MPFR provides a library for multiple-precision floating-point computation with correct rounding. The computation is both efficient and has a well-defined semantics. It copies the good ideas from the ANSI/IEEE-754 standard for double-precision floating-point arithmetic (53-bit mantissa).