perf: optimize writes of small payloads using internal []byte buffer pools

[mccutchen]

Optimize writes of small (<= 8KiB) payloads by using a set of internal []byte buffer pools in WriteFrame to minimize allocations.

[github-actions]

🔥 Run benchmarks comparing cc642fc against main:

gh workflow run bench.yaml -f pr_number=59

Note: this comment will update with each new commit.

[github-actions]

benchstats: 7aee74c...cc642fc

View full benchmark output on the workflow summary.

goos: linux
goarch: amd64
pkg: github.com/mccutchen/websocket
cpu: AMD EPYC 7763 64-Core Processor                
                  │ ./baseline/bench-results.txt │      ./head/bench-results.txt       │
                  │            sec/op            │   sec/op     vs base                │
ReadFrame/1KiB-4                     577.6n ± 1%   592.7n ± 1%   +2.61% (p=0.000 n=10)
ReadFrame/1MiB-4                     281.6µ ± 3%   290.9µ ± 3%   +3.30% (p=0.015 n=10)
WriteFrame/1KiB-4                    671.6n ± 0%   535.0n ± 1%  -20.34% (p=0.000 n=10)
WriteFrame/1MiB-4                    358.0µ ± 6%   330.6µ ± 4%   -7.65% (p=0.003 n=10)
geomean                              14.06µ        13.21µ        -6.03%

                  │ ./baseline/bench-results.txt │       ./head/bench-results.txt       │
                  │             B/s              │     B/s       vs base                │
ReadFrame/1KiB-4                    1.665Gi ± 1%   1.623Gi ± 1%   -2.53% (p=0.000 n=10)
ReadFrame/1MiB-4                    3.468Gi ± 3%   3.357Gi ± 3%   -3.19% (p=0.015 n=10)
WriteFrame/1KiB-4                   1.433Gi ± 0%   1.798Gi ± 1%  +25.53% (p=0.000 n=10)
WriteFrame/1MiB-4                   2.728Gi ± 6%   2.954Gi ± 4%   +8.28% (p=0.003 n=10)
geomean                             2.180Gi        2.320Gi        +6.42%

                  │ ./baseline/bench-results.txt │        ./head/bench-results.txt        │
                  │             B/op             │     B/op      vs base                  │
ReadFrame/1KiB-4                    1.164Ki ± 0%   1.164Ki ± 0%        ~ (p=1.000 n=10) ¹
ReadFrame/1MiB-4                    1.008Mi ± 0%   1.008Mi ± 0%    0.00% (p=0.036 n=10)
WriteFrame/1KiB-4                   1152.00 ± 0%     24.00 ± 0%  -97.92% (p=0.000 n=10)
WriteFrame/1MiB-4                   1.008Mi ± 0%   1.008Mi ± 0%   +0.00% (p=0.000 n=10)
geomean                             34.37Ki        13.06Ki       -62.01%
¹ all samples are equal

                  │ ./baseline/bench-results.txt │       ./head/bench-results.txt        │
                  │          allocs/op           │ allocs/op   vs base                   │
ReadFrame/1KiB-4                      5.000 ± 0%   5.000 ± 0%         ~ (p=1.000 n=10) ¹
ReadFrame/1MiB-4                      5.000 ± 0%   5.000 ± 0%         ~ (p=1.000 n=10) ¹
WriteFrame/1KiB-4                     1.000 ± 0%   1.000 ± 0%         ~ (p=1.000 n=10) ¹
WriteFrame/1MiB-4                     1.000 ± 0%   2.000 ± 0%  +100.00% (p=0.000 n=10)
geomean                               2.236        2.659        +18.92%
¹ all samples are equal

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.21%. Comparing base (7aee74c) to head (cc642fc).

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #59      +/-   ##
==========================================
+ Coverage   92.94%   93.21%   +0.27%     
==========================================
  Files           2        2              
  Lines         496      516      +20     
==========================================
+ Hits          461      481      +20     
  Misses         28       28              
  Partials        7        7              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mccutchen]

While this does yield the expected performance improvements, I'm going to close it for now as it feels like the additional complexity is not worth it. Would rather keep the code a bit cleaner/simpler than chase maximum performance at the moment.

Might revisit later!