As vr pointed out this only performs well on already sorted lists but falls behind pretty fast otherwise. This actually means that you can make it run faster on unsorted list by sorting twice (still doesn't beat most of the leading propositions though):
use warnings;
use strict;
use Benchmark 'cmpthese';
use constant DO_CHECK => 0;
use if DO_CHECK, 'Data::Compare', qw/Compare/;
my @input = (-57..50,52,0);
my @output = (0,0..50,52,-57..-1);
use List::Util 'shuffle';
srand 123;
@input = shuffle @input;
cmpthese(DO_CHECK ? 1 : -2, {
Eily => sub {
my @list = @input;
@list = sort { ~$b <=> ~$a } @list;
Compare(\@list,\@output) or die "@list" if DO_CHECK;
},
Eily2 => sub {
my @list = @input;
@list = sort { ~$b <=> ~$a } sort { $a <=> $b } @list;
Compare(\@list,\@output) or die "@list" if DO_CHECK;
},
});
__END__
Rate Eily Eily2
Eily 20139/s -- -58%
Eily2 47615/s 136% --
Going twice as fast by doing the job twice, talk about counter intuitive :D. |