feat: Reduce reallocations in RxStreamOrderer#3003
Draft
larseggert wants to merge 3 commits intomozilla:mainfrom
Draft
feat: Reduce reallocations in RxStreamOrderer#3003larseggert wants to merge 3 commits intomozilla:mainfrom
larseggert wants to merge 3 commits intomozilla:mainfrom
Conversation
Let's see if this helps performance.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #3003 +/- ##
==========================================
- Coverage 95.66% 93.36% -2.31%
==========================================
Files 123 123
Lines 35702 35712 +10
Branches 35702 35712 +10
==========================================
- Hits 34156 33342 -814
- Misses 1506 1528 +22
- Partials 40 842 +802
|
Contributor
|
Bencher Report| Branch | feat-inbound_frame-prealloc |
| Testbed | On-prem |
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 | 278.12 ms(-0.08%)Baseline: 278.34 ms | 282.73 ms (98.37%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| msquic vs. neqo (cubic, paced) | 📈 view plot 🚷 view threshold | 224.72 ms(+12.76%)Baseline: 199.30 ms | 236.94 ms (94.84%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| neqo vs. google (cubic, paced) | 📈 view plot 🚷 view threshold | 756.26 ms(-0.45%)Baseline: 759.69 ms | 774.82 ms (97.61%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| neqo vs. msquic (cubic, paced) | 📈 view plot 🚷 view threshold | 156.46 ms(-0.83%)Baseline: 157.78 ms | 160.59 ms (97.43%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| neqo vs. neqo (cubic) | 📈 view plot 🚷 view threshold | 94.69 ms(+3.42%)Baseline: 91.56 ms | 96.88 ms (97.74%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| neqo vs. neqo (cubic, paced) | 📈 view plot 🚷 view threshold | 94.16 ms(+1.35%)Baseline: 92.90 ms | 98.09 ms (95.99%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| neqo vs. neqo (reno) | 📈 view plot 🚷 view threshold | 93.24 ms(+1.86%)Baseline: 91.54 ms | 96.70 ms (96.43%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| neqo vs. neqo (reno, paced) | 📈 view plot 🚷 view threshold | 95.04 ms(+2.42%)Baseline: 92.79 ms | 97.78 ms (97.19%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| neqo vs. quiche (cubic, paced) | 📈 view plot 🚷 view threshold | 191.75 ms(-0.97%)Baseline: 193.64 ms | 196.97 ms (97.35%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| neqo vs. s2n (cubic, paced) | 📈 view plot 🚷 view threshold | 221.72 ms(+0.26%)Baseline: 221.14 ms | 224.10 ms (98.94%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| quiche vs. neqo (cubic, paced) | 📈 view plot 🚷 view threshold | 157.33 ms(+2.74%)Baseline: 153.14 ms | 158.50 ms (99.26%) |
| Benchmark | Latency | Benchmark Result milliseconds (ms) (Result Δ%) | Upper Boundary milliseconds (ms) (Limit %) |
|---|---|---|---|
| s2n vs. neqo (cubic, paced) | 📈 view plot 🚷 view threshold | 173.28 ms(-0.28%)Baseline: 173.77 ms | 178.00 ms (97.35%) |
Contributor
Benchmark resultsPerformance differences relative to b3d8f0d. 1-conn/1-100mb-resp/mtu-1504 (aka. Download)/client: No change in performance detected. time: [200.11 ms 200.47 ms 200.96 ms]
thrpt: [497.61 MiB/s 498.82 MiB/s 499.72 MiB/s]
change:
