Is there going to be issues with later CPU boards that don't have a TILER-like component to them without the rotator chip? Isn't solving this again one more thing to do for every CPU board in the future?
Well, I don't know anything about the chips or wirings, so I trust the project has better people to decide this, but I don't understand your argument. Possibly because I'm uniformed.
In case the OMAP can be made to produce rotated images without any performance or freedom issues, removing the rotator chip seems like:
- slightly lowers costs (BOM and maybe rotator driver development, assuming the rotator driver is not already complete, and the TILER driver does not cost more to develop)
- sligthly? lowers power budget
- potentially accelerates video ? (passing through one less chip?)
- simplifies design and so potentially reduces bug count
A future CPU board upgrade is not certain. If it happens and the new CPU has no TILER equivalent, it could integrate the rotator chip in the upgraded CPU board, couldn't it ? (maybe not, I have no idea, but I'd need a reason why not. Wouldn't it fit or something ? ).
So only those who need a rotator pay for it (in money, battery, heat, bugs, complexity, whatever). A future upgrade needing a rotator sounds like a risk with low probability and low impact (because can be solved by adding
a rotator to the upgraded CPU board). So I dont' see how mitigating it now compensates for duplicated functionality in the 1st generation design. Kind of premature optimization ?
Intuitively I can't see a reason for duplicating functionality. But until someone proves it can be done with the OMAP it is moot. I guess it is still not proven to be possible with the rotator yet ?
On the one hand one video seems to have it working, on the other I think I read drivers were not ready for the rotator, so maybe the video was with software rotation or something?
So very happy that both paths are kind of researched in parallel.
It looks like a reasonable justification for delaying a little to research the possibilities.
But I'm mostly pulling this out of my **** . I'm sure anyone who understand the electronics can think this by themselves. The only constructive thing I can say is "don't hurry, take the time you need to decide this".
Edit: I'm assuming the CPU board interface is intended as custom, not as a kind of standard like Luke does with EOMA-68. In that case I would understand the rotator belonged to the screen device, not the CPU board. But if CPU boards for pyras are not expected to be usable in other devices, then you are free to add the rotator to eventual CPU boards.
Edit2: No, wait. If it was a standard the standard might say something about whether screens are expected to be portrait or landscape and that might make it more reasonable to add a rotator chip to the cpu board or the display board. But it wouldn't have to say the screen needs to be portrait or that the rotator belongs to the device.