Arbitrary-precision correctly-rounded floating-point arithmetic, via MPFR.
