Profiler Camera.Render warnings

Hi there,

I am trying to improve frame rate of my game, so I profiled it on an iPad 4th gen. Now I’m seeing warnings under Camera.Render in JobAlloc.Overflow, but have no clue as to what could be wrong with it.
Can anyone shed some light on this? I’d like to be able to get rid of the warnings if they would lead to better rendering performance.

Hi!

JobAlloc.Overflow is an internal label which indicates slower memory allocation path during dynamic batching job preparation.
Unfortunately, there is not much to do about it now. We’ll remove warning from this label.

Hey alexey, can you give more info on this? What’s meant by “slower memory allocation path” and what effects will this have?

I’m seeing 4 of these warnings on my android build while I’m tackling performance issues - could this be the cause of my issues?

I’ve also got some of these. Is there anything I can do personally to improve the situation or avoid the slower path? Is it just a side effect of having a very large number of renderers in the scene? Is it an indicator that I should move towards some manual batching?

Bumping this for a probable answer.

Still showing up on 2018.3.0f2.

What is it?

Still showing up on 2021.1.3f1…

Can somebody respond?

This can basically happen whenever using our Memory Manager for native allocations, either by the engine’s native code or via UnsafeUtility.Malloc / NativeCollections with e.g. a Temp or TempJob allocator Label. Those are supposed to be quick allocations using an allocator dedicated to make these allocations quickly. But if the allocator is overburdened with allocations, or the allocation to big to be used with that particular allocator, it has to use a fallback allocator that is defined for every allocator. There might be multiple layers to the fallback but the main point is: it takes longer to get that memory than intended when the allocator was chosen, therefore you get a warning.

This could mean you can improve things by

  1. putting less memory into this allocator in other places,
  2. Making the allocation smaller (if you have control over it)
  3. Increasing the amount of memory available/bucket sizes for that allocator via Allocator Customization
  4. Report a bug if you think this is handled wrongly on the native engine code side