feat(transport/cubic): cubic region RFC9438 updates by omansfeld · Pull Request #2983 · mozilla/neqo

omansfeld · 2025-09-15T13:10:54Z

comments and docs
~~using max_datagram_size_f64 across functions~~
- (this was already merged in refactor(transport/cubic): only convert max_datagram_size_f64 once #2994)
~~using saturating_duration_since to get time since epoch start~~
- (this was merged in refactor(transport/cubic): use saturating_duration_since for t #2997)
new target formula clamping to min and max value (cwnd<=target<=cwnd*1.5)
removing EXPONENTIAL_GROWTH_REDUCTION since it's use is now covered by the max value mentioned above

Split off of #2535.
Part of #2967.

EDIT 2025-10-13:

I rebased after parts of this have already landed (see above checklist). This PR now contains the last updates to the cubic region and should be ready for review.

In the old code we'd also apply the clamping to 1.5 * curr_cwnd to the reno region, since it was applied to the return value, not to the cubic region calculation result. Since this clamping is supposed to prevent exponential growth ("less than the increase rate of slow start" - RFC9438) it doesn't need to be applied to the reno region, which is growing additively per design.

codecov · 2025-09-15T13:15:48Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.38%. Comparing base (b90ded6) to head (920de9a).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2983      +/-   ##
==========================================
- Coverage   93.41%   93.38%   -0.04%     
==========================================
  Files         124      124              
  Lines       35969    35970       +1     
  Branches    35969    35970       +1     
==========================================
- Hits        33602    33591      -11     
- Misses       1523     1535      +12     
  Partials      844      844

Components	Coverage Δ
neqo-common	`97.50% <ø> (ø)`
neqo-crypto	`83.25% <ø> (-0.48%)`	⬇️
neqo-http3	`93.34% <ø> (ø)`
neqo-qpack	`94.18% <ø> (ø)`
neqo-transport	`94.46% <100.00%> (-0.02%)`	⬇️
neqo-udp	`79.32% <ø> (+0.48%)`	⬆️
mtu	`85.57% <ø> (-0.20%)`	⬇️

Copilot

Pull Request Overview

This PR updates the CUBIC congestion control implementation with significant refinements to target calculation and documentation improvements. The changes enhance the algorithm's adherence to RFC 9438 specifications while removing obsolete constants.

Updated documentation and comments to reference RFC 9438 instead of RFC 8312
Implemented new target formula with clamping between cwnd and cwnd*1.5
Removed EXPONENTIAL_GROWTH_REDUCTION constant in favor of the new target clamping approach

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

neqo-transport/src/cc/cubic.rs

Copilot

Pull Request Overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 2 comments.

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

neqo-transport/src/cc/cubic.rs

mxinden

Well done! Thank you.

neqo-transport/src/cc/cubic.rs

mxinden · 2025-10-13T13:31:00Z

Once the latest congestion control changes landed in Nightly, we should closely monitor packet loss:

https://glam.telemetry.mozilla.org/firefox/probe/http3_loss_ratio/explore?activeBuckets=%5B%220%22%2C%221%22%2C%222%22%2C%224%22%2C%223%22%2C%225%22%5D&process=parent&timeHorizon=ALL&visiblePercentiles=%5B99.9%2C99%2C95%2C75%2C50%5D

mxinden · 2025-10-13T13:33:20Z

In the old code we'd also apply the clamping to 1.5 * curr_cwnd to the reno region, since it was applied to the return value, not to the cubic region calculation result. Since this clamping is supposed to prevent exponential growth ("less than the increase rate of slow start" - RFC9438) it doesn't need to be applied to the reno region, which is growing additively per design.

Worth adding a debug_assert asserting that the Reno target never grows exponentially?

