This module is a thin Python wrapper over C library dcmt, created by Makoto Matsumoto and Takuji Nishimura, authors of Mersenne twister RNG algorithm. The original library provides functions for creating multiple independent RNGs (which can be used, for example, in parallel Monte-Carlo simulations).
The package contains intact dcmt
sources (except for example
folder) and exports some functions and structures from its API.
For licensing on original library and other information please see the link above, or src/dcmt/README
file and sources.
The version of the module corresponds to the version of original library, with the addition of the build number.
For example, the wrapper on top of dcmt 0.6.1
may have version like 0.6.1-10
.
Module documentation can be found here
- Fix issues with setuptools/distutils (one of them does not support Cython, other does not support Sphinx)
- Fixed bug when mt_range() return 'common' parameters dict with non-common parameters included;
- Fixed bug when DcmtRandomState.rand() returned array instead of Python float;
- Added raw integer randoms generation to DcmtRandomState;
- Added 'inplace' function to fill provided numpy array with randoms
- Using slightly optimized random number generation taken from
numpy
1.5.1; - Module is rewritten using Cython;
- Exporting analogues of Python standard
random.Random
andnumpy
'snumpy.random.mtrand.RandomState
; - Renamed
create_mts
tomt_range
, which now returns dictionary andnumpy
array instead ofctypes
structures;
- Shortened function names: create_mts(), init_mt()
- Removed get_random() and added rand(), which returns numpy array
- Added create_mts_stripped(), currently as a 'half-documented' feature
- Three basic functions: create_generators(), init_generator() and get_random() (returns integer)