I did notice though that unless you need precision exceeding 30+ decimal places that perl scalars (in 5.8 at least) will give the same results and are much much faster.

You'll only get 30+ decimal digits if you're using some special hardware and a special version of perl to take advantage of it (Sparc-V9 hardware has 128bit quad floating point numbers). If you're using IEEE754 64bit doubles (like on Intel hardware) you'll only get about 15 significant decimal digits.

perl -e 'printf "%.30f\n",(1/3)'