github-actions · 2025-10-13T14:08:18Z

Bencher Report

Branch	cubic_region_updates
Testbed	On-prem

Click to view all benchmark results

Benchmark	Latency	Benchmark Result nanoseconds (ns) (Result Δ%)	Upper Boundary nanoseconds (ns) (Limit %)
1-conn/1-100mb-req/mtu-1504 (aka. Upload)/client	📈 view plot 🚷 view threshold	205,030,000.00 ns (-1.74%) Baseline: 208,655,425.53 ns	218,012,106.18 ns (94.05%)
1-conn/1-100mb-resp/mtu-1504 (aka. Download)/client	📈 view plot 🚷 view threshold	199,230,000.00 ns (-1.69%) Baseline: 202,660,797.87 ns	212,889,465.14 ns (93.58%)
1-conn/1-1b-resp/mtu-1504 (aka. HPS)/client	📈 view plot 🚷 view threshold	28,498,000.00 ns (+0.29%) Baseline: 28,415,441.49 ns	28,858,220.30 ns (98.75%)
1-conn/10_000-parallel-1b-resp/mtu-1504 (aka. RPS)/client	📈 view plot 🚷 view threshold	282,720,000.00 ns (-3.92%) Baseline: 294,259,521.28 ns	306,283,616.21 ns (92.31%)
1-streams/each-1000-bytes/simulated-time	📈 view plot 🚷 view threshold	119,050,000.00 ns (+0.58%) Baseline: 118,357,978.72 ns	120,888,615.11 ns (98.48%)
1-streams/each-1000-bytes/wallclock-time	📈 view plot 🚷 view threshold	591,170.00 ns (-1.09%) Baseline: 597,663.46 ns	622,601.76 ns (94.95%)
1000-streams/each-1-bytes/simulated-time	📈 view plot 🚷 view threshold	14,996,000,000.00 ns (+0.03%) Baseline: 14,992,058,510.64 ns	15,010,807,004.61 ns (99.90%)
1000-streams/each-1-bytes/wallclock-time	📈 view plot 🚷 view threshold	13,803,000.00 ns (-2.76%) Baseline: 14,194,686.17 ns	14,973,867.89 ns (92.18%)
1000-streams/each-1000-bytes/simulated-time	📈 view plot 🚷 view threshold	19,061,000,000.00 ns (+0.74%) Baseline: 18,921,611,702.13 ns	19,189,763,596.48 ns (99.33%)
1000-streams/each-1000-bytes/wallclock-time	📈 view plot 🚷 view threshold	48,754,000.00 ns (-6.45%) Baseline: 52,117,143.62 ns	58,599,715.44 ns (83.20%)
RxStreamOrderer::inbound_frame()	📈 view plot 🚷 view threshold	107,120,000.00 ns (-2.43%) Baseline: 109,791,595.74 ns	111,981,697.30 ns (95.66%)
coalesce_acked_from_zero 1+1 entries	📈 view plot 🚷 view threshold	88.66 ns (+0.04%) Baseline: 88.62 ns	89.29 ns (99.29%)
coalesce_acked_from_zero 10+1 entries	📈 view plot 🚷 view threshold	105.63 ns (-0.42%) Baseline: 106.08 ns	107.08 ns (98.65%)
coalesce_acked_from_zero 1000+1 entries	📈 view plot 🚷 view threshold	88.96 ns (-0.95%) Baseline: 89.82 ns	94.42 ns (94.21%)
coalesce_acked_from_zero 3+1 entries	📈 view plot 🚷 view threshold	106.39 ns (-0.19%) Baseline: 106.59 ns	107.55 ns (98.92%)
decode 1048576 bytes, mask 3f	📈 view plot 🚷 view threshold	1,590,600.00 ns (-0.12%) Baseline: 1,592,520.21 ns	1,599,563.62 ns (99.44%)
decode 1048576 bytes, mask 7f	📈 view plot 🚷 view threshold	5,059,500.00 ns (+0.04%) Baseline: 5,057,326.06 ns	5,076,842.39 ns (99.66%)
decode 1048576 bytes, mask ff	📈 view plot 🚷 view threshold	3,033,100.00 ns (+0.05%) Baseline: 3,031,698.94 ns	3,043,449.39 ns (99.66%)
decode 4096 bytes, mask 3f	📈 view plot 🚷 view threshold	8,317.10 ns (+0.25%) Baseline: 8,296.37 ns	8,343.26 ns (99.69%)
decode 4096 bytes, mask 7f	📈 view plot 🚷 view threshold	20,047.00 ns (+0.19%) Baseline: 20,008.22 ns	20,087.14 ns (99.80%)
decode 4096 bytes, mask ff	📈 view plot 🚷 view threshold	11,656.00 ns (-0.60%) Baseline: 11,726.27 ns	11,970.14 ns (97.38%)
sent::Packets::take_ranges	📈 view plot 🚷 view threshold	4,668.30 ns (-1.67%) Baseline: 4,747.42 ns	4,986.14 ns (93.63%)
transfer/pacing-false/same-seed/simulated-time/run	📈 view plot 🚷 view threshold	25,710,000,000.00 ns (+1.72%) Baseline: 25,275,000,000.00 ns	25,817,812,451.62 ns (99.58%)
transfer/pacing-false/same-seed/wallclock-time/run	📈 view plot 🚷 view threshold	26,156,000.00 ns (+0.63%) Baseline: 25,993,139.78 ns	27,070,349.68 ns (96.62%)
transfer/pacing-false/varying-seeds/simulated-time/run	📈 view plot 🚷 view threshold	25,176,000,000.00 ns (+0.03%) Baseline: 25,168,268,817.20 ns	25,215,941,018.90 ns (99.84%)
transfer/pacing-false/varying-seeds/wallclock-time/run	📈 view plot 🚷 view threshold	25,500,000.00 ns (-2.66%) Baseline: 26,197,709.68 ns	27,647,732.39 ns (92.23%)
transfer/pacing-true/same-seed/simulated-time/run	📈 view plot 🚷 view threshold	25,675,000,000.00 ns (+0.26%) Baseline: 25,607,177,419.35 ns	25,691,809,468.26 ns (99.93%)
transfer/pacing-true/same-seed/wallclock-time/run	📈 view plot 🚷 view threshold	27,187,000.00 ns (-0.76%) Baseline: 27,395,930.11 ns	28,804,606.05 ns (94.38%)
transfer/pacing-true/varying-seeds/simulated-time/run	📈 view plot 🚷 view threshold	24,959,000,000.00 ns (-0.14%) Baseline: 24,994,043,010.75 ns	25,044,262,140.71 ns (99.66%)
transfer/pacing-true/varying-seeds/wallclock-time/run	📈 view plot 🚷 view threshold	26,007,000.00 ns (-2.68%) Baseline: 26,724,134.41 ns	28,219,820.17 ns (92.16%)

🐰 View full continuous benchmarking report in Bencher

omansfeld · 2025-10-13T14:53:09Z

In the old code we'd also apply the clamping to 1.5 * curr_cwnd to the reno region, since it was applied to the return value, not to the cubic region calculation result. Since this clamping is supposed to prevent exponential growth ("less than the increase rate of slow start" - RFC9438) it doesn't need to be applied to the reno region, which is growing additively per design.

