Feature Request Performance improvement for massive amount of colliders

@JosepMariaPujol Not really sure but is that currently have something called incremental update bounding volume hierarchy or something similar that only that will change so basically those static collider that will never will be excluded from bounding volume hierarchy update since most of the colliders are static.

 

If I understand correctly, looks like the current implementation is static and dynamic physics collider are splitted as separate chunk to process. What if we combine both of them into same single chunk and do incremental bounding volume hierarchy update for the physics collider that actually moving i.e. change position/rotation. Looks like it will much more faster that we process all the physics colliders in one single chunk no matter it's static or dynamic physics collider?

 

Currently due to new transform systems statics will rebuild every frame due to LocalToWorld change filter triggering. I'm sure this is just an oversight due to transform V2 reworks and it'll eventually be solved.

I believe you can fix this I believe by changing this line

Code (CSharp):

1. chunk.DidChange(ref LocalToWorldType, m_LastSystemVersion) ||

To something like,

Code (CSharp):

1. (!chunk.Has(ref LocalTransformType) && chunk.DidChange(ref LocalToWorldType, m_LastSystemVersion)) ||

To only do the local to world change filter check if it oesn't have local transform (in which case it'll rely on localtransform type change filter instead)

Not thoroughly tested, I just randomly looked into why statics were updating every frame a few weeks ago.

 

@JosepMariaPujol Any plan to implement the improvement suggested at above?

 

Yeah it was fixed in one of the earlier previews

From discord,

https://discord.com/channels/489222168727519232/1072222747192721589/1076186844892508251

 

Hi. Any new update for this feature request?