Open lukaschod opened 1 year ago
Also, I did not dig deep enough to find out why these differences surfaces as from the surface level sfloat and float should do similar math, where sfloat simply would have no platform/processor specific differences. My guess maybe there is a bug in sfloat that causes these problems.
Truly indeed!This project is just a total bull-shit
Thanks for posting this! Off to another rabbit-hole somewhere else :)
Please reply if you can recommend an alternative for device and platform-agnostic (pc, linux, osx, ios, android) fully deterministic physics. Personally I just need basic physics like ray, cube, sphere, cylinder, collision terrain vs each other with sweeping (delta movement) support and resolving the final position. Currently considering bullet physics in a unity plugin as the top option. Evaluated bepuPhysics (C# incompatibilities), and box2d (not 3d).
Thanks for posting this! Off to another rabbit-hole somewhere else :)
Please reply if you can recommend an alternative for device and platform-agnostic (pc, linux, osx, ios, android) fully deterministic physics. Personally I just need basic physics like ray, cube, sphere, cylinder, collision terrain vs each other with sweeping (delta movement) support and resolving the final position. Currently considering bullet physics in a unity plugin as the top option. Evaluated bepuPhysics (C# incompatibilities), and box2d (not 3d).
I've used this https://github.com/aaa719717747/TrueSyncExample
which was a light version of https://www.photonengine.com/
I tried playing around for few days with these packages, sadly I can confirm it is far from production ready.
Cylinder/Mesh colliders are very unstable. Also tried the same setup with float version and issues was gone.
Frictions are quite off, as objects loses the force almost 2x faster than in float version.
sfloat:
float:
P.S. Yes the scenarios here shown are not exact the same for comparison. But it is quite obvious then u play around in the scene and in float version everything slides completely differently.
In any case, it is cool package modification and nice attempt for deterministic physics. However, if you dont plan using only boxes and sphere likes in example, you wont find this package doing what you want.