[Skeime]

Wouldn't it also be much better to evaluate the Taylor polynomials using Horner's method, instead? (Maybe C++ can do this automatically, but given that there might be rounding differences, it probably won't.)