Conversation
|
Just as a note with this implementation since the zone renderer's vertex shader currently has no access to model height or model local positions, exact prior behavior matching is not possible unfortunately. Plan to explore other implementation options |
I have a branch which needs the dust blowing off which would provide this, essentially it adds support for SSBOs which allow individual models both static and dynamic to access unique model data. I'll be getting around to that once the latest update is out and stable. |
|
In the meantime though, we could get around the snapping by sampling the tile height instead. That would allow the heightBlending to be applied again, so the grass blades at the tile plane are fixed into position. |
|
Courtesy of @Jin-Jiyunsun, without the height blending the grass sliding around it very noticeable in areas with strong wind: java_BwDbTiFdu5.mp4 |
|
Also the legacy renderer is broken by this PR: |
Lmao I have a fix for this i did last night, I just forgot to push the update I guess, will push update when I get home |
data in vNormal.w alongside the existing height ratio and wind reach
|
Very gross implementation to get a bit closer to legacy, will almost certainly not go this path but wanted to share regardless packs the actual model origin (X, Z) into the 16-bit wind wind.3.mp4 |
|
Can you improve the granularity so that each individual object is affected by the wind? Right now they are in rather large blocks rather than each individual tile |
Unfortunately with this gross implementation i am limited with the number of bits i can pass, so by increasing granularity i have to give up how tall an object is affecting how much it moves and using a standard approximation instead which doesn't look right either as seen below wind.4.mp4This is part of the reason why i don't think this implementation method is necessarily the path forward but figure its important to share these experiments |
Ah ok, in that case this the one in this reply is probably good enough as a stop gap to bring it back properly, my two cents Perhaps we could implement it this way and move it back to experimental for the time being? |
|
Up to the group, happy either way but i would definitely suggest we move this to experimental if we do want to roll it out as you suggested This would definitely just be a stop gap until a better method/tools come around, I don't want to shoe horn us into this implementation permanently knowing its limits |
Adds wind back into the plugin by adding support for the new renderer
wind.2.mp4