README.md

   1 # sfpy
   2 softfloat and softposit in Python
   3   * support for softfloat float16, float32, and float64
   4   * support for softposit posit8, quire8, posit16, quire16, posit32, and quire32
   5
   6 ## Installation
   7 On most linux distros with CPython 2.7, 3.4, 3.5, 3.6, or 3.7, sfpy should work out of the box:
   8
   9 ```
  10 pip install sfpy
  11 ```
  12
  13 Under the hood, sfpy uses Cython to create bindings for the softposit and softfloat C libraries.
  14 The building instructions are tested on Ubuntu 16.04 - for other platforms they may need some
  15 adaptation.
  16
  17 ## Demo
  18 ```
  19 >>> import sfpy
  20 >>> from sfpy import *
  21 >>> a, b = Float16(1.3), Float16(1.4)
  22 >>> a * b - a / b
  23 Float16(0.89208984375)
  24 >>> sfpy.float.flag_get_inexact()
  25 True
  26 >>> a += b
  27 >>> a
  28 Float16(2.69921875)
  29 >>>
  30 >>> x, y = Posit16(3.0), Posit16(3)
  31 >>> x
  32 Posit16(3.0)
  33 >>> x.bits
  34 22528
  35 >>> y
  36 Posit16(2.9802322387695312e-08)
  37 >>> y.bits
  38 3
  39 >>> x * y
  40 Posit16(8.940696716308594e-08)
  41 >>> acc = Posit16(0)
  42 >>> for i in range(10000):
  43 ...   acc = acc.fma(x, y)
  44 ...
  45 >>> acc
  46 Posit16(1.9073486328125e-06)
  47 >>> acc.bits
  48 24
  49 >>> q = Quire16(0)
  50 >>> for i in range(10000):
  51 ...   q.iqma(x, y)
  52 ...
  53 >>> q
  54 Quire16(0.00089263916015625)
  55 >>> q.bits
  56 64424509440000
  57 >>> q.to_posit()
  58 Posit16(0.00089263916015625)
  59 >>> q.to_posit().bits
  60 490
  61 ```
  62
  63 ## Building
  64 See [BUILDING](https://github.com/billzorn/sfpy/blob/master/BUILDING.md).