Post

I've struggled with performance in my tower defense game as well, going to try out some of these optimizations! I've noticed using collision shapes on enemies is also very costly but necessary as enemies can get knocked back and need to collide with towers. Do you use collision on your enemies?

That's an amazing thread, I'll put some of these in use for my games, even if performance is pretty good for now, it's nice to keep those in mind for when I add more and more to it

📌

You should probably do a blog post with all this ! It’s very interesting and information worth keeping around instead of an ephemeral BS thread !

That's very weird, I have no issues on my side. But if that can help you, I used a unroll service to share you his great thread: skyview.social?url=https%3A...

Depends on the developer. I've never been a fan myself.

Not a big time OOP fan, but I don't read the thread this way: - different physics object type (still using inheritance) - cache strings - avoid unnecessary nav stuff - time slicing :( - move non physics code out of physics updates/callbacks All mostly independent of using OOP or not.

thanks Lukas!

📌 Thanks for sharing your findings!

Taking notes... Noticed similar things about navigation, some functions are really "heavy"

📌

This is so cool! Thanks for posting stuff like this, I find optimization in games to be really fascinating to see, especially in engines I'm not actively working in.

It's the opposite - RigidBody allows either code control or physics simulated movement, CharacterBody is code control only. Ideally, you could extend from PhysicsBody, but the engine doesn't allow this for some reason 🫠

The extra gains possibly only come into the picture when you have have an up/down and is_on_floor? 🤔

So did you store the function or the stringName?

+++ we replaced polling nav_agent.is_navigation_finished() with a local variable set by the navigation_finished signal. we looked at the source code of is_navigation_finished() and it turns out the function fully updates pathfinding and avoidance, which is super costly.

Wow those are some nice gains. Function flattening is particularly interesting, as I had no idea they had such a performance hit. Thread bookmarked.

I don't recall having an interaction with you? No idea where this is coming from, but it wasn't intended to be antagonistic.

If gdscript has no intermediate compilation step and is purely interpreted, I don’t see how it can ever be ready for the big leagues. It just won’t ever work out. I kinda assumed it would have some intermediate language compilation, like almost every other language nowadays.

amazing thread thanks for putting it together!

Not all heroes wear capes

THANKS! 📌