On 11/29/24 09:18, Bald Eagle wrote:
> I had something bouncing around in the back of my head this morning, and since
> this thread concerns the PRNG, I'll drop this here.
> 
> If you look at my issue here:
> https://news.povray.org/web.650c19af5bb87ec31f9dae3025979125%40news.povray.org
> 
> this exposed what I believe to be a long-standing non-uniformity in the returned
> values from POV-Ray's built-in PRNG.
> 
> Have we taken a deep look at the values that the PRNG returns?
> 
> I was examining this from a color_map point of view way back, and I think it
> would be good to have something uniform/linear unless otherwise specified.
> 
> - BW
> 

Due the noise_generator changes in yuqk, I've been working on 
documentation which I think will help. See the attached text file.

I believe the noise you were trying to use is a smoothed noise which in 
addition to not being locally random at all, roughly follows a normal / 
Gaussian centered distribution over all values returned.

The yuqk fork has implemented a few different methods for uniform random 
noise in a range beyond what we can do with seed() and random() at parse 
time.

Bill P.

Post a reply to this message

Attachments:
Download 'ip_noise_generator.txt' (10 KB)