Re: Proposed solution to the "predetermined states" problem
Wolfgang Corcoran-Mathe 23 May 2026 15:15 UTC
On 2026-05-23 09:43 -0400, Peter McGoron wrote:
> On 5/23/26 00:50, Shiro Kawai wrote:
> >
> > It appears that the implementation that produced the output you showed
> > doesn't run "warmup" cycles during initialization---does it?
> > Scrambled linear PNG needs to discard several dozen initial values
> > to reach the state where the number of 1's and 0's are
> > approximately the same
> > in the state [1]. Otoh, I think once it is warmed up, any seed
> > (except all 0's)
> > produces the same quality of sequence.
>
> You're right. (I just copied the xoshiro256++ code into a C file).
> An implementer might make the same mistake I did, however.
Preposterous; only a rank amateur could make such a mistake!
In completely unrelated news, I will be making a few small fixes to
the sample implementation shortly.
(Thank you, Shiro.)
--
Wolfgang Corcoran-Mathe <xxxxxx@sigwinch.xyz>