|
|
|
|
|
by Impossible
2745 days ago
|
|
I did a ton of research because I was looking into creating a VR falling sand game (which someone actually made about a year ago https://www.oculus.com/experiences/rift/1337446473017832/). There are a couple of other 3D falling sand games (http://csiuo.com/Sand3d/, https://github.com/TheTomster/sandish, https://powdertoy.co.uk/Discussions/Thread/View.html%3FThrea...) but none of them are very good due to the factors you mentioned. UX is also a pretty big problem, as traditional mouse or touch controls don't work well in 3D, which is why I wanted to go the VR route, but something like a very dynamic Minecraft as you mentioned could work as well. Most falling sand game implementations are sequential, so parallelizing the simulation using a margolus neighborhood (https://en.wikipedia.org/wiki/Block_cellular_automaton) or double buffer could greatly improve performance. Most falling sand games tend to not be pure cellular automata, and some are implemented as particle systems with particle in cell grid collision instead of CA, but particle and fluid sims are easy to parallelize. I haven't seen a falling sand game (2D or 3D) that runs entirely on the GPU, and my goal was to do that, eventually ending up with some complex Claybook (https://www.claybookgame.com/) style engine. |
|