[Numpy-discussion] ANN: randomgen 1.19.0 release

Kevin Sheppard kevin.k.sheppard at gmail.com
Wed Jun 24 10:31:15 EDT 2020


randomgen 1.19.0 has been released with bug fixes and new features.

*New Features*

   - A helper class that lets users define custom bit generators in Python
   (slow) or numba (fast). This simplifies experimenting with alternative
   configurations. The UserBitGenerator can be used with
   numpy.random.Generator to produce variants from a wide range of
   distributions.
   https://bashtage.github.io/randomgen/custom-bit-generators.html
   - New bit generators:
      - PCG64DXSM - The cheap-multiplier variant that produces output
      before updating the state. This generator may become the default in NumPy
      in the future and is the 2.0 version of PCG64.
      - LCG128Mix - A 128-bit LCG with a settable multiplier, increment,
      and output function. This generator nests both PCG64 and PCG64DXSM as
      special cases. It also can act as a standard 128bit LCG/MCG.
      - EFIIX64 (x48 variant) - Chris Doty's stream cipher-like generator
      - SFC64 with settable counter increment which allows distinct streams
      to be produced from the same SeedSequence.
      - LXM - A generator that mixes the output of a 64 bit-LCG and a
      256bit Xorshift.
   - ExtendedGenerator contains methods for producing variants that are not
   included in numpy.random.Generator.

*Deprecations*
Both Generator and RandomState have been officially deprecated. The NumPy
versions are both better maintained and feature-rich.

*Other*
All bit generators have been tested output at least 1TB (many to 4TB) both
as single streams, interleaved and interleaved with 4 or 8196 other bit
generators. When interleaved, the additional generators were constructed
using both SeedSequence.spawn and jumped (when available).

https://bashtage.github.io/randomgen/testing.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20200624/3f85226c/attachment-0001.html>


More information about the NumPy-Discussion mailing list