Without delving into the internals of Math::BigFloat, I don't see any way to speed this up. Perhaps you could try a different approach? A different algorythm maybe?

Or take the big leap and write this routine in C and then link it in. Inline::CPP may help.

/\/\averick
perl -l -e "eval pack('h*','072796e6470272f2c5f2c5166756279636b672');"