I wasn't sure to post this little snippet, but I've confirmation that I made well. I had come up with something similar to your solution, but I hadn't realised that the "g" modifier would work so excellently in this case. Your solution is more "general", in that it works for non-hex chars.

For my particular case, anyway, I'll stick to my solution. I made a bit of benchmarking, and my solution is faster for this particular hex application:

#!/usr/bin/perl -w
use strict;
use Benchmark qw(cmpthese);

sub PackUnpack {
    unpack "h*", pack "H*", shift;
    $v; # Thanks Roy!
}
sub SubstRegex {
    (my $v = shift) =~ s/(.)(.)/$2$1/g;
}

my @tests = map { unpack "h*", $_ }
    qw(ciao a tutti quanti 43242 432423423423 kjlf94324 dfas8902342);

cmpthese (-1, {
        PackUnpack => sub { PackUnpack($_) foreach (@tests) },
        SubstRegex => sub { SubstRegex($_) foreach (@tests) },
    }
);
__END__
              Rate SubstRegex PackUnpack
SubstRegex  3079/s         --       -79%
PackUnpack 14463/s       370%         --
[download]

This is no surprise, of course - and your solution remains more useful for the general case.

Update: SubstRegex returns $v as per Roy Johnson's suggestion.