Worth adding a debug_assert asserting that the Reno target never grows exponentially?

I don't think it's necessary, as it really mathematically cannot happen (unless with unrealistically small cwnd values).

Just as an example, if we had a cwnd of 10 * mss (which is already really small) we would need to ack 50 * mss bytes to reach the curr_cwnd * 1.5 threshold. Which cannot really happen, because the congestion window is only 10 * mss. That is with the increased CUBIC_ALPHA = 1.0 which we currently don't even implement. And the magnitude of acked bytes needed for this gets bigger the bigger the cwnd is. Should be something like curr_cwnd^2 / 2 acked bytes needed to reach the threshold.

That said, I'm not opposed to putting in a debug_assert there. You said "worth it" implying there is some downside to having it?

mxinden · 2025-10-13T15:13:36Z

That said, I'm not opposed to putting in a debug_assert there. You said "worth it" implying there is some downside to having it?

Just one more line of code. No other downside.

neqo-transport/src/cc/cubic.rs

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

neqo-transport/src/cc/cubic.rs

github-actions · 2025-10-13T16:49:17Z

Bencher Report

Branch	cubic_region_updates
Testbed	On-prem

🚨 1 Alert

Iteration	Benchmark	Measure Units	View	Benchmark Result (Result Δ%)	Upper Boundary (Limit %)
9	neqo vs. s2n (cubic, paced)	Latency milliseconds (ms)	📈 plot 🚷 threshold 🚨 alert (🔔)	226.24 ms (+2.44%) Baseline: 220.85 ms	223.80 ms (101.09%)

Click to view all benchmark results

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
google vs. neqo (cubic, paced)	📈 view plot 🚷 view threshold	275.35 ms (-0.84%) Baseline: 277.68 ms	280.33 ms (98.23%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
msquic vs. neqo (cubic, paced)	📈 view plot 🚷 view threshold	196.83 ms (+1.33%) Baseline: 194.24 ms	227.67 ms (86.45%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
neqo vs. google (cubic, paced)	📈 view plot 🚷 view threshold	757.64 ms (+0.00%) Baseline: 757.63 ms	764.85 ms (99.06%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
neqo vs. msquic (cubic, paced)	📈 view plot 🚷 view threshold	159.64 ms (+1.50%) Baseline: 157.27 ms	159.70 ms (99.96%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
neqo vs. neqo (cubic)	📈 view plot 🚷 view threshold	89.32 ms (-1.44%) Baseline: 90.63 ms	94.43 ms (94.59%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
neqo vs. neqo (cubic, paced)	📈 view plot 🚷 view threshold	90.10 ms (-1.90%) Baseline: 91.84 ms	95.56 ms (94.29%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
neqo vs. neqo (reno)	📈 view plot 🚷 view threshold	89.47 ms (-1.23%) Baseline: 90.59 ms	93.95 ms (95.24%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
neqo vs. neqo (reno, paced)	📈 view plot 🚷 view threshold	93.08 ms (+1.37%) Baseline: 91.82 ms	95.42 ms (97.54%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
neqo vs. quiche (cubic, paced)	📈 view plot 🚷 view threshold	193.10 ms (-0.35%) Baseline: 193.77 ms	197.23 ms (97.91%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
neqo vs. s2n (cubic, paced)	📈 view plot 🚷 view threshold 🚨 view alert (🔔)	226.24 ms (+2.44%) Baseline: 220.85 ms	223.80 ms (101.09%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
quiche vs. neqo (cubic, paced)	📈 view plot 🚷 view threshold	155.72 ms (+2.05%) Baseline: 152.59 ms	158.25 ms (98.41%)

Benchmark	Latency	Benchmark Result milliseconds (ms) (Result Δ%)	Upper Boundary milliseconds (ms) (Limit %)
s2n vs. neqo (cubic, paced)	📈 view plot 🚷 view threshold	175.94 ms (+1.15%) Baseline: 173.94 ms	178.08 ms (98.80%)

🐰 View full continuous benchmarking report in Bencher

Copilot

Pull Request Overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.

Comments suppressed due to low confidence (1)

neqo-transport/src/cc/cubic.rs:1

The phrase should be 'no more than' instead of 'less than' to accurately reflect the RFC constraint that CUBIC's increase rate should not exceed slow start's rate.

// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

neqo-transport/src/cc/cubic.rs

- comments and docs - new `target` formula clamping to min and max value (`cwnd<=target<=cwnd*1.5`) - removing `EXPONENTIAL_GROWTH_REDUCTION` since it's use is now covered by the max value mentioned above

- linux reference links - renaming variables - `debug_assert!` to check exponential growth of `w_est`

- comments - moved debug assert

- add predicate for small cwnd values - move assert and change the message

mxinden

For others, debug_assert removed after out-of-band discussion. RFC does not require the limit on the Reno region. Assumption is, that it is fine to follow the RFC only.

github-actions · 2025-10-14T15:52:12Z

Failed Interop Tests

QUIC Interop Runner, client vs. server, differences relative to 791fd40.

