[Issue 6594] Xorshift as default generator

[d-bugmail at puremagic.com]

http://d.puremagic.com/issues/show_bug.cgi?id=6594



--- Comment #5 from Joseph Rushton Wakeling <joseph.wakeling at webdrake.net> 2013-07-08 06:09:36 PDT ---
(In reply to comment #4)
> I have to agree.
> Closed.

Just to note, for future reference -- the Xorshift generators have been
criticized by other authors on a number of grounds -- see e.g.:
http://dx.doi.org/10.1145/1113316.1113319
[PDF preprint copy:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.63.7497&rep=rep1&type=pdf
]

However, there are some successors which look promising:
http://arxiv.org/abs/1004.3115
[This work was originally published in an Australian journal but the preprint
copy seems to be the only one readily available.]

So, I think there is future scope for replacing MT with a faster, simpler and
high-quality algorithm.

I also recently came across some further work extending MT to work with SIMD
and GPU implementations, which might be worth looking into:
http://www.math.sci.hiroshima-u.ac.jp/~saito/articles/sfmt.pdf
http://arxiv.org/abs/1005.4973v3

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------