numfi 0.2.0

a numpy.ndarray subclass that does fixed-point arithmetic

numfi

numfi is a numpy.ndarray subclass that does fixed-point arithmetic.

Feature:

• Automatically perform fixed-point arithmetic through overloaded operators
• Maximum compatibility with numpy and other library, just like a normal numpy.ndarray
• Optimized calculation speed by minimizing quantization as much as possible

Install

Prerequisite: python3 and numpy

currently numfi is not ready for pypi, so you can clone this repo and build/install locally by:

python setup.py bdist_wheel

python -m pip install numfi --upgrade --force-reinstall --find-links=./dist 

or you can just copy numfi.py and use as what you want, after all it's only 200 lines of code in one file

Quick start

from numfi import numfi

# numfi(array, signed, bits_word, bits_frac, rounding='round', overflow='wrap')
x = numfi([1,2,3],1,16,8) 

# any arithmetic operation with numfi will return a numfi object with proper precision and value
y = x + 1 
z = x * 2
w = x / (0.1 + x)
...

Document

Details can be found here: https://numfi.readthedocs.io/en/latest/?

License

The project is licensed under the MIT license.