[New-bugs-announce] [issue38767] Replace Mersenne Twister RNG with a PCG family algorithm

[Tilman Roeder]

New submission from Tilman Roeder <tilmroe at gmail.com>:

Currently, the `random.random()` function / the random module uses the Mersenne Twister algorithm for generating random numbers.

While this algorithm has acceptable statistical properties for most use-cases (and does feature a ridiculously large period), it is slow and very memory intensive when compared with algorithms from the PCG family (see http://www.pcg-random.org). PCG family algorithms also have much better statistical properties than the Mersenne Twister.

I would propose to replace the underlying generator in `random` with a suitable PCG family algorithm, while not changing any of the external interfaces of the module. I think that the change would make the module faster, better in terms of memory usage, and improve the statistical properties of Python's default random numbers.

I have not done anything in the direction in terms of writing any code, but if this sounds like something that would be sensible, I would be happy to work on the change and submit a PR.

Also, this is the first time I am contributing to Python/ writing an issue here, so apologies if this is not the correct place to make a suggestion like this.

----------
components: Extension Modules
messages: 356371
nosy: dyedgreen, mark.dickinson, rhettinger
priority: normal
severity: normal
status: open
title: Replace Mersenne Twister RNG with a PCG family algorithm
type: enhancement
versions: Python 3.9

_______________________________________
Python tracker <report at bugs.python.org>
<https://bugs.python.org/issue38767>
_______________________________________