boostmultiprecision

Boost.Multiprecision is a component of the Boost libraries that provides facilities for arithmetic with numbers whose precision exceeds that of built-in types. It supports integer, floating-point, and rational arithmetic with arbitrarily large precision and is designed to integrate with the C++ type system and standard library.

The library defines several concrete backends, including cpp_int for arbitrary-precision integers, cpp_dec_float and cpp_bin_float for decimal and binary floating-point numbers with user-specified digit counts, and cpp_rational for exact rational arithmetic. A generic number<Backend, ExpressionTemplates> type allows creating custom types by combining existing backends.

Boost.Multiprecision is largely header-only; most functionality requires only the header files. Some backends can be backed by external libraries such as GMP/MPFR for improved performance and larger sizes; when available, these backends provide faster arithmetic and exact rounding semantics.

Key features include seamless operator overloading and interoperation with built-in types, streaming I/O, numeric limits, and support for Boost.Math-style numerical algorithms. The library focuses on correctness and portability, with attention to cross-platform reproducibility and deterministic behavior when using fixed precision.

Typical use cases include numerical simulations requiring high precision, verification of mathematical identities, and tests for algorithms dependent on exact or controlled-precision arithmetic. While powerful, multiprecision arithmetic incurs higher runtime and memory costs compared with native types.