check for and enable the bmi1 and bmi2 target feature

[folkertdev]

closes #361

Locally this is very effective, CI shows a lot more noise. But it does make sense that this would be faster right?

Benchmark 2 (73 runs): target/release/examples/blogpost-uncompress rs-chunked 4 silesia-small.tar.gz
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          68.9ms ±  522us    68.3ms … 70.7ms          2 ( 3%)          -  1.2% ±  0.3%
  peak_rss           24.1MB ± 75.4KB    23.9MB … 24.1MB          0 ( 0%)          -  0.1% ±  0.1%
  cpu_cycles          270M  ± 1.59M      268M  …  278M           4 ( 5%)        ⚡-  1.5% ±  0.2%
  instructions        781M  ±  400       781M  …  781M           2 ( 3%)          -  0.1% ±  0.0%
  cache_references   3.14M  ± 97.1K     2.85M  … 3.36M           0 ( 0%)          -  2.1% ±  1.1%
  cache_misses       83.8K  ± 14.4K     51.4K  …  129K           1 ( 1%)        💩+  7.0% ±  5.4%
  branch_misses      4.09M  ± 11.0K     4.08M  … 4.12M           0 ( 0%)          +  0.2% ±  0.1%
Benchmark 2 (152 runs): target/release/examples/blogpost-uncompress rs-chunked 12 silesia-small.tar.gz
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          32.9ms ±  842us    32.3ms … 41.8ms         10 ( 7%)        ⚡-  2.8% ±  0.5%
  peak_rss           24.1MB ± 58.4KB    24.0MB … 24.1MB          0 ( 0%)          +  0.0% ±  0.1%
  cpu_cycles          111M  ± 2.73M      110M  …  141M          13 ( 9%)        ⚡-  3.7% ±  0.4%
  instructions        265M  ±  276       265M  …  265M           0 ( 0%)        ⚡-  6.4% ±  0.0%
  cache_references   3.38M  ±  153K     3.13M  … 4.62M           5 ( 3%)        ⚡-  2.3% ±  0.9%
  cache_misses       40.3K  ± 12.2K     30.6K  …  167K          10 ( 7%)        💩+  8.7% ±  6.0%
  branch_misses      1.79M  ± 1.95K     1.79M  … 1.81M           8 ( 5%)          -  0.1% ±  0.0%
Benchmark 2 (62 runs): target/release/examples/blogpost-compress 1 rs silesia-small.tar
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          81.8ms ±  950us    80.1ms … 86.5ms          3 ( 5%)          -  0.1% ±  0.4%
  peak_rss           26.7MB ± 68.6KB    26.5MB … 26.7MB          0 ( 0%)          -  0.0% ±  0.1%
  cpu_cycles          288M  ± 2.76M      284M  …  303M           1 ( 2%)          +  0.3% ±  0.3%
  instructions        582M  ±  296       582M  …  582M           0 ( 0%)          -  1.0% ±  0.0%
  cache_references   19.9M  ±  211K     19.4M  … 20.5M           5 ( 8%)          +  0.0% ±  0.3%
  cache_misses        407K  ± 83.5K      293K  …  644K           5 ( 8%)        ⚡-  9.6% ±  5.7%
  branch_misses      2.97M  ± 2.46K     2.97M  … 2.98M           1 ( 2%)          +  0.2% ±  0.0%
Benchmark 2 (12 runs): target/release/examples/blogpost-compress 9 rs silesia-small.tar
  measurement          mean ± σ            min … max           outliers         delta
  wall_time           429ms ± 2.82ms     426ms …  434ms          0 ( 0%)        ⚡-  2.5% ±  0.8%
  peak_rss           24.4MB ±  102KB    24.2MB … 24.5MB          0 ( 0%)          +  0.0% ±  0.3%
  cpu_cycles         1.84G  ± 9.16M     1.83G  … 1.86G           2 (17%)        ⚡-  2.5% ±  0.8%
  instructions       3.43G  ±  270      3.43G  … 3.43G           0 ( 0%)          -  0.6% ±  0.0%
  cache_references    197M  ±  759K      196M  …  198M           0 ( 0%)          -  0.9% ±  0.8%
  cache_misses       2.36M  ±  438K     1.69M  … 3.00M           0 ( 0%)          - 34.7% ± 40.4%
  branch_misses      15.8M  ± 41.1K     15.7M  … 15.8M           0 ( 0%)          +  0.1% ±  0.2%

