perf: eliminate bounds checks in applyMask

[mccutchen]

Context

A small follow-up to #50, inspired by this blog post by @jub0bs that walks through optimizing away bounds checks (as well as enabling inlining).

Note, there are still two bounds checks per applyMask(), on line 494 and line 504:

$ go build -gcflags '-d=ssa/check_bce/debug=1'
# github.com/mccutchen/websocket
./proto.go:322:28: Found IsInBounds
./proto.go:368:60: Found IsSliceInBounds
./proto.go:439:49: Found IsSliceInBounds
./proto.go:494:19: Found IsSliceInBounds
./proto.go:504:22: Found IsSliceInBounds

But that's an improvement on the 9 bounds checks we were seeing previously:

$ go build -gcflags '-d=ssa/check_bce/debug=1'
# github.com/mccutchen/websocket
./proto.go:322:28: Found IsInBounds
./proto.go:368:60: Found IsSliceInBounds
./proto.go:439:49: Found IsSliceInBounds
./proto.go:493:10: Found IsInBounds
./proto.go:494:10: Found IsInBounds
./proto.go:495:10: Found IsInBounds
./proto.go:496:10: Found IsInBounds
./proto.go:497:10: Found IsInBounds
./proto.go:498:10: Found IsInBounds
./proto.go:499:10: Found IsInBounds
./proto.go:500:10: Found IsInBounds
./proto.go:503:10: Found IsInBounds

Results

A nice little ~4.5% improvement to throughput according to the results posted below:

                  │ ./baseline/bench-results.txt │      ./head/bench-results.txt       │
                  │            sec/op            │   sec/op     vs base                │
ReadFrame/1KiB-4                     982.4n ± 1%   906.8n ± 2%   -7.71% (p=0.000 n=10)
ReadFrame/1MiB-4                     628.9µ ± 0%   561.0µ ± 1%  -10.79% (p=0.000 n=10)
WriteFrame/1KiB-4                    1.341µ ± 0%   1.360µ ± 1%   +1.42% (p=0.000 n=10)
WriteFrame/1MiB-4                    1.039m ± 1%   1.035m ± 1%        ~ (p=0.105 n=10)
geomean                              30.46µ        29.08µ        -4.51%

                  │ ./baseline/bench-results.txt │       ./head/bench-results.txt        │
                  │             B/s              │      B/s       vs base                │
ReadFrame/1KiB-4                   1002.8Mi ± 1%   1086.4Mi ± 2%   +8.35% (p=0.000 n=10)
ReadFrame/1MiB-4                    1.553Gi ± 0%    1.741Gi ± 1%  +12.10% (p=0.000 n=10)
WriteFrame/1KiB-4                   735.1Mi ± 0%    724.5Mi ± 1%   -1.44% (p=0.000 n=10)
WriteFrame/1MiB-4                   962.5Mi ± 1%    966.6Mi ± 1%        ~ (p=0.105 n=10)
geomean                             1.006Gi         1.054Gi        +4.71%

---

Edit: realized after the fact that I didn't actually link to the blog post that showed me how to do this!

[github-actions]

🔥 Run benchmarks comparing c168278 against main:

gh workflow run bench.yaml -f pr_number=53

Note: this comment will update with each new commit.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.81%. Comparing base (3cb163b) to head (c168278).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #53      +/-   ##
==========================================
+ Coverage   92.76%   92.81%   +0.04%     
==========================================
  Files           2        2              
  Lines         484      487       +3     
==========================================
+ Hits          449      452       +3     
  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.

[github-actions]

benchstats: 3cb163b...c168278

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                     982.4n ± 1%   906.8n ± 2%   -7.71% (p=0.000 n=10)
ReadFrame/1MiB-4                     628.9µ ± 0%   561.0µ ± 1%  -10.79% (p=0.000 n=10)
WriteFrame/1KiB-4                    1.341µ ± 0%   1.360µ ± 1%   +1.42% (p=0.000 n=10)
WriteFrame/1MiB-4                    1.039m ± 1%   1.035m ± 1%        ~ (p=0.105 n=10)
geomean                              30.46µ        29.08µ        -4.51%

                  │ ./baseline/bench-results.txt │       ./head/bench-results.txt        │
                  │             B/s              │      B/s       vs base                │
ReadFrame/1KiB-4                   1002.8Mi ± 1%   1086.4Mi ± 2%   +8.35% (p=0.000 n=10)
ReadFrame/1MiB-4                    1.553Gi ± 0%    1.741Gi ± 1%  +12.10% (p=0.000 n=10)
WriteFrame/1KiB-4                   735.1Mi ± 0%    724.5Mi ± 1%   -1.44% (p=0.000 n=10)
WriteFrame/1MiB-4                   962.5Mi ± 1%    966.6Mi ± 1%        ~ (p=0.105 n=10)
geomean                             1.006Gi         1.054Gi        +4.71%

                  │ ./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%       ~ (p=0.121 n=10)
WriteFrame/1KiB-4                   1.125Ki ± 0%   1.125Ki ± 0%       ~ (p=1.000 n=10) ¹
WriteFrame/1MiB-4                   1.008Mi ± 0%   1.008Mi ± 0%       ~ (p=0.481 n=10)
geomean                             34.37Ki        34.37Ki       +0.00%
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%   1.000 ± 0%       ~ (p=1.000 n=10) ¹
geomean                               2.236        2.236       +0.00%
¹ all samples are equal