time: [−0.0850% +0.1485% +0.4403%] (p = 0.28 > 0.05)
thrpt: [−0.4384% −0.1483% +0.0851%]
1-conn/10_000-parallel-1b-resp/mtu-1504 (aka. RPS)/client: No change in performance detected. time: [299.66 ms 301.36 ms 303.08 ms]
thrpt: [32.994 Kelem/s 33.183 Kelem/s 33.371 Kelem/s]
change:
time: [−0.3116% +0.4839% +1.2104%] (p = 0.21 > 0.05)
thrpt: [−1.1959% −0.4816% +0.3126%]
1-conn/1-1b-resp/mtu-1504 (aka. HPS)/client: No change in performance detected. time: [28.416 ms 28.512 ms 28.630 ms]
thrpt: [34.928 B/s 35.073 B/s 35.191 B/s]
change:
time: [−0.3750% +0.0906% +0.5737%] (p = 0.71 > 0.05)
thrpt: [−0.5704% −0.0905% +0.3764%]
1-conn/1-100mb-req/mtu-1504 (aka. Upload)/client: 💚 Performance has improved. time: [202.29 ms 202.62 ms 203.01 ms]
thrpt: [492.58 MiB/s 493.52 MiB/s 494.33 MiB/s]
change:
time: [−3.7342% −3.4882% −3.2606%] (p = 0.00 < 0.05)
thrpt: [+3.3705% +3.6142% +3.8791%]
decode 4096 bytes, mask ff: No change in performance detected. time: [11.613 µs 11.651 µs 11.694 µs]
change: [−0.8027% −0.1804% +0.3326%] (p = 0.57 > 0.05)
decode 1048576 bytes, mask ff: No change in performance detected. time: [3.0185 ms 3.0278 ms 3.0387 ms]
change: [−0.8118% −0.1895% +0.3544%] (p = 0.54 > 0.05)
decode 4096 bytes, mask 7f: No change in performance detected. time: [19.948 µs 19.998 µs 20.056 µs]
change: [−0.3160% +0.1413% +0.5881%] (p = 0.57 > 0.05)
decode 1048576 bytes, mask 7f: No change in performance detected. time: [5.0328 ms 5.0426 ms 5.0540 ms]
change: [−1.2024% −0.5035% +0.0438%] (p = 0.12 > 0.05)
decode 4096 bytes, mask 3f: No change in performance detected. time: [8.2789 µs 8.3159 µs 8.3580 µs]
change: [+0.0483% +0.5635% +1.1728%] (p = 0.06 > 0.05)
decode 1048576 bytes, mask 3f: No change in performance detected. time: [1.5881 ms 1.5949 ms 1.6035 ms]
change: [−2.0388% −0.4004% +0.7757%] (p = 0.67 > 0.05)
1-streams/each-1000-bytes/wallclock-time: Change within noise threshold. time: [589.89 µs 591.70 µs 593.79 µs]
change: [−1.1506% −0.6516% −0.1492%] (p = 0.01 < 0.05)
1000-streams/each-1-bytes/wallclock-time: Change within noise threshold. time: [14.032 ms 14.061 ms 14.092 ms]
change: [−0.9199% −0.6459% −0.3527%] (p = 0.00 < 0.05)
1000-streams/each-1000-bytes/wallclock-time: No change in performance detected. time: [50.776 ms 50.939 ms 51.101 ms]
change: [−0.1426% +0.5306% +1.1027%] (p = 0.10 > 0.05)
1000-streams/each-1000-bytes/simulated-time: No change in performance detected. time: [18.741 s 18.912 s 19.085 s]
thrpt: [51.170 KiB/s 51.638 KiB/s 52.110 KiB/s]
change:
time: [−0.9411% +0.3276% +1.5536%] (p = 0.62 > 0.05)
thrpt: [−1.5298% −0.3265% +0.9501%]
coalesce_acked_from_zero 1+1 entries: No change in performance detected. time: [88.191 ns 88.538 ns 88.895 ns]
change: [−0.1959% +0.3684% +1.1270%] (p = 0.28 > 0.05)
coalesce_acked_from_zero 3+1 entries: No change in performance detected. time: [106.03 ns 106.55 ns 107.30 ns]
change: [−0.2849% +0.3210% +1.0704%] (p = 0.39 > 0.05)
coalesce_acked_from_zero 10+1 entries: No change in performance detected. time: [105.58 ns 106.07 ns 106.64 ns]
change: [−0.4911% −0.0224% +0.4220%] (p = 0.92 > 0.05)
coalesce_acked_from_zero 1000+1 entries: No change in performance detected. time: [88.723 ns 91.556 ns 98.100 ns]
