Skip to content

fix(bench): expect stable MSF catalogs during estimate churn #7264

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
kixelated merged 1 commit into from
Oct 30, 2026
Merged

fix(bench): expect stable MSF catalogs during estimate churn #7264

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 9 additions & 7 deletions rs/moq-mux/benches/catalog.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,11 +3,12 @@
//! Every track's jitter rises 1 ms per frame for [`RISES`] frames inside one rate window, then one
//! more frame lands after the window. Each catalog a subscriber receives makes a player update its
//! subscription on every track, so the work fans out as `publishes x tracks x subscribers`. The
//! routine asserts the rate limit holds that to two publishes (the leading and trailing edge)
//! hang case asserts the rate limit holds that to two publishes (the leading and trailing edge)
//! whatever the shape; without it the count would be `RISES x tracks x subscribers`.
//!
//! The MSF reference case publishes renditions in distinct sibling namespaces and converts them
//! back for every subscriber, exercising namespace resolution over both fanout axes.
//! The MSF reference case declares renditions in distinct sibling namespaces and converts the
//! initial catalog back for every subscriber. MSF keeps the estimates it first declared, so this
//! case asserts that measured rises cause no further catalogs or subscription updates.
//! The rate window runs on tokio's paused clock, so the counts are deterministic.
//!
//! The teardown case drops a reserved importer whose declared tracks never got a config, as at the
Expand Down Expand Up @@ -82,7 +83,7 @@ impl Fanout {
waiter: kio::Waiter::noop(),
};
// The initial track set is configuration, not estimate traffic.
fanout.receive();
assert_eq!(fanout.receive(), (1, fanout.tracks.len() * subscribers));
fanout
}

Expand Down Expand Up @@ -125,7 +126,7 @@ impl Fanout {
(publishes, updates)
}

async fn run(mut self) {
async fn run(mut self, expected: usize) {
let (mut publishes, mut updates) = (0, 0);
let mut tally = |(p, u): (usize, usize)| {
publishes += p;
Expand All @@ -142,7 +143,7 @@ impl Fanout {
let shape = self.tracks.len() * self.subscribers.len();
assert_eq!(
(publishes, updates),
(2, 2 * shape),
(expected, expected * shape),
"estimate churn regression with {} tracks x {} subscribers",
self.tracks.len(),
self.subscribers.len()
Expand All @@ -158,6 +159,7 @@ fn bench_estimate_fanout(c: &mut Criterion) {
.unwrap();
let mut group = c.benchmark_group("catalog_estimate_fanout");
for (mode, references) in [("hang", false), ("msf_references", true)] {
let expected = if references { 0 } else { 2 };
for tracks in TRACKS {
for subscribers in SUBSCRIBERS {
group.throughput(Throughput::Elements(RISES * tracks as u64));
Expand All @@ -167,7 +169,7 @@ fn bench_estimate_fanout(c: &mut Criterion) {
|b, &(tracks, subscribers)| {
b.iter_batched(
|| runtime.block_on(Fanout::new(tracks, subscribers, references)),
|fanout| runtime.block_on(fanout.run()),
|fanout| runtime.block_on(fanout.run(expected)),
BatchSize::SmallInput,
);
},
Expand Down
Loading