Do you mean you added the my $z=12; to both the subroutines and the lexical one became 50% slower? I can't reproduce that behaviour. Adding it to only one of the subs slows it (35% in my case), but then we are comparing apples and oranges.

I meant to say that comparing lexicals only: one with a my $z; inside the loop is so much slower. Even 35% is much

here is the code:

use Benchmark 'cmpthese';

cmpthese(-2, {
    predecl => '
        my $y;
        my $x;
        for $x (1..10000) { $y+=$x }
    #print "$y\n";
    ',
    lexical => '
        my $y;
        for my $x (1..10000) { $y+=$x }
    #print "$y\n";
    ',
});

__END__
[download]

          Rate predecl lexical
predecl 1836/s      --    -54%
lexical 3996/s    118%      --
[download]

For loops do aliasing!

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery}