On Apr 22, 2012, at 8:47 AM, Jeffrey Green <[email protected]> wrote: > I'm assuming the proprietary stuff that is the hurdle here is a ROM based > boot sequence (and language). I would think that the ARM processor > architecture is standard. If so about the ROM, is the general public > completely in the dark about it?
There are is a binary containing the code that runs on the GPU; this is loaded by the boot loader on the GPU. Then there are libGLES.so and friends that link with your openGL ES Linux app. Not much different from what NVidia does. The "kernel" is also booted by the GPU. In some ways this is simpler. To allow kernel choice at run time, the GPU loaded "kernel" could be 9load or grub or something. Linux driver to talk to the GPU is open source & CPU-GPU message types to implement a dumb frame buffer are documented so getting up /dev/draw shouldn't be hard. There is a schematic for the board and a separate spec for peripherals - GPIO, UART, PWM, etc etc. So a port should be no worse than to other ARM based SBCs. And once a port is done, you can have a ready to run SDHC card - no more fighting with installing plan9!
