# check-cfg perf evaluation

Was done using `hyperfine` on an Intel Core i7-7700HQ (4 cores, 3.8GHz max).

They represent both extreme, helloworld (0 feature and 1 cfg) and windows-rs (~700 features and >99000 cfgs).

## `windows`

$ hyperfine --prepare "cargo clean" "cargo +nightly check" "cargo +nightly check -Zcheck-cfg"
Benchmark 1: cargo +nightly check
  Time (mean ± σ):     805.4 ms ±  12.5 ms    [User: 675.2 ms, System: 150.8 ms]
  Range (min … max):   792.1 ms … 824.6 ms    10 runs

Benchmark 2: cargo +nightly check -Zcheck-cfg
  Time (mean ± σ):     822.5 ms ±  11.9 ms    [User: 691.1 ms, System: 150.7 ms]
  Range (min … max):   803.6 ms … 836.6 ms    10 runs

Summary
  cargo +nightly check ran
    1.02 ± 0.02 times faster than cargo +nightly check -Zcheck-cfg

Or around +17.1ms for 4 (local) crates, with the being one being at ~700 features and 99102 total `#[cfg]`s in the code!

(Also note that there was 10/15 warnings and they were not suppressed so the linting/display machinery was printing them)

## helloworld (lib)

$ hyperfine --prepare "cargo clean" "cargo +nightly check" "cargo +nightly check -Zcheck-cfg"
Benchmark 1: cargo +nightly check
  Time (mean ± σ):      67.5 ms ±   1.1 ms    [User: 30.3 ms, System: 34.9 ms]
  Range (min … max):    65.9 ms …  69.2 ms    20 runs

Benchmark 2: cargo +nightly check -Zcheck-cfg
  Time (mean ± σ):      67.4 ms ±   0.9 ms    [User: 31.0 ms, System: 34.0 ms]
  Range (min … max):    65.4 ms …  69.4 ms    21 runs

Summary
  cargo +nightly check -Zcheck-cfg ran
    1.00 ± 0.02 times faster than cargo +nightly check

No visible impact.