change: [−1.4101% +2.7728% +10.065%] (p = 0.60 > 0.05)
RxStreamOrderer::inbound_frame(): 💚 Performance has improved. time: [102.63 ms 102.79 ms 103.07 ms]
change: [−7.6514% −7.3735% −7.0581%] (p = 0.00 < 0.05)
sent::Packets::take_ranges: No change in performance detected. time: [4.5298 µs 4.6645 µs 4.8068 µs]
change: [−2.0069% +1.3391% +5.2122%] (p = 0.52 > 0.05)
transfer/pacing-false/varying-seeds/wallclock-time/run: Change within noise threshold. time: [26.951 ms 27.000 ms 27.052 ms]
change: [+0.9397% +1.2158% +1.4901%] (p = 0.00 < 0.05)
transfer/pacing-false/varying-seeds/simulated-time/run: No change in performance detected. time: [25.129 s 25.166 s 25.204 s]
thrpt: [162.52 KiB/s 162.76 KiB/s 163.00 KiB/s]
change:
time: [−0.2319% −0.0321% +0.1692%] (p = 0.76 > 0.05)
thrpt: [−0.1689% +0.0321% +0.2325%]
transfer/pacing-true/varying-seeds/wallclock-time/run: Change within noise threshold. time: [27.309 ms 27.379 ms 27.452 ms]
change: [+0.4884% +0.8607% +1.2416%] (p = 0.00 < 0.05)
transfer/pacing-true/varying-seeds/simulated-time/run: Change within noise threshold. time: [24.987 s 25.033 s 25.079 s]
thrpt: [163.32 KiB/s 163.62 KiB/s 163.92 KiB/s]
change:
time: [+0.0800% +0.3075% +0.5500%] (p = 0.01 < 0.05)
thrpt: [−0.5470% −0.3065% −0.0799%]
transfer/pacing-false/same-seed/wallclock-time/run: Change within noise threshold. time: [26.419 ms 26.434 ms 26.449 ms]
change: [+0.9941% +1.1013% +1.2043%] (p = 0.00 < 0.05)
transfer/pacing-false/same-seed/simulated-time/run: No change in performance detected. time: [25.152 s 25.152 s 25.152 s]
thrpt: [162.85 KiB/s 162.85 KiB/s 162.85 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: [28.142 ms 28.160 ms 28.179 ms]
change: [+0.0679% +0.1823% +0.2954%] (p = 0.00 < 0.05)
transfer/pacing-true/same-seed/simulated-time/run: No change in performance detected. time: [25.588 s 25.588 s 25.588 s]
thrpt: [160.07 KiB/s 160.07 KiB/s 160.07 KiB/s]
change:
time: [+0.0000% +0.0000% +0.0000%] (p = NaN > 0.05)
thrpt: [+0.0000% +0.0000% +0.0000%]
Download data for |
Collaborator
Author
|
Hm. Transfer test regression, but one bench shows an improvement. Time to look at flamegraphs... |
Collaborator
Author
|
Hm. Simplifying |
Collaborator
Author
|
So what we save in |
CodSpeed Performance ReportMerging #3003 will improve performances by 17.7%Comparing Summary
Benchmarks breakdown
|
Contributor
Failed Interop TestsQUIC Interop Runner, client vs. server, differences relative to b9c32c7. neqo-latest as client
neqo-latest as server
All resultsSucceeded Interop TestsQUIC Interop Runner, client vs. server neqo-latest as client
neqo-latest as server
Unsupported Interop TestsQUIC Interop Runner, client vs. server neqo-latest as client
neqo-latest as server
|
Contributor
Client/server transfer resultsPerformance differences relative to b9c32c7. Transfer of 33554432 bytes over loopback, min. 100 runs. All unit-less numbers are in milliseconds.
Download data for |
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
1 participant
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.
Let's see if this helps performance.