[codecov]

Codecov Report

Attention: Patch coverage is 52.50000% with 19 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
zlib-rs/src/cpu_features.rs	34.48%	19 Missing ⚠️

Flag	Coverage Δ
fuzz-compress	41.41% <76.47%> (-0.45%)	⬇️
fuzz-decompress	31.09% <70.58%> (+0.08%)	⬆️
test-aarch64-apple-darwin	88.91% <22.22%> (-0.24%)	⬇️
test-x86_64-apple-darwin	85.63% <80.00%> (-0.04%)	⬇️
test-x86_64-unknown-linux-gnu	87.07% <95.00%> (-0.01%)	⬇️

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

Files with missing lines	Coverage Δ
zlib-rs/src/adler32.rs	98.82% <100.00%> (ø)
zlib-rs/src/adler32/avx2.rs	92.50% <100.00%> (ø)
zlib-rs/src/deflate/compare256.rs	98.02% <100.00%> (ø)
zlib-rs/src/deflate/slide_hash.rs	98.96% <100.00%> (ø)
zlib-rs/src/inflate.rs	95.97% <100.00%> (-0.17%)	⬇️
zlib-rs/src/inflate/writer.rs	91.54% <100.00%> (ø)
zlib-rs/src/cpu_features.rs	46.25% <34.48%> (-13.40%)	⬇️

... 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.

[folkertdev]

pclmulqdq implies sse2 https://github.com/rust-lang/rust/blob/master/compiler/rustc_target/src/target_features.rs#L451

[bjorn3]

How much does disabling bmi2 hurt? It contains the pdep and pext instructions which are slow on Zen 2 and lower.

[folkertdev]

Pretty significant, at least on my CPU

Benchmark 2 (74 runs): target/release/examples/blogpost-uncompress rs-chunked 4 silesia-small.tar.gz
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          68.0ms ±  376us    67.4ms … 69.8ms          1 ( 1%)        💩+  1.7% ±  0.2%
  peak_rss           24.1MB ± 70.5KB    23.9MB … 24.1MB          0 ( 0%)          +  0.0% ±  0.1%
  cpu_cycles          275M  ± 1.37M      274M  …  285M           6 ( 8%)        💩+  1.9% ±  0.2%
  instructions        786M  ±  383       786M  …  786M           2 ( 3%)          +  0.7% ±  0.0%
  cache_references   3.19M  ±  387K     2.97M  … 5.72M           3 ( 4%)          +  0.2% ±  3.4%
  cache_misses       64.8K  ± 11.4K     43.4K  …  123K           1 ( 1%)        ⚡- 52.3% ±  2.7%
  branch_misses      4.08M  ± 3.61K     4.08M  … 4.10M           2 ( 3%)          -  0.2% ±  0.1%
Benchmark 2 (153 runs): target/release/examples/blogpost-uncompress rs-chunked 12 silesia-small.tar.gz
  measurement          mean ± σ            min … max           outliers         delta
  wall_time          32.8ms ±  803us    32.1ms … 41.1ms          9 ( 6%)        💩+  2.4% ±  0.4%
  peak_rss           24.1MB ± 64.3KB    23.9MB … 24.1MB          0 ( 0%)          +  0.0% ±  0.1%
  cpu_cycles          114M  ± 2.26M      113M  …  139M          17 (11%)        💩+  2.9% ±  0.3%
  instructions        285M  ±  257       285M  …  285M           0 ( 0%)        💩+  7.8% ±  0.0%
  cache_references   3.61M  ± 99.4K     3.35M  … 4.06M           4 ( 3%)        💩+  8.4% ±  0.7%
  cache_misses       36.6K  ± 6.04K     31.4K  … 86.8K           8 ( 5%)          -  1.1% ±  2.9%
  branch_misses      1.79M  ± 1.28K     1.79M  … 1.80M           5 ( 3%)          -  0.0% ±  0.0%

We already don't go you of our way to support older hardware, instead preferring the best performance on modern hardware, so I'm in favor of enabling it.