Implement [ Variable ] Precision Floating Point [ Arithmetic ] in CAS Mode

[parisse]

Giac/Xcas has both, based on MPFR and MPFI. You need to have both, because interval arithmetic gives most of the time intervals that are overly pessimistic. Indeed rounding errors are normally independant, which means if you make say n additions, error is most of the time of order sqrt(n)*epsilon, while interval arithmetic will return an interval of size n*epsilon.