neqo-latest as client

neqo-latest vs. aioquic: ⚠️C1
neqo-latest vs. go-x-net: BP BA
neqo-latest vs. haproxy: 🚀C1 BP BA
neqo-latest vs. kwik: ⚠️BP BA
neqo-latest vs. linuxquic: L1 C1
neqo-latest vs. lsquic: E L1 C1
neqo-latest vs. msquic: R ⚠️Z A L1 C1
neqo-latest vs. mvfst: A 🚀C1 ⚠️L1
neqo-latest vs. neqo: A
neqo-latest vs. neqo-latest: A
neqo-latest vs. nginx: BP BA
neqo-latest vs. ngtcp2: E CM
neqo-latest vs. picoquic: Z E A L1
neqo-latest vs. quic-go: A
neqo-latest vs. quiche: BP BA
neqo-latest vs. quinn: A
neqo-latest vs. s2n-quic: E 🚀BP BA CM
neqo-latest vs. tquic: S BP BA
neqo-latest vs. xquic: H DC LR C20 M R Z 3 B U A L1 L2 C1 C2 6 BP BA

neqo-latest as server

All results

Succeeded Interop Tests

QUIC Interop Runner, client vs. server

neqo-latest as client

neqo-latest vs. aioquic: H DC LR C20 M S R Z 3 B U A L1 L2 ⚠️C1 C2 6 V2 BP BA
neqo-latest vs. go-x-net: H DC LR M B U A L2 C2 6
neqo-latest vs. haproxy: H DC LR C20 M S R Z 3 B U A L1 L2 🚀C1 C2 6 V2
neqo-latest vs. kwik: 🚀~~H DC LR C20 M S R Z 3 B U A L1 L2 C1 C2 6 V2~~
neqo-latest vs. linuxquic: H DC LR C20 M S R Z 3 B U E A L2 C2 6 V2 BP BA CM
neqo-latest vs. lsquic: H DC LR C20 M S R Z 3 B U A L2 C2 6 V2 BP BA CM
neqo-latest vs. msquic: H DC LR C20 M S ⚠️Z B U L2 C2 6 V2 BP BA
neqo-latest vs. mvfst: H DC LR M R Z 3 B U ⚠️L1 L2 🚀C1 C2 6 BP BA
neqo-latest vs. neqo: H DC LR C20 M S R Z 3 B U E L1 L2 C1 C2 6 V2 BP BA CM
neqo-latest vs. neqo-latest: H DC LR C20 M S R Z 3 B U E L1 L2 C1 C2 6 V2 BP BA CM
neqo-latest vs. nginx: H DC LR C20 M S R Z 3 B U A L1 L2 C1 C2 6
neqo-latest vs. ngtcp2: H DC LR C20 M S R Z 3 B U A L1 L2 C1 C2 6 V2 BP BA
neqo-latest vs. picoquic: H DC LR C20 M S R 3 B U L2 C1 C2 6 V2 BP BA
neqo-latest vs. quic-go: H DC LR C20 M S R Z 3 B U L1 L2 C1 C2 6 BP BA
neqo-latest vs. quiche: H DC LR C20 M S R Z 3 B U A L1 L2 C1 C2 6
neqo-latest vs. quinn: H DC LR C20 M S R Z 3 B U E L1 L2 C1 C2 6 BP BA
neqo-latest vs. s2n-quic: H DC LR C20 M S R 3 B U A L1 L2 C1 C2 6 🚀BP
neqo-latest vs. tquic: H DC LR C20 M R Z 3 B U A L1 L2 C1 C2 6

neqo-latest as server

aioquic vs. neqo-latest: H DC LR C20 M S R Z 3 B U A L1 L2 C1 C2 6 V2 BP BA
chrome vs. neqo-latest: 3
go-x-net vs. neqo-latest: H DC LR M B U A L2 C2 6 BP BA
kwik vs. neqo-latest: H DC LR C20 M S R Z 3 B U A L1 L2 C1 C2 6 V2
linuxquic vs. neqo-latest: H DC LR C20 M S R Z 3 B U E A L1 L2 C1 C2 6 V2 BP BA CM
lsquic vs. neqo-latest: H DC LR C20 M S R 3 B E A ⚠️L1 L2 C1 C2 6 V2 BP BA CM
msquic vs. neqo-latest: H DC LR C20 M S R Z B A L1 L2 C1 C2 6 V2 BP BA
mvfst vs. neqo-latest: H DC LR M 3 B L2 C2 6 BP BA
neqo vs. neqo-latest: H DC LR C20 M S R Z 3 B U E L1 L2 C1 C2 6 V2 BP BA CM
ngtcp2 vs. neqo-latest: H DC LR C20 M S R Z 3 B U E A L1 L2 C1 C2 6 V2 BP BA CM
openssl vs. neqo-latest: H DC C20 S R 3 B L2 C2 6 BP BA
picoquic vs. neqo-latest: H DC LR C20 M S R Z 3 B U E A L1 L2 C1 C2 6 V2 BP BA CM
quic-go vs. neqo-latest: H DC LR C20 M S R Z 3 B U A L1 L2 C1 C2 6 BP BA
quiche vs. neqo-latest: H DC LR M S R Z 3 B A L1 L2 C1 C2 6 BP BA
quinn vs. neqo-latest: H DC LR C20 M S R Z 3 B U E A ⚠️L1 L2 C1 C2 6 BP BA
s2n-quic vs. neqo-latest: H DC LR M S R 3 B E A L1 L2 C1 C2 6 BP BA
tquic vs. neqo-latest: H DC LR M S R Z 3 B A L1 L2 C1 C2 6 BP BA
xquic vs. neqo-latest: H DC LR C20 S R Z 3 B U A L1 L2 C1 C2 6 BP BA

