Try target-cpu=native

[Skgland]

Setting target-cpu to native might enable the compiler to apply more optimization.

I had some significant improvements, but also some regressions on my machine.

My local bench results from adding this

Created using the compare target in the justfile.

compare.log

running 3 tests
test tests::test_all_s ... ignored
test tests::test_large ... ignored
test tests::test_simple ... ignored

test result: ok. 0 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 0.00s

run_switches/baseline_unicode
                        time:   [3.5822 ms 3.5846 ms 3.5870 ms]
                        thrpt:  [265.87 MiB/s 266.05 MiB/s 266.23 MiB/s]
                 change:
                        time:   [-0.8116% -0.7452% -0.6735%] (p = 0.00 < 0.05)
                        thrpt:  [+0.6781% +0.7508% +0.8182%]
                        Change within noise threshold.
run_switches/baseline   time:   [3.1508 ms 3.1551 ms 3.1597 ms]
                        thrpt:  [301.82 MiB/s 302.26 MiB/s 302.68 MiB/s]
                 change:
                        time:   [-0.3682% -0.1833% +0.0162%] (p = 0.06 > 0.05)
                        thrpt:  [-0.0162% +0.1837% +0.3695%]
                        No change in performance detected.
Found 3 outliers among 100 measurements (3.00%)
  2 (2.00%) high mild
  1 (1.00%) high severe
run_switches/opt1_idiomatic
                        time:   [152.82 µs 152.86 µs 152.91 µs]
                        thrpt:  [6.0908 GiB/s 6.0927 GiB/s 6.0944 GiB/s]
                 change:
                        time:   [-76.164% -76.123% -76.061%] (p = 0.00 < 0.05)
                        thrpt:  [+317.72% +318.81% +319.54%]
                        Performance has improved.
Found 11 outliers among 100 measurements (11.00%)
  5 (5.00%) low severe
  3 (3.00%) low mild
  1 (1.00%) high mild
  2 (2.00%) high severe
run_switches/opt2_count_s
                        time:   [93.749 µs 93.887 µs 94.042 µs]
                        thrpt:  [9.9032 GiB/s 9.9196 GiB/s 9.9342 GiB/s]
                 change:
                        time:   [-66.435% -66.356% -66.271%] (p = 0.00 < 0.05)
                        thrpt:  [+196.48% +197.23% +197.93%]
                        Performance has improved.
Found 24 outliers among 100 measurements (24.00%)
  2 (2.00%) low severe
  13 (13.00%) low mild
  1 (1.00%) high mild
  8 (8.00%) high severe
run_switches/opt3_count_s_branchless
                        time:   [74.144 µs 74.154 µs 74.165 µs]
                        thrpt:  [12.557 GiB/s 12.559 GiB/s 12.561 GiB/s]
                 change:
                        time:   [-69.834% -69.725% -69.658%] (p = 0.00 < 0.05)
                        thrpt:  [+229.58% +230.30% +231.49%]
                        Performance has improved.
Found 16 outliers among 100 measurements (16.00%)
  6 (6.00%) high mild
  10 (10.00%) high severe
run_switches/opt6_chunk_count
                        time:   [19.981 µs 19.993 µs 20.005 µs]
                        thrpt:  [46.554 GiB/s 46.583 GiB/s 46.610 GiB/s]
                 change:
                        time:   [+6.1832% +6.6117% +7.2029%] (p = 0.00 < 0.05)
                        thrpt:  [-6.7189% -6.2017% -5.8231%]
                        Performance has regressed.
Found 6 outliers among 100 measurements (6.00%)
  2 (2.00%) low mild
  2 (2.00%) high mild
  2 (2.00%) high severe
run_switches/opt6_chunk_exact_count
                        time:   [14.590 µs 14.596 µs 14.603 µs]
                        thrpt:  [63.777 GiB/s 63.807 GiB/s 63.832 GiB/s]
                 change:
                        time:   [+1.1436% +1.6182% +1.9169%] (p = 0.00 < 0.05)
                        thrpt:  [-1.8808% -1.5924% -1.1307%]
                        Performance has regressed.
Found 16 outliers among 100 measurements (16.00%)
  1 (1.00%) low severe
  2 (2.00%) low mild
  6 (6.00%) high mild
  7 (7.00%) high severe