Perl is really handy for text manipulation. That doesn't preclude that manipulation being CPU-intensive - it often is. For text manipulation tasks, regardless of any IO, Perl can generally easily keep up with compiled languages because the CPU-intensive work is actually done in compiled C (the Perl interpreter is compiled C). If you are using regular expressions, map, grep, index, substr and all the other Perl goodness then you really are not at a speed disadvantage compared with compiled languages.
On the other hand there are good hash implementations for C++ - a good tool is a good tool in most languages. They aren't as much fun to use as Perl's hashes because strict typing means Perlish techniques are cumbersome, but they are there when you need em.
Perl's payment curve coincides with its learning curve.