Metadata-Version: 2.0
Name: fpq
Version: 0.6.2
Summary: This package provides modules for manipulating floating point numbers quantization using NumPy.
Home-page: https://github.com/Hasenpfote/fpq
Author: Hasenpfote
Author-email: Hasenpfote36@gmail.com
License: UNKNOWN
Description-Content-Type: UNKNOWN
Keywords: quantization,floating-point,vector,quaternion
Platform: UNKNOWN
Classifier: Programming Language :: Python
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.3
Classifier: Programming Language :: Python :: 3.4
Classifier: Programming Language :: Python :: 3.5
Classifier: Programming Language :: Python :: 3.6
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Classifier: Development Status :: 3 - Alpha
Classifier: Environment :: Other Environment
Classifier: Intended Audience :: Developers
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Requires-Python: >=3.3
Requires-Dist: numpy
Provides-Extra: doc
Requires-Dist: sphinx; extra == 'doc'
Provides-Extra: test
Requires-Dist: coverage; extra == 'test'
Requires-Dist: nose; extra == 'test'

|License| |Build Status|

fpq
===

About
-----

This package provides modules for manipulating floating point numbers
quantization using NumPy.

Feature
-------

-  Supports multidimensional arrays.
-  Supports encoding and decoding between 64/32/16-bits floating point
   numbers and N-bits unsigned normalized integers.
-  Supports encoding and decoding between 64/32/16-bits floating point
   numbers and N-bits signed normalized integers.
-  Supports encoding and decoding between 3d-vectors and N-bits unsigned
   integers.
-  Supports encoding and decoding between Quaternions and N-bits
   unsigned integers.

Compatibility
-------------

-  Python 3.3+

Installation
------------

::

    pip install fpq

Usage
-----

encoding and decoding between 32-bits floating point numbers and 5-bits
unsigned normalized integers.

.. code:: python

    >>> import numpy as np
    >>> from fpq.fp import *
    >>> fp = np.array([0., 0.25, 1.], dtype=np.float32)
    >>> enc = encode_fp_to_std_unorm(fp, dtype=np.uint8, nbits=5)
    >>> enc
    array([ 0,  8, 31], dtype=uint8)
    >>> dec = decode_std_unorm_to_fp(enc, dtype=np.float32, nbits=5)
    >>> dec
    array([0.       , 0.2580645, 1.       ], dtype=float32)

encoding and decoding between 32-bits floating point numbers and 5-bits
signed normalized integers.

.. code:: python

    >>> import numpy as np
    >>> from fpq.fp import *
    >>> fp = np.array([-1., -0.25, -0., 0., 0.25, 1.], dtype=np.float32)
    >>> enc = encode_fp_to_std_snorm(fp, dtype=np.uint8, nbits=5)
    >>> enc
    array([31,  9,  1,  0,  8, 30], dtype=uint8)
    >>> dec = decode_std_snorm_to_fp(enc, dtype=np.float32, nbits=5)
    >>> dec
    array([-1.        , -0.26666668, -0.        ,  0.        ,  0.26666668,
            1.        ], dtype=float32)
    >>> enc = encode_fp_to_ogl_snorm(fp, dtype=np.uint8, nbits=5)
    >>> enc
    array([17, 28,  0,  0,  4, 15], dtype=uint8)
    >>> dec = decode_ogl_snorm_to_fp(enc, dtype=np.float32, nbits=5)
    >>> dec
    array([-1.        , -0.26666668,  0.        ,  0.        ,  0.26666668,
            1.        ], dtype=float32)

encoding and decoding between 3d-vectors and 64-bits(2:20:20:22)
unsigned integers.

.. code:: python

    >>> import math
    >>> import random
    >>> import numpy as np
    >>> from fpq.vector import *
    >>> v = np.array([vec_random(norm=100.) for _ in range(3)], dtype=np.float64)
    >>> v
    array([[-54.70386501, -22.45578546, -52.18237577],
           [-85.46791152,  -5.69032986,   1.21334561],
           [ 16.02886205,   1.94634654, -30.35219431]])
    >>> enc = encode_vec_to_uint(v, dtype=np.uint64, nbits=20)
    >>> enc
    array([ 1313110064653969262,   306332797892602581, 11373476070061802081],
          dtype=uint64)
    >>> dec = decode_uint_to_vec(enc, dtype=np.float64, nbits=20)
    >>> dec
    array([[-54.69957531, -22.45404536, -52.17828412],
           [-85.4662517 ,  -5.69027392,   1.21334561],
           [ 16.02845595,   1.94632843, -30.3514349 ]])

encoding and decoding between Quaternions and 64-bits(2:20:20:20)
unsigned integers.

.. code:: python

    >>> import numpy as np
    >>> from fpq.quaternion import *
    >>> q = np.array([quat_random() for _ in range(3)], dtype=np.float64)
    >>> q
    array([[ 0.25679071, -0.15512517,  0.88804262,  0.34838917],
           [ 0.71399177,  0.05729705, -0.69728753, -0.02688697],
           [-0.66527338, -0.62596543, -0.40672262, -0.01246296]])
    >>> enc = encode_quat_to_uint(q, dtype=np.uint64)
    >>> enc
    array([2724532880236077588,   93422189206870975, 1020620101889574962],
          dtype=uint64)
    >>> dec = decode_uint_to_quat(enc, dtype=np.float64)
    >>> dec
    array([[ 0.25679011, -0.15512497,  0.88804308,  0.34838854],
           [ 0.71399243,  0.0572969 , -0.69728688, -0.02688637],
           [ 0.66527395,  0.62596484,  0.40672258,  0.01246335]])

Here are a few
`examples <https://github.com/Hasenpfote/fpq/tree/master/example>`__.

Documentation
-------------

For users, docs are now available at https://hasenpfote.github.io/fpq/.

References and links
--------------------

| `D3D: Data Conversion
  Rules <https://msdn.microsoft.com/en-us/library/windows/desktop/dd607323(v=vs.85).aspx>`__
| `OGL: Normalized
  Integer <https://www.khronos.org/opengl/wiki/Normalized_Integer>`__
| `Vulkan: Fixed-Point Data
  Conversions <http://vulkan-spec-chunked.ahcox.com/ch02s08.html>`__

License
-------

This software is released under the MIT License, see LICENSE.

.. |License| image:: https://img.shields.io/badge/license-MIT-brightgreen.svg
   :target: https://github.com/Hasenpfote/fpq/blob/master/LICENSE
.. |Build Status| image:: https://travis-ci.org/Hasenpfote/fpq.svg?branch=master
   :target: https://travis-ci.org/Hasenpfote/fpq