Unsupported Interop Tests

QUIC Interop Runner, client vs. server

neqo-latest as client

neqo-latest vs. aioquic: E CM
neqo-latest vs. go-x-net: C20 S R Z 3 E L1 C1 V2 CM
neqo-latest vs. haproxy: E CM
neqo-latest vs. kwik: E CM
neqo-latest vs. msquic: 3 E CM
neqo-latest vs. mvfst: C20 S E V2 CM
neqo-latest vs. nginx: E V2 CM
neqo-latest vs. picoquic: CM
neqo-latest vs. quic-go: E V2 CM
neqo-latest vs. quiche: E V2 CM
neqo-latest vs. quinn: V2 CM
neqo-latest vs. s2n-quic: Z V2
neqo-latest vs. tquic: E V2 CM
neqo-latest vs. xquic: S E V2 CM

neqo-latest as server

aioquic vs. neqo-latest: E
chrome vs. neqo-latest: H DC LR C20 M S R Z B U E A L1 L2 C1 C2 6 V2 BP BA CM
go-x-net vs. neqo-latest: C20 S R Z 3 E L1 C1 V2
kwik vs. neqo-latest: E
lsquic vs. neqo-latest: Z U
msquic vs. neqo-latest: 3 E
mvfst vs. neqo-latest: C20 S R U E V2
openssl vs. neqo-latest: Z U E L1 C1 V2
quic-go vs. neqo-latest: E V2
quiche vs. neqo-latest: C20 U E V2
s2n-quic vs. neqo-latest: C20 Z U V2
tquic vs. neqo-latest: C20 U E V2
xquic vs. neqo-latest: E V2

github-actions · 2025-10-14T15:56:45Z

Client/server transfer results

Performance differences relative to b90ded6.

Transfer of 33554432 bytes over loopback, min. 100 runs. All unit-less numbers are in milliseconds.

Client vs. server (params)	Mean ± σ	Min	Max	MiB/s ± σ	Δ `main`	Δ `main`
google vs. google	457.0 ± 5.2	448.6	466.5	70.0 ± 6.2
google vs. neqo (cubic, paced)	275.4 ± 4.5	266.5	283.8	116.2 ± 7.1	-0.2	-0.1%
msquic vs. msquic	177.6 ± 45.5	137.6	354.4	180.2 ± 0.7
msquic vs. neqo (cubic, paced)	196.8 ± 47.0	141.5	406.6	162.6 ± 0.7	-2.5	-1.3%
neqo vs. google (cubic, paced)	757.6 ± 4.5	749.3	773.6	42.2 ± 7.1	💔 2.3	0.3%
neqo vs. msquic (cubic, paced)	159.6 ± 8.4	149.4	186.4	200.5 ± 3.8	1.3	0.8%
neqo vs. neqo (cubic)	89.3 ± 4.7	79.7	106.0	358.3 ± 6.8	-0.5	-0.5%
neqo vs. neqo (cubic, paced)	90.1 ± 5.2	81.7	106.7	355.2 ± 6.2	💔 1.5	1.7%
neqo vs. neqo (reno)	89.5 ± 4.5	78.9	104.7	357.6 ± 7.1	-0.8	-0.9%
neqo vs. neqo (reno, paced)	93.1 ± 5.0	82.4	108.4	343.8 ± 6.4	1.4	1.5%
neqo vs. quiche (cubic, paced)	193.1 ± 4.6	186.4	200.8	165.7 ± 7.0	💔 1.4	0.8%
neqo vs. s2n (cubic, paced)	226.2 ± 4.5	215.2	237.4	141.4 ± 7.1	💔 6.0	2.7%
quiche vs. neqo (cubic, paced)	155.7 ± 4.6	146.4	166.9	205.5 ± 7.0	💔 1.8	1.1%
quiche vs. quiche	145.6 ± 5.1	137.8	161.6	219.7 ± 6.3
s2n vs. neqo (cubic, paced)	175.9 ± 4.9	166.1	187.2	181.9 ± 6.5	0.0	0.0%
s2n vs. s2n	254.1 ± 30.9	230.8	353.2	125.9 ± 1.0

Download data for profiler.firefox.com or download performance comparison data.

github-actions · 2025-10-14T16:30:26Z

Benchmark results

Performance differences relative to b90ded6.

