16777216 consumers=4 api='webstream'        ***     10.91 %       ±2.35% ±3.13%  ±4.09%

Be aware that when doing many comparisons the risk of a false-positive
result increases. In this case, there are 18 comparisons, you can thus
expect the following amount of false-positive results:
  0.90 false positives, when considering a   5% risk acceptance (*, **, ***),
  0.18 false positives, when considering a   1% risk acceptance (**, ***),
  0.02 false positives, when considering a 0.1% risk acceptance (***)

codecov · 2026-05-15T07:12:57Z

Codecov Report

❌ Patch coverage is 80.43478% with 9 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.04%. Comparing base (2edd842) to head (7476a98).
⚠️ Report is 15 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/internal/streams/iter/broadcast.js	80.43%	9 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main   #63322   +/-   ##
=======================================
  Coverage   90.04%   90.04%           
=======================================
  Files         714      714           
  Lines      225338   225364   +26     
  Branches    42598    42603    +5     
=======================================
+ Hits       202897   202940   +43     
+ Misses      14236    14198   -38     
- Partials     8205     8226   +21

Files with missing lines	Coverage Δ
lib/internal/streams/iter/broadcast.js	`84.41% <80.43%> (-0.26%)`	⬇️

... and 30 files with indirect coverage changes

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

nodejs-github-bot · 2026-05-15T19:34:59Z

CI: https://ci.nodejs.org/job/node-test-pull-request/73475/

nodejs-github-bot added needs-ci PRs that need a full CI run. stream Issues and PRs related to the stream subsystem. labels May 15, 2026

trivikr self-assigned this May 15, 2026

jasnell approved these changes May 15, 2026

View reviewed changes

trivikr added request-ci Add this label to start a Jenkins CI on a PR. author ready PRs that have at least one approval, no pending requests for changes, and a CI started. labels May 15, 2026

github-actions Bot removed the request-ci Add this label to start a Jenkins CI on a PR. label May 15, 2026

This comment was marked as outdated.

Sign in to view

Ethan-Arrowood approved these changes May 15, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

stream: cache minimum cursor count in broadcast#63322

stream: cache minimum cursor count in broadcast#63322
trivikr wants to merge 1 commit into
nodejs:mainfrom
trivikr:stream-iter-broadcast-min-cursor

trivikr commented May 15, 2026

Uh oh!

nodejs-github-bot commented May 15, 2026

Uh oh!

trivikr commented May 15, 2026 •

edited

Loading

Uh oh!

trivikr commented May 15, 2026

Uh oh!

codecov Bot commented May 15, 2026 •

edited

Loading

Uh oh!

This comment was marked as outdated.

nodejs-github-bot commented May 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Uh oh!

Conversation

trivikr commented May 15, 2026

Description

Benchmark

Uh oh!

nodejs-github-bot commented May 15, 2026

Uh oh!

trivikr commented May 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

trivikr commented May 15, 2026

Uh oh!

codecov Bot commented May 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

This comment was marked as outdated.

nodejs-github-bot commented May 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

trivikr commented May 15, 2026 •

edited

Loading

codecov Bot commented May 15, 2026 •

edited

Loading