I am trying to re-implement vcsm_square example from RaspiStill to my project.
Everthing seems to work, I am able to work with GPU memory "directly" and don't use "slow" glReadPixels.
But I have found quite limiting condition : VCSM buffer dimensions must be a power of two. See:
https://github.com/raspberrypi/userland ... ent.c#L316
If is used camera in resolution 1280x960 then must be created VCSM buffer of size 2048x1024. Which must be later also
processed e.g.in OpenCV - but this buffer is almost twice as big.
Any ideas or suggestions for me?
Thank you very much!