On 27/08/16 03:01, Matthijs van Duin wrote:
On 25 August 2016 at 15:31, Tomi Valkeinen tomi.valkeinen@ti.com wrote:
On 25/08/16 16:22, Matthijs van Duin wrote:
Most convincingly in the second NOTE of 10.2.4.6.5 DISPC Predecimation
Well, it's not talking about TILER there. The macro-pixel size is clear, but I was hoping to see a "if you use TILER, you need to use this and that". Or a mention that the macro-pixel size maps directly to the tiler unit sizes (or whatever they are called).
I agree, but this still seems like the best guess. I think I've also seen other hints toward 32-bit but I'm really not sure anymore where.
I can't get the YUYV rotation to work correctly, but I get better results with 16-bit TILER unit.
I did get NV12 working, I think (I didn't do too much testing). Attached quick kernel patches for this. The first one makes me think that this has never worked, and the dispc code for TILER rotation is probably quite broken (as seen with YUYV).
Tomi