Of course a loop is even faster...

#!perl 
use strict;
use Benchmark qw(cmpthese);
my %counter;

sub normal {
    return 0 unless $_[0];
    @_ = ($_[0] - 1);
    $counter{(caller(0))[3]}++;
    return normal(@_);
}

sub tail {
    return 0 unless $_[0];
    @_ = ($_[0] - 1);
    $counter{(caller(0))[3]}++;
    goto &tail;
}

sub faster {
    return 0 unless $_[0];
    @_ = ($_[0] - 1);
    $counter{(caller(0))[3]}++;
    &faster;
}

sub loop {
    for (0..$_[0]) { $counter{(caller(0))[3]}++; }
    return 0;
}

cmpthese( -3, {
        "normal" => sub { normal(50000) },
        "tail"   => sub {   tail(50000) },
        "faster" => sub { faster(50000) },
        "loop"   => sub { loop(50000) },
    }
);

__END__
         Rate normal   tail faster   loop
normal 2.15/s     --    -8%   -18%   -38%
tail   2.35/s     9%     --   -11%   -32%
faster 2.64/s    23%    12%     --   -24%
loop   3.45/s    60%    47%    31%     --
[download]

I put the caller thingee in there to ensure that the for loop didnt get optimized away. Which is in itself also a good argument for just using a loop in the first place.

cmpthese( -10, {
        "normal" => sub { normal(500000) },
        "tail"   => sub {   tail(500000) },
        "faster" => sub { faster(500000) },
    }
);
[download]

       s/iter normal   tail faster
normal   2.26     --   -22%   -40%
tail     1.77    28%     --   -23%
faster   1.36    66%    30%     --
[download]