libdecimal

Deterministic decimal arithmetic for C++

libdecimal is a header-only C++ library for IEEE 754-2008 decimal floating-point arithmetic, built on Intel's LIBBID reference implementation. It wraps the BID (binary-integer-decimal) encoding in a typed, constexpr-friendly interface with exact base-10 semantics and deterministic cross-platform results. Decimal32 / Decimal64 / Decimal128 are supported, together with transcendental functions, string and binary-float conversions, and full decomposition into sign, significand, and exponent — plus direct access to the underlying BID bit patterns for interop and persistence. Lightweight fixed-point, scaled-integer, and BCD representations are also included.

Layer	Role
libdecimal	Header-only C++ decimal arithmetic interface
Intel LIBBID	Vendored IEEE 754 decimal (BID) reference implementation

Supported platforms

OS / Compiler	GCC 13	GCC 14	GCC 15	Clang 17	Clang 18	Clang 19	Clang 20	Apple Clang	MSVC
Ubuntu 22.04
Ubuntu 24.04
macOS 14
Windows

macOS is built for x86_64 and exercised under Rosetta 2: Intel's float128 (DPML) sources are x86_64-only, so libdecimal targets amd64 / x86_64 on every platform.

Release packages

Native installers are attached to each tagged release. All packages are x86_64 / amd64 (Intel's DPML is x86_64-only).

Platform	Package
Debian / Ubuntu (amd64)	libdecimal-2.2.3-Linux-amd64.deb
Red Hat / Fedora (x86_64)	libdecimal-2.2.3-Linux-x86_64.rpm
macOS (x86_64)	libdecimal-2.2.3-macOS-x86_64.pkg
Windows (x64)	libdecimal-2.2.3-Windows-x64.exe

Download them from the Releases page. Each installer ships the headers, the Intel LIBBID static library, and the libdecimal CMake package files.

Requirements

Requirement	Minimum	Tested ceiling
C++ standard	C++23	C++23
CMake	3.22	—
Toolchains	GCC 13 · Clang 18 · Apple Clang · MSVC 19.4x	GCC 14 · Clang 20

Quick start

find_package(libdecimal REQUIRED)
add_executable(app main.cpp)
target_link_libraries(app PRIVATE libdecimal::libdecimal)

#include <decimal/bid.hpp>
#include <iostream>

int main() {
    using namespace math::literals;

    math::decimal_t<uint64_t> price(12345, -2);       // 123.45
    math::decimal_t<uint64_t> fee_rate = 0.0025_dec;  // 0.25 %

    auto fee   = price * fee_rate;
    auto total = price + fee;

    std::cout << "price: " << price << "  fee: " << fee << "  total: " << total << '\n';
    std::cout << "exp(price): " << math::exp(price) << '\n';   // transcendentals on the BID path
}

Building from source

cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build --parallel
sudo cmake --install build

The only build artifact is the vendored Intel LIBBID static library (libbid_static); everything libdecimal adds is header-only.

Decimal representations

The examples/ directory contains standalone programs demonstrating each representation:

File	Representation
examples/bid_example.cpp	IEEE 754 BID via Intel LIBBID
examples/fixed_example.cpp	Compile-time fixed-point
examples/bcd_example.cpp	Binary-coded decimal
examples/scaled_example.cpp	Lightweight scaled-integer pair

cmake -S . -B build -Dlibdecimal_BUILD_EXAMPLES=ON
cmake --build build --parallel
./build/bid_example

Development

cmake -S . -B build -DCMAKE_BUILD_TYPE=Debug -Dlibdecimal_BUILD_TESTS=ON  # configure (debug + tests)
cmake --build build --parallel                                           # build everything
ctest --test-dir build --output-on-failure                               # run the test suite
cmake --build build --target docs_serve                                  # serve docs @ http://127.0.0.1:9000/

Attribution

libdecimal builds on Intel's LIBBID implementation of IEEE 754 decimal arithmetic. The heavy lifting — the arithmetic itself — comes from the work of Marius Cornea, John Harrison, Cristina Anderson, and Evgeny Gvozdev. The underlying model traces back to Mike Cowlishaw and the IEEE 754 standard.

The template/macro layer that makes it usable in modern C++, along with the full decomposition of BID into sign, significand, and exponent, is the contribution of this project.