Performance Query

[charlie_says]

I've got a really odd problem with performance, which is not something I've had with PIXI before... And it appears to be to do with rendering.

What I have is 15000 sprites, which are initially pixel sized, from that100 of these are updated every frame (so that could be size, position, tint), doing this gets me about 7fps. If I do a smaller test say 2000 sprites I'll get 30fps, but I'm still only updating 100 of them per frame - so, it seems to me to be the renderer (also, just creating the image, and not doing anything on the update performs the same way.)

So, the question is: what can I do to boost the performance here - I've seen plenty of pixi tests like bunnymark, which perform loads better and all those sprites move every frame.

It did seem that I could perhaps use a particle container which might help.

I could perhaps break the image up into sections then specifically render the sections in turns.

But, I'm keen to hear of any other ideas I could work with.

 

[charlie_says]

I thought about this a little, and realised there were some set up issues (mainly that my sprite was in a container for historic reasons) once I started working with the sprite directly I was able to achieve about a 3x performance improvement on both tests.

My project isn't performance driven, so, if it works at 15fps it's ok, it just makes the controls a little unresponsive, but anything lower and they're not really useable.

I still don't really understand why when I'm only updating 100 of the sprites I get such different speeds between the 2 tests.

Any performance tips would still be gratefully received!

 

Edited December 15, 2020 by charlie_says

[ivan.popelyshev]

> I still don't really understand why when I'm only updating 100 of the sprites I get such different speeds between the 2 tests.

ParticleContainer?

[charlie_says]

Updated the project to particle container, large test works at 30fps smaller one works at 60fps - both perfectly acceptable for what I'm doing.

Thanks!