1-conn/1-100mb-resp/mtu-1504 (aka. Download)/client: Change within noise threshold.

       time:   [198.86 ms 199.23 ms 199.67 ms]
       thrpt:  [500.83 MiB/s 501.94 MiB/s 502.88 MiB/s]
change:
       time:   [−0.5566% −0.2927% −0.0386%] (p = 0.03 < 0.05)
       thrpt:  [+0.0387% +0.2936% +0.5597%]
Found 3 outliers among 100 measurements (3.00%)

1 (1.00%) low mild

1 (1.00%) high mild

1 (1.00%) high severe

1-conn/10_000-parallel-1b-resp/mtu-1504 (aka. RPS)/client: Change within noise threshold.

       time:   [280.98 ms 282.72 ms 284.52 ms]
       thrpt:  [35.147 Kelem/s 35.370 Kelem/s 35.590 Kelem/s]
change:
       time:   [−2.5094% −1.6343% −0.6886%] (p = 0.00 < 0.05)
       thrpt:  [+0.6934% +1.6614% +2.5740%]
Found 1 outliers among 100 measurements (1.00%)

1 (1.00%) high mild

1-conn/1-1b-resp/mtu-1504 (aka. HPS)/client: No change in performance detected.

       time:   [28.445 ms 28.498 ms 28.571 ms]
       thrpt:  [35.000   B/s 35.090   B/s 35.155   B/s]
change:
       time:   [−0.3734% −0.0046% +0.3304%] (p = 0.98 > 0.05)
       thrpt:  [−0.3293% +0.0046% +0.3748%]
Found 14 outliers among 100 measurements (14.00%)

3 (3.00%) low severe

1 (1.00%) high mild

10 (10.00%) high severe

1-conn/1-100mb-req/mtu-1504 (aka. Upload)/client: Change within noise threshold.

       time:   [204.78 ms 205.03 ms 205.28 ms]
       thrpt:  [487.14 MiB/s 487.73 MiB/s 488.34 MiB/s]
change:
       time:   [+0.6909% +0.8881% +1.0673%] (p = 0.00 < 0.05)
       thrpt:  [−1.0560% −0.8803% −0.6862%]
Found 2 outliers among 100 measurements (2.00%)

1 (1.00%) low severe

1 (1.00%) high mild

decode 4096 bytes, mask ff: No change in performance detected.

       time:   [11.605 µs 11.656 µs 11.721 µs]
       change: [−0.0692% +0.4745% +1.1337%] (p = 0.14 > 0.05)
Found 16 outliers among 100 measurements (16.00%)

2 (2.00%) low severe

2 (2.00%) low mild

12 (12.00%) high severe

decode 1048576 bytes, mask ff: No change in performance detected.

       time:   [3.0212 ms 3.0331 ms 3.0475 ms]
       change: [−0.4890% +0.0817% +0.6473%] (p = 0.79 > 0.05)
Found 11 outliers among 100 measurements (11.00%)

2 (2.00%) low mild

1 (1.00%) high mild

8 (8.00%) high severe

decode 4096 bytes, mask 7f: No change in performance detected.

       time:   [19.984 µs 20.047 µs 20.112 µs]
       change: [−0.1467% +0.3306% +0.8997%] (p = 0.24 > 0.05)
Found 24 outliers among 100 measurements (24.00%)

2 (2.00%) low severe

3 (3.00%) low mild

1 (1.00%) high mild

18 (18.00%) high severe

decode 1048576 bytes, mask 7f: No change in performance detected.

       time:   [5.0479 ms 5.0595 ms 5.0728 ms]
       change: [−0.2863% +0.0679% +0.4199%] (p = 0.70 > 0.05)
Found 14 outliers among 100 measurements (14.00%)

14 (14.00%) high severe

decode 4096 bytes, mask 3f: No change in performance detected.

       time:   [8.2820 µs 8.3171 µs 8.3584 µs]
       change: [−3.3451% −0.7110% +1.0209%] (p = 0.66 > 0.05)
Found 16 outliers among 100 measurements (16.00%)

4 (4.00%) low mild

1 (1.00%) high mild

11 (11.00%) high severe

decode 1048576 bytes, mask 3f: No change in performance detected.

       time:   [1.5854 ms 1.5906 ms 1.5969 ms]
       change: [−0.4174% +0.0858% +0.5841%] (p = 0.74 > 0.05)
Found 8 outliers among 100 measurements (8.00%)

3 (3.00%) high mild

5 (5.00%) high severe

1-streams/each-1000-bytes/wallclock-time: No change in performance detected.

       time:   [589.03 µs 591.17 µs 593.62 µs]
       change: [−1.1062% −0.2596% +0.4791%] (p = 0.54 > 0.05)
Found 8 outliers among 100 measurements (8.00%)

1 (1.00%) high mild

7 (7.00%) high severe

1-streams/each-1000-bytes/simulated-time

time:   [118.82 ms 119.05 ms 119.29 ms]

thrpt:  [8.1862 KiB/s 8.2027 KiB/s 8.2190 KiB/s]

change:

time:   [−0.1071% +0.1614% +0.4457%] (p = 0.26 > 0.05)

thrpt:  [−0.4438% −0.1611% +0.1072%]

No change in performance detected.

Found 1 outliers among 100 measurements (1.00%)

1 (1.00%) high mild

1000-streams/each-1-bytes/wallclock-time: Change within noise threshold.

       time:   [13.779 ms 13.803 ms 13.827 ms]
       change: [−1.4949% −0.8003% −0.2708%] (p = 0.01 < 0.05)
Found 1 outliers among 100 measurements (1.00%)

1 (1.00%) high mild

1000-streams/each-1-bytes/simulated-time

time:   [14.981 s 14.996 s 15.011 s]

thrpt:  [66.619   B/s 66.684   B/s 66.750   B/s]

change:

time:   [−0.1933% −0.0598% +0.0488%] (p = 0.35 > 0.05)

thrpt:  [−0.0488% +0.0598% +0.1937%]

No change in performance detected.

