pyfixed 0.3.0

Arbitrary-precision fixed-point library for Python

pyfixed

pyfixed is an arbitrary-precision fixed-point arithmetics library for Python.
It currently is in early development.

pyfixed is licensed under the MIT License.

Rational

pyfixed aims to aid in simulation of fixed-point arithmetics and testing of software and hardware solutions.
It supports multiple configurations, with the goal of supporting a wide variety of hardware implementations.

pyfixed is currently not optimized, and has poor real-time performance.
While optimizations are planned in general, the project does not aim to achieve any performance goals.

Installation

Install latest:

pip3 install pyfixed

Build from source:

python -m build
pip3 install dist/*.whl

You can also use it directly by adding the root directory to your PYTHONPATH.

Features

• pyfixed.Fixed: arbitrary-precision fixed-point.
  • Native API.
  • Simple arithmetic and bitwise operations.
  • 10 rounding modes (see pyfixed.Fixed.FixedRounding).
  • Optional exceptions on overflow, underflow and invalid operations (e.g. divide by 0).
    • Sticky exception flags.
  • Support for Python, NumPy and mpmath numeric types (integers and floats).
• pyfixed.Complex: arbitrary-precision complex fixed-point.
• fixed_cmp: visual comparison between fixed-point, IEEE-754 and posits.

Planned Features

• Posit support (against softposit).
• More functions:
  • $e^z$, $ln(z)$ and their derivatives.
  • FFT.
  • Matrix operations.
  • Other DSP-related operations.