|
|
Пишет nancygold (![[info]](http://lj.rossia.org/img/userinfo.gif){kind=small} nancygold)@ 2024-08-29 00:21:00 |
 |
|
 |
|
|
|
|
|
|
 |
|
 |
| Настроение: |  annoyed |
| Музыка: | https://www.youtube.com/watch?v=VKpTk3pAQCk |
| Entry tags: | computing |
World Generation Algorithms
World generation algorithms I tried and discarded, in the order of appearance:
* Perlin noise.
* Hierarchical corner matched tiles.
* Rule based cellular automata from emitters.
* Subtracted midpoint displacement fractals.
* Same fractals but stencil controlled.
* DLA.
* Union of river fractals constraining the height map. Reverse DLA.
Reverse DLA is nice since it guarantees to give proper rivers.
Compared to distributary nonsense produced by other methods.
Perlin noise is especially prone to produce circular river flow.
Gradient smoothing fixes that a bit, but not completely.
I think Perlin noise is the worst generation algorithm.
It is a huge pitfall.
Even voronois could be helpful at the plannin scale, if you don't trust
cellular automata. But Perlin noise is just harmful and leads astray.
Never trust normies, who promote perlin noise, when the DLA family is so
simple, efficient and gives perfect result instantly.
They are like the toxic bros who suggest you to man-up, take testosterone and
do gym, instead of understanding that masculinity is the root of all evil.