1000-streams/each-1000-bytes/wallclock-time: Change within noise threshold.

       time:   [48.591 ms 48.754 ms 48.923 ms]
       change: [−1.2857% −0.7961% −0.3212%] (p = 0.00 < 0.05)
Found 3 outliers among 100 measurements (3.00%)

3 (3.00%) high mild

1000-streams/each-1000-bytes/simulated-time

time:   [18.883 s 19.061 s 19.238 s]

thrpt:  [50.763 KiB/s 51.233 KiB/s 51.716 KiB/s]

change:

time:   [−1.4684% −0.2121% +0.9641%] (p = 0.74 > 0.05)

thrpt:  [−0.9549% +0.2125% +1.4903%]

No change in performance detected.

Found 3 outliers among 100 measurements (3.00%)

3 (3.00%) low mild

coalesce_acked_from_zero 1+1 entries: No change in performance detected.

       time:   [88.300 ns 88.657 ns 88.998 ns]
       change: [−0.7076% −0.0900% +0.4401%] (p = 0.76 > 0.05)
Found 6 outliers among 100 measurements (6.00%)

6 (6.00%) high mild

coalesce_acked_from_zero 3+1 entries: No change in performance detected.

       time:   [106.04 ns 106.39 ns 106.76 ns]
       change: [−3.3446% −0.7650% +0.8209%] (p = 0.68 > 0.05)
Found 11 outliers among 100 measurements (11.00%)

11 (11.00%) high severe

coalesce_acked_from_zero 10+1 entries: No change in performance detected.

       time:   [105.29 ns 105.63 ns 106.07 ns]
       change: [−0.3605% +0.2551% +0.8561%] (p = 0.45 > 0.05)
Found 8 outliers among 100 measurements (8.00%)

8 (8.00%) high severe

coalesce_acked_from_zero 1000+1 entries: No change in performance detected.

       time:   [88.850 ns 88.961 ns 89.097 ns]
       change: [−1.1690% −0.2602% +0.6253%] (p = 0.58 > 0.05)
Found 13 outliers among 100 measurements (13.00%)

6 (6.00%) high mild

7 (7.00%) high severe

RxStreamOrderer::inbound_frame(): 💚 Performance has improved.

       time:   [107.05 ms 107.12 ms 107.19 ms]
       change: [−1.3466% −1.2436% −1.1448%] (p = 0.00 < 0.05)
Found 30 outliers among 100 measurements (30.00%)

12 (12.00%) low severe

2 (2.00%) low mild

5 (5.00%) high mild

11 (11.00%) high severe

sent::Packets::take_ranges: No change in performance detected.

       time:   [4.5999 µs 4.6683 µs 4.7373 µs]
       change: [−3.0742% +0.1722% +3.6012%] (p = 0.92 > 0.05)
Found 2 outliers among 100 measurements (2.00%)

2 (2.00%) high mild

transfer/pacing-false/varying-seeds/wallclock-time/run: Change within noise threshold.

       time:   [25.450 ms 25.500 ms 25.558 ms]
       change: [−0.7068% −0.4516% −0.1821%] (p = 0.00 < 0.05)
Found 3 outliers among 100 measurements (3.00%)

2 (2.00%) high mild

1 (1.00%) high severe

transfer/pacing-false/varying-seeds/simulated-time/run: No change in performance detected.

       time:   [25.139 s 25.176 s 25.214 s]
       thrpt:  [162.45 KiB/s 162.69 KiB/s 162.94 KiB/s]
change:
       time:   [−0.3739% −0.1767% +0.0336%] (p = 0.09 > 0.05)
       thrpt:  [−0.0335% +0.1770% +0.3753%]

transfer/pacing-true/varying-seeds/wallclock-time/run: Change within noise threshold.

       time:   [25.936 ms 26.007 ms 26.083 ms]
       change: [+1.5088% +1.8631% +2.2585%] (p = 0.00 < 0.05)
Found 1 outliers among 100 measurements (1.00%)

1 (1.00%) high severe

transfer/pacing-true/varying-seeds/simulated-time/run: No change in performance detected.

       time:   [24.921 s 24.959 s 25.000 s]
       thrpt:  [163.84 KiB/s 164.11 KiB/s 164.36 KiB/s]
change:
       time:   [−0.3303% −0.1258% +0.0979%] (p = 0.27 > 0.05)
       thrpt:  [−0.0978% +0.1259% +0.3314%]
Found 1 outliers among 100 measurements (1.00%)

1 (1.00%) high mild

transfer/pacing-false/same-seed/wallclock-time/run: Change within noise threshold.

       time:   [26.137 ms 26.156 ms 26.175 ms]
       change: [+2.1238% +2.3229% +2.4737%] (p = 0.00 < 0.05)
Found 2 outliers among 100 measurements (2.00%)

2 (2.00%) high mild

transfer/pacing-false/same-seed/simulated-time/run: No change in performance detected.

       time:   [25.710 s 25.710 s 25.710 s]
       thrpt:  [159.31 KiB/s 159.31 KiB/s 159.31 KiB/s]
change:
       time:   [+0.0000% +0.0000% +0.0000%] (p = NaN > 0.05)
       thrpt:  [+0.0000% +0.0000% +0.0000%]

transfer/pacing-true/same-seed/wallclock-time/run: Change within noise threshold.

       time:   [27.163 ms 27.187 ms 27.212 ms]
       change: [+0.8877% +1.1122% +1.2727%] (p = 0.00 < 0.05)
Found 3 outliers among 100 measurements (3.00%)

