Int128: Portable and Performant 128-bit integers

Boost.Int128 is a portable implementation of a signed, and an unsigned 128-bit integer and related functionality (i.e. <bit>, <iostream>, etc). Importantly, on all platforms, the sizeof the types provided in this library are exactly 128-bits.

The library is header-only, has no dependencies, and requires only C++14. When using C++20 or newer, the library can be consumed as a module import boost.int128.

128-bit integers are useful across many domains, but C++ provides no portable way to use them. GCC and Clang offer __int128 as a non-standard extension on 64-bit targets, but it lacks std::numeric_limits specializations, <iostream> support, and is absent entirely on MSVC. Multiprecision libraries can fill the gap, but typically at the cost of a larger sizeof and additional overhead (e.g., Boost.Multiprecision always has an extra word). Boost.Int128 solves this by providing types that are exactly 128-bits on every platform. Operation implementations rely on compiler intrinsic where available for native performance, and optimized software implementations elsewhere. The types provided by the library also have native support being running on GPU using CUDA, along with many of the functions.

Boost.Int128 is tested natively on Ubuntu (x86_64, x86_32, s390x, aarch64, ARM32v7), macOS (x86_64, and Apple Silicon), and Windows (x86_64, x86_32, and ARM64); as well as emulated PPC64LE using QEMU with the following compilers:

Tested on Github Actions and Drone. Coverage can be found on Codecov.