Performance issue

I have this small system. But this system takes 0.100ms of time every frame. There are only 2 entities that match query. I do profiling wrong, or entities package without burst are so slow? Or math methods are slow? What is wrong with this code?
0.1ms is too much for this simple code. With only 60 of such small systems I will spent all of frame budged. (6ms for scripts and 5ms for rendering for 90fps on oculus quest 2).
Unity 2020, entities 0.50, Android (Oculus Quest 2).

 

at the moment it is known, that ECS is still quite "slow" with a few entities, but that should change (1.0?)
In the editor everything runs slower.
Is there a reason why you run without burst? Burst is precisely there so that it gets faster and at first glance I don't see why the code can't run without burst.

 

First of all, use Burst! Entities without Burst is not that great.
I don't see any managed types so I don't understand why you use WithoutBurst.

Second, it's important to turn off Safety checks in Jobs->Burst. Also turn off Leak Detection. With these turned off you have nearly the same timings as in a build.

 

Please don't use screenshots of code. It's low effort.

 

This system takes 0.100ms in build on Android. So I think Leak Detection and Safety checks are disabled in build. I'll check this system with burst, but it still is very strange that without burst it takes so many time.

 

You should be using burst but .1ms is insanely slow for this calculation for 2 entities, something else is up.

Do you see it taking .1ms on the worker in the profiler? The entity debugger can be somewhat misleading since it will accumulate time spent in other jobs at sync points and report that time on whichever job the sync point happens.

 

This number is from profiler, not from entity debugger.

 

There's no point of not using Burst whenever you can, unless you're debugging.

 

Why you're all pointing on not using Burst? We can agree that it's faster with burst, but this execution time is still very high even without it.

 

It's not slow for Mono C#.
There's quite a bit of hefty arithmetics involved. Many of which Burst can take advantage of. There's also overhead that's not apparent like the EntityQuery. But really, most is the heavy mathematics being used. You're really underestimating unoptimized matrix multiplications.

If you're so inclined of this being slow. Why not just write the same loop of 2 entities out in classic MonoBehaviour and report the numbers. If you get it down factor 5-10x to 0.01 - 0.05 I'd say there's an issue but I doubt that.

edit: Just realized @miknios is not OP. Point still stands.