Some technical stuff

One of the side effects of the bugfix for Bay Trail devices is disabling of terrain vertex buffer/index buffer caching. I am worried about that, and I was wondering how much effect it will have on framerate consistency.

I made some measurements on Nexus 4 at 256 visibility range (i.e. heavy load). The table below contains standard deviations of frame times and percentages of frames above 100ms (i.e. stuttering).

	Uncached (1.29.2)		Cached (1.29.1)
Frames	stddev	>100ms %	stddev	>100ms %
0	0.37	1.6	0.124	0.4
250	0.26	0.4	0.434	1.6
500	0.236	0.4	0.468	0.4
750	0.322	1.2	0.371	0.8
1000	0.249	0.4	0.342	0.8
1250	0.273	0.4	0.397	0.8
1500	0.349	1.2	0.317	0.8
1750	0.417	1.2	0.507	0.8
2000	0.38	2	0.47	1.6
2250	0.31	0.8	0.42	2.4
2500	0.306	1.2	0.209	1.2
2750	0.222	0.8	0.405	1.6
3000	0.371	2.4	0.486	1.6
3250	0.402	2	0.321	0.8
3500	0.265	1.2	0.322	0.4
3750	0.293	2	0.363	1.2
Average	0.314	1.2	0.372	1.075

Each row is calculated from 250 frames, the entire table is 4000 frames.

As you can see, there’s not much difference at all. In the uncached scenario 1.2% of frames were above 100ms, while in the cached scenario 1.075%. I suspect this small difference is well within the measurement error. Especially that standard deviations point the other way, uncached is actually more consistent when it comes to frame length.

Sorry for boring you with this technical stuff :-) Should be releasing full version to Google Play soonish.