I need some help testing an issue with the cpu animation system. I’ve found that there is a sudden increase in animation time when the number of actors cross a certain (computer, actor dependent) threshold. For example, on my Core 2 Duo (2.8ghz) and using the following test file
The threshold can vary, but it’s generally about a dozen actors. Different actors have different trigger thresholds.
I would appreciate it, if you could run the test.py and post your results (number of actors that cause a sudden jump in animation time, cpu secs, OS) so that we can understand the issue better.
1 actor, 1200 fps
3 1100
5 1000
15 750
30 420
60 265
80 120
90 65
100 50
110 40
150 24
200 18
300 12
500 7
I am using my own build from the repository (code is from a few weeks ago). I have threading disable in my build, as well as components I didn’t need disabled.
I havent looked at the code, but I remember some detail that if the animation state does not need to get updated, panda would not recompute the state. Thus, if the frame rate for the animation is below the frame rate for rendering, you can slip a few frames in without doing any more animation. That would explain major changes in frame rate based on animation speed and other load on the renderer.