mp++ is a C++11 header-only library for multiprecision arithmetic, currently supporting integers, rationals and quadruple-precision floats. Based on well-known libraries such as GMP, MPFR, and others, mp++ has two main objectives:
- to maximise performance for small integers and rationals,
- to provide a modern, consistent and unified C++ interface to several lower-level multiprecision libraries.
mp++ adopts various techniques to improve performance on small multiprecision integers and rationals. In particular, a small buffer optimisation and custom implementations of basic mathematical primitives are instrumental in achieving a performance increase, with respect to GMP and other libraries, which can be substantial (see the benchmarks section). The price to pay is a small overhead when operating on large integers.
mp++ is a spinoff of the Piranha library. It was created to cater to the requirements of computer algebra systems, which typically need to be able to manipulate arbitrarily-large numbers but which, in practice, often end up storing many small values (e.g., as coefficients in a polynomial or entries in a matrix).
mp++ is written in modern C++, and it requires a C++11-capable compiler. Currently it is continuously tested on the following setups:
- GCC 4.8 and later versions on GNU/Linux 32/64-bit,
- Clang 3.8 and later versions on GNU/Linux 64-bit,
- MSVC 2015 on Windows 32/64-bit,
- Clang 4 on Windows 32/64-bit (with the
clang-cldriver for MSVC), - MinGW GCC 6 on Windows 64-bit,
- Clang on OSX 64-bit (Xcode 6.4 and later).
The Intel compiler ICC 17 on GNU/Linux 64-bit is tested occasionally.