I had this problem come up a while back (in fact I think it was my question that prompted the thread on fwp about it . . .). At any rate, this was something I had lying around from it that benchmarks three solutions. You can decide for yourself if the first is cheating or if you really, really care about speed. :)

Also, note that these are returning the length of the common string not the string itself.

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

use Inline C => <<EOF;
int comlen(char *p, char *q) {
  int i = 0;
  while( *p && (*p++ == *q++) )
    i++;
  return i;
}
EOF

sub comlen_or {
  length((($_[0]^$_[1])=~m/^(\0+)/)[0]);
}

sub comlen_tr {
  my( $t );

  return ($t=$_[0]^$_[1])=~ y/\0/\0/;
}

$a = "abcdefghijk";
$b = "abcdefg";

cmpthese( shift || 2_000_000,
           {
            inline_c => sub { comlen( $a, $b ) },
            comlen_or => sub { comlen_or( $a, $b ) },
            comlen_tr => sub { comlen_tr( $a, $b ) },
           },
         );

exit 0;
[download]