Threw this up cause its what I do almost any new platform I explore ;)
I'm in the process of converting my GLES2.0 framework away from GLFW and back to Xlib as I've had different issues with different embedded platforms using GLFW, oddly I've not seen these issues with my Xlib code... The Allwinner A20 board here is showing 128 moving "3d" objects falling on a (very) simple mesh, the shaders in use are very simple texture and single light point shading with ambient lighting. This is running on top of Debian (armhf) running LXDE. This isn't pushing a single thread to its limit on this platform and if I separated the dynamics and rendering into their own threads I could probably get 50% more performance (at a guess!) If you're interested in the Allwinner A20 be sure and checkout my previous two posts...