keen linden
Some physics query tests, primarily using OverlapSphereNonAlloc to detect colliders.
https://docs.unity3d.com/ScriptReference/Physics.OverlapSphereNonAlloc.html
Some Info:
• Program and testing all done in Unity's editor environment using a basic URP templated build and scene.
• Each graph is an accumulation of data throughout 100 tests where each sequence has a larger collider capacity to check against; upwards to 10k colliders at the end of the testing cycle.
• I've reduced as much overhead in the scene as possible, and for each test I yield between instantiations of colliders to prevent any hiccups upon executing a query. I'm also running these tests on a decently built (eh, good enough) computer using a Window's OS.
• I've selectively picked the least chaotic tests results to share as larger sample sizes do produce some hiccups which made these sample plotting graphs less compact than I liked.
Some findings throughout my time evaluating this data:
• Querying OverlapSphere doesn't have that bad of a cost, but there's no doubt that more colliders in the vicinity equates to more time it takes to fully evaluate.
• The cost of finding object colliders seems to be somewhat linear (at least with my sample size) but hiccups become somewhat more noticeable with larger amount of colliders.
• Layer masks and filtering colliders does help runtime performance by filtering out unwanted colliders... but, there is still some noticeable cost relative to filtering out those colliders.
• Square magnitude vs distance check comparisons; they seem very comparable so might as well just get the info you need. I think there's some discussion about ARM processors having trouble with sqr operations, but that's something my mate probably knows more about.
~
