[QUOTE=Davioware] 3 seconds per frame is 0.33 fps, Vs 3 fps. That's not 100% faster; it's 10x as fast; 900% faster
Did you make sure to set collisions to bounding box as well in both c2 and cc?[/QUOTE]
Lol whoops, no. I really screwed up that one in multiple ways. :P Let me try that again.
To get the math right and a more accurate percentage I ran a second test, being more careful this time: 730 instances of both sprite and sprite 2, no rotation, just collision checks, none overlapping, using bounding boxes (the test would be incomparable using full collision, because it would wary wildly depending on sprite size and number of collision polys):
CC 55fps, 0.0181s per frame
C2 17 fps, 0.0588s per frame
I think that results in C2 being about 31% the speed of CC there.
[QUOTE=Davioware]4fps VS 9fps is over twice the speed, over 100% faster.[/QUOTE]
Technically that was wavering into 5 fps, so I just made it to about 100% - I should have been more accurate, it was 4 a bit more often than 5.
More accurate version:
CC 53 fps, 0.0188 per frame
C2 21 fps, 00476 per frame
C2 being about 40% the speed of CC.
The first test creating instances, 1.2s/1.8s results in c2 being two thirds the speed.
So it seems the result is higher - CC seems to be about 1.5-3.2x faster.
Unless I screwed up again. :P
[QUOTE=Davioware]Also, more work CANNOT be done per frame at low fps, assuming rendering always finishes first; It just can't spit the logic out fast enough. Rendering finishes in a fraction of the time, and the cpu is just chugging, trying to finish calculating the event sheet.[/QUOTE]
Of course it can - that's the entire result reason why it's a low FPS in the first place. I'm asking it to do more work than It can manage in a 60th of a second. If a processor could manage 60 calculations per second, at 60 fps it could manage 1 per frame, at 1 fps it could manage 60 per frame. Therefore, more work is done per frame at 1 fps. Regardless if it is the CPU or GPU that cannot keep up, either one has to wait for the other (though the CPU works one step ahead since the gpu renders the previous frame concurrently). In these tests, the GPU was waiting on the CPU.
Interestingly, even seeing the result is back where I estimated originally before these tests, I don't feel like I care. I guess it's because c2's speed is sufficient for what I want to do, and device speed and browser optimizations are only going to improve.
While I wouldn't mind that extra speed, I guess the reason I feel this way is because I'm not sure I would use it. I can understand why others would want it though.Arima2013-02-02 06:40:55