Hard-code the hash function for the quick deflate algorithm

[brian-pane]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Flag	Coverage Δ
test-aarch64-apple-darwin	87.45% <100.00%> (-0.04%)	⬇️
test-x86_64-apple-darwin	87.34% <100.00%> (-0.05%)	⬇️
test-x86_64-unknown-linux-gnu	87.08% <100.00%> (-0.03%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
zlib-rs/src/deflate/algorithm/quick.rs	97.67% <100.00%> (ø)

... and 3 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[brian-pane]

This change seems to improve performance at compression level 1 (the only level where the quick alogrithm is used). Oddly, though, I tried the same technique for the fast and medium algorithms, but it produced a regression there, so I only changed quick.rs in this PR.

Benchmark 1 (67 runs): ./blogpost-compress-baseline 1 rs silesia-small.tar
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          75.4ms ±  841us    74.6ms … 79.2ms          3 ( 4%)        0%
  peak_rss           26.6MB ± 60.2KB    26.4MB … 26.7MB          3 ( 4%)        0%
  cpu_cycles          291M  ± 1.78M      289M  …  304M           3 ( 4%)        0%
  instructions        555M  ±  275       555M  …  555M           0 ( 0%)        0%
  cache_references    264K  ± 3.84K      261K  …  285K           6 ( 9%)        0%
  cache_misses        227K  ± 7.31K      193K  …  243K           9 (13%)        0%
  branch_misses      3.06M  ± 5.98K     3.04M  … 3.07M           0 ( 0%)        0%
Benchmark 2 (69 runs): ./target/release/examples/blogpost-compress 1 rs silesia-small.tar
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          73.4ms ±  425us    72.6ms … 74.7ms          3 ( 4%)        ⚡-  2.7% ±  0.3%
  peak_rss           26.6MB ± 49.0KB    26.5MB … 26.7MB          0 ( 0%)          +  0.0% ±  0.1%
  cpu_cycles          283M  ±  582K      282M  …  284M           1 ( 1%)        ⚡-  2.9% ±  0.2%
  instructions        544M  ±  368       544M  …  544M           2 ( 3%)        ⚡-  2.1% ±  0.0%
  cache_references    264K  ± 5.52K      260K  …  292K           6 ( 9%)          -  0.1% ±  0.6%
  cache_misses        228K  ± 7.98K      193K  …  237K           4 ( 6%)          +  0.4% ±  1.1%
  branch_misses      2.91M  ± 6.91K     2.90M  … 2.93M           2 ( 3%)        ⚡-  4.7% ±  0.1%

[folkertdev]

I see improvements for cycles/instructions locally, though runtime is not significant. Still I think that confirms that the effect is real. Thanks!

Benchmark 2 (65 runs): target/release/examples/blogpost-compress 1 rs silesia-small.tar
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          78.1ms ± 1.19ms    76.0ms … 81.6ms          0 ( 0%)          -  0.5% ±  0.5%
  peak_rss           26.6MB ± 73.0KB    26.5MB … 26.7MB          0 ( 0%)          -  0.0% ±  0.1%
  cpu_cycles          283M  ± 2.41M      278M  …  291M           6 ( 9%)        ⚡-  1.4% ±  0.3%
  instructions        566M  ±  282       566M  …  566M           0 ( 0%)        ⚡-  2.7% ±  0.0%