horace {l Wrote}:polycount isn't the only thing that matters for performance. often stuff like overdraw affects it more. it would be good to have some statistics which shows where the time gets spent in the engine, so that people don't start to optimize wrong bottlenecks.Varivar {l Wrote}:I know this problem, I have it too when playing with many ai carts or playing multiplayer mode. I can try to reduce the polycount, but I really hope the rendering performance will be improved. Is 15000 polygons acceptable for irrlicht? Just for comparison: Tux Tollway (the 0.6 version) still had more polygons than this track (around 30000) while it ran smoother on multiplayer mode (0.6).
@sj04736: i have also provided 512x512 versions of the skybox images. they would look much better. 256x256 isn't really enough for a skybox. thanks for converting the track!
I agree. Atm I think we have two issues:
- Irrlicht always renders the whole track, no culling (since it's apparently faster to render everything with vbo support than using the octree for culling).
- Due to the character animations karts are not really optimised anymore (we previously had display lists). We probably need some advise here - one option might be to split the 'karts' in the karts (without wheels could use vbo) and characters, which need animations.
I've started to look at what stats irrlicht provides, and will add output of that later on.
Note also that at least some of the binaries (Macs) are not optimised.
...
but couldn't you check if changes got done to the track, and if yes, make new screenshots from some points at the camera fly-over? only if the track gets started for the first time after the changes got made. i always found it a bit cumbersome to update my screenshots after changes of textures and other details of my tracks. you have to start supertuxkart, somehow achieve a nice camera perspective (which can be hard), take a screenshot, rework the screenshot in gimp since the aspect ratio doesn't fit,...
OK, I'll put it on the todo list. But probably only for post 0.7 - once I've worked out how the camera flight over might work, it should be possible to add one camera position for the screenshot.
Cheers,
Joerg