1 (1.00%) low mild

2 (2.00%) high mild

transfer/pacing-true/same-seed/simulated-time/run: No change in performance detected.

       time:   [25.675 s 25.675 s 25.675 s]
       thrpt:  [159.53 KiB/s 159.53 KiB/s 159.53 KiB/s]
change:
       time:   [+0.0000% +0.0000% +0.0000%] (p = NaN > 0.05)
       thrpt:  [+0.0000% +0.0000% +0.0000%]

Download data for profiler.firefox.com or download performance comparison data.

omansfeld mentioned this pull request Sep 15, 2025

CUBIC update to RFC 9438 tracking issue #2967

Closed

6 tasks

omansfeld force-pushed the cubic_region_updates branch from fa05f07 to a2d3411 Compare October 13, 2025 11:32

Copilot AI review requested due to automatic review settings October 13, 2025 11:32

Copilot AI reviewed Oct 13, 2025

View reviewed changes

neqo-transport/src/cc/cubic.rs Outdated Show resolved Hide resolved

omansfeld force-pushed the cubic_region_updates branch from a2d3411 to 1d42561 Compare October 13, 2025 11:47

omansfeld changed the title ~~cubic: cubic region updates~~ feat(transport/cubic): cubic region RFC9438 updates Oct 13, 2025

omansfeld marked this pull request as ready for review October 13, 2025 12:08

omansfeld requested review from KershawChang and martinthomson as code owners October 13, 2025 12:08

Copilot AI review requested due to automatic review settings October 13, 2025 12:08

omansfeld requested review from larseggert and mxinden as code owners October 13, 2025 12:08

Copilot AI reviewed Oct 13, 2025

View reviewed changes

neqo-transport/src/cc/cubic.rs Show resolved Hide resolved

neqo-transport/src/cc/cubic.rs Show resolved Hide resolved

mxinden approved these changes Oct 13, 2025

View reviewed changes

neqo-transport/src/cc/cubic.rs Outdated Show resolved Hide resolved

neqo-transport/src/cc/cubic.rs Show resolved Hide resolved

omansfeld force-pushed the cubic_region_updates branch from 35d6048 to bcd1bb5 Compare October 13, 2025 15:41

mxinden reviewed Oct 13, 2025

View reviewed changes

neqo-transport/src/cc/cubic.rs Outdated Show resolved Hide resolved

mxinden reviewed Oct 13, 2025

View reviewed changes

neqo-transport/src/cc/cubic.rs Outdated Show resolved Hide resolved

Copilot AI review requested due to automatic review settings October 13, 2025 16:14

Copilot AI reviewed Oct 13, 2025

View reviewed changes

mxinden reviewed Oct 13, 2025

View reviewed changes

neqo-transport/src/cc/cubic.rs Outdated Show resolved Hide resolved

Copilot AI review requested due to automatic review settings October 14, 2025 15:26

Copilot AI reviewed Oct 14, 2025

View reviewed changes

neqo-transport/src/cc/cubic.rs Show resolved Hide resolved

feat(transport/cubic): cubic region RFC9438 updates

8f6f53d

- comments and docs - new `target` formula clamping to min and max value (`cwnd<=target<=cwnd*1.5`) - removing `EXPONENTIAL_GROWTH_REDUCTION` since it's use is now covered by the max value mentioned above

omansfeld added 4 commits October 14, 2025 17:27

adressing review comments

bdcb193

- linux reference links - renaming variables - `debug_assert!` to check exponential growth of `w_est`

fixes

a636eb2

- comments - moved debug assert

fix the assert

1527ebc

- add predicate for small cwnd values - move assert and change the message

remove assert

920de9a

omansfeld force-pushed the cubic_region_updates branch from 28ffece to 920de9a Compare October 14, 2025 15:27

mxinden approved these changes Oct 14, 2025

View reviewed changes

omansfeld enabled auto-merge October 14, 2025 15:47

omansfeld added this pull request to the merge queue Oct 14, 2025

Merged via the queue into mozilla:main with commit 7334a18 Oct 14, 2025
94 checks passed

omansfeld deleted the cubic_region_updates branch October 14, 2025 16:05

Conversation

omansfeld commented Sep 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov bot commented Sep 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Uh oh!

Uh oh!

mxinden left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

mxinden commented Oct 13, 2025

Uh oh!

mxinden commented Oct 13, 2025

Uh oh!

github-actions bot commented Oct 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Bencher Report

Uh oh!

omansfeld commented Oct 13, 2025

Uh oh!

mxinden commented Oct 13, 2025

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions bot commented Oct 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Bencher Report

🚨 1 Alert

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Uh oh!

mxinden left a comment

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Oct 14, 2025

Failed Interop Tests

neqo-latest as client

neqo-latest as server

Succeeded Interop Tests

neqo-latest as client

neqo-latest as server

Unsupported Interop Tests

neqo-latest as client

neqo-latest as server

Uh oh!

github-actions bot commented Oct 14, 2025

Client/server transfer results

Uh oh!

Uh oh!

github-actions bot commented Oct 14, 2025

Benchmark results

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

omansfeld commented Sep 15, 2025 •

edited

Loading

codecov bot commented Sep 15, 2025 •

edited

Loading

github-actions bot commented Oct 13, 2025 •

edited

Loading

github-actions bot commented Oct 13, 2025 •

edited

Loading