FixMath

FixMath is a software library and specification for fixed-point mathematics designed to provide deterministic, portable numeric computations across diverse hardware. It defines fixed-point numeric types with configurable word length and fractional bits, along with arithmetic operations, conversions, and utilities to support predictable behavior in embedded and real-time systems.

Originating in the early 2010s as a response to nondeterministic floating-point behavior in simulations and control software, FixMath has grown into multi-language implementations and a reference suite for benchmarking.

Core features include saturating and wrapping arithmetic, explicit scale management, overflow detection, status flags, and SIMD-accelerated kernels where available. The design emphasizes deterministic results, explicit rounding modes, and clear API semantics to minimize precision errors.

Math functions include fixed-point implementations of sqrt, abs, min/max, and select transcendental functions (sin, cos, exp, log) using range reduction and iterative methods suited to fixed-point domains. The library offers fast paths for common ranges and provides configurable precision guarantees.

Implementation spans core libraries in C, C++, Rust, and Python bindings, with reference implementations and porting guidelines. It includes unit tests, a demo suite, and documentation to ease integration into new architectures.

Typical applications include embedded control loops, digital signal processing pipelines, and safety-critical systems where reproducibility matters, as well as numerical simulations that require deterministic results.

Critics note that fixed-point configuration can be complex and error-prone if precision is mis-specified, and performance depends on compiler and hardware support; coverage for all general-purpose math edge cases may be incomplete.

See also: Fixed-point arithmetic, Deterministic computing, Numerical libraries.