Re^2: Choosing the right sort (Or: WUDsamatter?)

[salva]

I have added an entry for Sort::Key to Pustular Postulant benchmarks

use Sort::Key;
sub mykeysort {
    ikeysort { /foo(\d+)\.tla/; $1 } @_
}

and those are the results for 200 elements:

              Rate       naive schwartzian     guttros   Sort::Key
naive       17.9/s          --        -73%        -79%        -83%
schwartzian 67.1/s        276%          --        -20%        -36%
guttros     84.0/s        371%         25%          --        -19%
Sort::Key    104/s        483%         55%         24%          --

for 5000 elements:

               Rate       naive schwartzian     guttros   Sort::Key
naive       0.413/s          --        -77%        -86%        -89%
schwartzian  1.79/s        334%          --        -38%        -52%
guttros      2.88/s        597%         61%          --        -23%
Sort::Key    3.74/s        805%        109%         30%          --

for 50000 elements

            s/iter       naive schwartzian     guttros   Sort::Key
naive         30.2          --        -78%        -87%        -90%
schwartzian   6.53        363%          --        -42%        -56%
guttros       3.81        693%         72%          --        -24%
Sort::Key     2.90        942%        125%         31%          --

and for 500000 elements

          s/iter   guttros Sort::Key
guttros     39.6        --      -25%
Sort::Key   29.7       33%        --

so Sort::Key is consistently ~30% faster than guttros ;-)