Thank you for the response Faatal. I made my own pool for testing as well. Not a huge fan of what Unity has. I noticed you had a few pools for things but one in specific stood out as missing so I did some trial and error.
Indeed, I prefer to directly change the vectors. I wondered why you sometimes do that and sometimes dont. Makes a lot of sense. Appreciate the info.
I did some experiments with transform pooling. You are instantiating them off the entity class when you spawn an entity. I could not get it stable with all of the entity types but falling trees and players were easy to achieve. At the point it instantiates, I had the main thread check for a pooled one, if none exist it ran it like normal but then started a job in the unity job system to pool an extra 25. This cut down the runtime and improved performance when spawning entity. I was able to pool them by their entity class name but ran in to very random crashing, particularly when trying to pool entity like zombies. Some things could not be stored at all such as the transform for falling blocks or snakes and animals. I know the typical thing is to store game objects but in this case, you are instantiating a transform which is slow. Honestly I am not sure on the limits of multi thread safety and unity api calls, even if run through the job system.
I tried to store the transform of entity after use instead of destroy the game object/transform. This does work but there are a lot of components attached to the game object within the transform. If you try to reuse the transform, the components will get doubled up from the entity spawn process and zombies with dismemberment become an issue.
As a side note I noticed you mentioned wanting to improve game performance to someone else on the forum. I have been working on an optimizer for the game. I have managed to get the players and non players in a dedicated server to update multi threaded for various portions of the game code. Offloaded work from the main thread to improve stability. Improved chunk sync for players. Increased the rate and stability of block collapse. Removed some bloat that was not needed since clients already send certain things to the dedi server. Vehicle response and general movement is smoother.
It appears to improve the single player client side performance as well. Tests are just starting on that.
Player capacity and zombie capacity are dramatically increased in dedicated servers. Rough numbers so far show 40-60 players and up to 300 zombies at a 60 fps target depending on hardware. If all goes well, I will try and get it out to the public before christmas. Would you have any interest in what I am doing? I dont want to step over my bounds or on any bodies feet.
