The only speed up that I can think of right now is

If I know I'm going to be sorting a large dataset later on, and I want that sorted loop to be a little tighter I will prepackage the elements I want to actually output.. something along the lines of

# a sample file, lets say a *really* long /etc/passwd file
# with lines like
root:x:0:0:Super User:/:/sbin/sh
someone:x:45:14:Some One:/home/someone:/bin/csh
sometwo:x:46:14:Some Two:/home/sometwo:/bin/ksh

# Lets say im gonna sort on uid then gid, I'll do
while (<IN>) {
   chomp;
   ($uid,$gid) = ( split(/:/) )[2,3];
   push(@tosort, [$uid, $gid, $_]);
}
...

for ( sort { $a->[0] <=> $b->[0] || $a->[1] <=> $b->[1] } @tosort ) {
   print "$_->[2]\n";
}
[download]

Now I haven't actually benchmarked it, so I'm not sure about a speed increase, but in the actual flow of the program, provided the comments are there to describe whats going on, it makes it easier for me to follow later on, and the for (sort) {} block is just that little bit tighter without needing a split(/:/) in there...

p.s after proof reading my post for typos I notice another idiom I use, again without any thought as to performance. I only return the values from a split I want/need. I also don't know if that is faster (its a little late to be benchmarking, but Ill probably get around to it tommorrow, uh.. today..) but its something I do consistantly. That way a few lines later I'm not pulling arbitrary values out of some array, that may or may not be labeled appropriately for that chunk of code..

/* And the Creator, against his better judgement, wrote man.c */